d/freebsd-dev
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
2b519b1707
freebsd-dev/sys/fs/nfsserver/nfs_nfsdstate.c

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

8948 lines
252 KiB
C
Raw Normal View History

Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*-
sys/fs: further adoption of SPDX licensing ID tags. Mainly focus on files that use BSD 2-Clause license, however the tool I was using misidentified many licenses so this was mostly a manual - error prone - task. The Software Package Data Exchange (SPDX) group provides a specification to make it easier for automated tools to detect and summarize well known opensource licenses. We are gradually adopting the specification, noting that the tags are considered only advisory and do not, in any way, superceed or replace the license texts.
2017-11-27 15:15:37 +00:00
* SPDX-License-Identifier: BSD-2-Clause-FreeBSD
*
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
* Copyright (c) 2009 Rick Macklem, University of Guelph
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
Add support for INET6 addresses to the kernel code that dumps open/lock state. PR#223036 reported that INET6 callback addresses were not printed by nfsdumpstate(8). This kernel patch adds INET6 addresses to the dump structure, so that nfsdumpstate(8) can print them out, post-r346190. The patch also includes the addition of #ifdef INET, INET6 as requested by bz@. PR: 223036 Reviewed by: bz, rgrimes MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D19839
2019-04-13 22:00:09 +00:00
#include "opt_inet.h"
#include "opt_inet6.h"
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
#include <sys/extattr.h>
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
#include <fs/nfs/nfsport.h>
int nfsrv_issuedelegs = 0;
Change the default value for vfs.newnfs.enable_locallocks to 0 for the experimental NFS server, since local locking is known to be broken and the patch to fix it is still a work in progress. MFC after: 5 days
2010-02-14 00:18:32 +00:00
int nfsrv_dolocallocks = 0;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
struct nfsv4lock nfsv4rootfs_lock;
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
time_t nfsdev_time = 0;
int nfsrv_layouthashsize;
volatile int nfsrv_layoutcnt = 0;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
NFSD_VNET_DEFINE(struct nfsrv_stablefirst, nfsrv_stablefirst);
NFSD_VNET_DECLARE(int, nfsrv_numnfsd);
NFSD_VNET_DECLARE(struct nfsstatsv1 *, nfsstatsv1_p);
extern uint32_t nfs_srvmaxio;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
extern int nfsrv_lease;
extern struct timeval nfsboottime;
extern u_int32_t newnfs_true, newnfs_false;
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
extern struct mtx nfsrv_dslock_mtx;
extern struct mtx nfsrv_recalllock_mtx;
extern struct mtx nfsrv_dontlistlock_mtx;
Add the BindConnectiontoSession operation to the NFSv4.1 server. Under some fairly unusual circumstances, the Linux NFSv4.1 client is doing a BindConnectiontoSession operation for TCP connections. It is also used by the ESXi6.5 NFSv4.1 client. This patch adds this operation to the NFSv4.1 server. Reported by: andreas.nagy@frequentis.com Tested by: andreas.nagy@frequentis.com MFC after: 2 weeks
2018-06-01 19:47:41 +00:00
extern int nfsd_debuglevel;
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
extern u_int nfsrv_dsdirsize;
extern struct nfsdevicehead nfsrv_devidhead;
extern int nfsrv_doflexfile;
extern int nfsrv_maxpnfsmirror;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
NFSV4ROOTLOCKMUTEX;
NFSSTATESPINLOCK;
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
extern struct nfsdontlisthead nfsrv_dontlisthead;
extern volatile int nfsrv_devidcnt;
extern struct nfslayouthead nfsrv_recalllisthead;
Fix the kernel part of pnfsdscopymr() to handle holes in the file being copied. If a mirrored DS is being recovered that has a lot of large sparse files, pnfsdscopymr(8) would use a lot of space on the recovered mirror since it would write the "holes" in the file being mirrored. This patch adds code to check for a "hole" and skip doing the write. The check is done on a "per PNFSDS_COPYSIZ size block", which is currently 64K. I think that most file server file systems will be using a blocksize at least this large. If the file server is using a smaller blocksize and smaller holes need to be preserved, PNFSDS_COPYSIZ could be decreased. The block of 0s is malloc()d, since pnfsdcopymr(8) should be an infrequent occurrence.
2018-07-08 18:15:55 +00:00
extern char *nfsrv_zeropnfsdat;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
Make the size of the hash tables used by the NFSv4 server tunable. No appreciable change in performance was observed after increasing the sizes of these tables and then testing with a single client. However, there was an email that indicated high CPU overheads for a heavily loaded NFSv4 and it is hoped that increasing the sizes of the hash tables via these tunables might help. The tables remain the same size by default. Differential Revision: https://reviews.freebsd.org/D2596 MFC after: 2 weeks
2015-05-27 22:00:05 +00:00
SYSCTL_DECL(_vfs_nfsd);
int nfsrv_statehashsize = NFSSTATEHASHSIZE;
SYSCTL_INT(_vfs_nfsd, OID_AUTO, statehashsize, CTLFLAG_RDTUN,
&nfsrv_statehashsize, 0,
"Size of state hash table set via loader.conf");
int nfsrv_clienthashsize = NFSCLIENTHASHSIZE;
SYSCTL_INT(_vfs_nfsd, OID_AUTO, clienthashsize, CTLFLAG_RDTUN,
&nfsrv_clienthashsize, 0,
"Size of client hash table set via loader.conf");
int nfsrv_lockhashsize = NFSLOCKHASHSIZE;
SYSCTL_INT(_vfs_nfsd, OID_AUTO, fhhashsize, CTLFLAG_RDTUN,
&nfsrv_lockhashsize, 0,
"Size of file handle hash table set via loader.conf");
int nfsrv_sessionhashsize = NFSSESSIONHASHSIZE;
SYSCTL_INT(_vfs_nfsd, OID_AUTO, sessionhashsize, CTLFLAG_RDTUN,
&nfsrv_sessionhashsize, 0,
"Size of session hash table set via loader.conf");
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
int nfsrv_layouthighwater = NFSLAYOUTHIGHWATER;
SYSCTL_INT(_vfs_nfsd, OID_AUTO, layouthighwater, CTLFLAG_RDTUN,
&nfsrv_layouthighwater, 0,
"High water mark for number of layouts set via loader.conf");
Make the size of the hash tables used by the NFSv4 server tunable. No appreciable change in performance was observed after increasing the sizes of these tables and then testing with a single client. However, there was an email that indicated high CPU overheads for a heavily loaded NFSv4 and it is hoped that increasing the sizes of the hash tables via these tunables might help. The tables remain the same size by default. Differential Revision: https://reviews.freebsd.org/D2596 MFC after: 2 weeks
2015-05-27 22:00:05 +00:00
static int nfsrv_v4statelimit = NFSRV_V4STATELIMIT;
SYSCTL_INT(_vfs_nfsd, OID_AUTO, v4statelimit, CTLFLAG_RWTUN,
&nfsrv_v4statelimit, 0,
"High water limit for NFSv4 opens+locks+delegations");
A problem w.r.t. interoperation between the FreeBSD NFSv4.1 server with delegations enabled and the Linux NFSv4.1 client was reported in reviews.freebsd.org/D7891. I believe that the FreeBSD server behaviour conforms to the RFC and that the Linux client has a bug. Therefore, I do not think the proposed patch is appropriate. When nfsrv_writedelegifpos is non-zero, the FreeBSD server will issue a write delegation for a read open if possible. The Linux client then erroneously assumes that the credentials used for the read open can write the file. This patch reverses the default value for nfsrv_writedelegifpos to 0 so that the default behaviour is Linux compatible and adds a sysctl that can be used to set nfsrv_writedelegifpos. This change should only affect users that are mounting a FreeBSD server with delegations enabled (they are not enabled by default) with a Linux NFSv4.1 client mount. Reported by: fatih.acar@gandi.net Tested by: fatih.acar@gandi.net MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D7891
2016-10-20 23:53:16 +00:00
static int nfsrv_writedelegifpos = 0;
SYSCTL_INT(_vfs_nfsd, OID_AUTO, writedelegifpos, CTLFLAG_RW,
&nfsrv_writedelegifpos, 0,
"Issue a write delegation for read opens if possible");
Allow use of a write open stateid for reading in the NFSv4 server. The NFSv4 RFCs give a server the option of allowing the use of an open stateid for write access to be used for a Read operation. This patch enables this by default and adds a sysctl to disable it, for anyone who does not want this capability. Allowing this is particularily useful for a pNFS Data Server (DS), since they are not permitted to allow the use of special stateids. Discovered during recent testing of the pNFS server under development. MFC after: 2 weeks
2017-04-24 20:46:19 +00:00
static int nfsrv_allowreadforwriteopen = 1;
SYSCTL_INT(_vfs_nfsd, OID_AUTO, allowreadforwriteopen, CTLFLAG_RW,
&nfsrv_allowreadforwriteopen, 0,
"Allow Reads to be done with Write Access StateIDs");
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
int nfsrv_pnfsatime = 0;
SYSCTL_INT(_vfs_nfsd, OID_AUTO, pnfsstrictatime, CTLFLAG_RW,
&nfsrv_pnfsatime, 0,
"For pNFS service, do Getattr ops to keep atime up-to-date");
int nfsrv_flexlinuxhack = 0;
SYSCTL_INT(_vfs_nfsd, OID_AUTO, flexlinuxhack, CTLFLAG_RW,
&nfsrv_flexlinuxhack, 0,
"For Linux clients, hack around Flex File Layout bug");
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
Fix a bogus comment that mentions lru lists that don't exist. Reported by: zack.kirsch at isilon.com MFC after: 2 weeks
2010-07-13 22:44:14 +00:00
* Hash lists for nfs V4.
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
*/
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
NFSD_VNET_DEFINE(struct nfsclienthashhead *, nfsclienthash);
NFSD_VNET_DEFINE(struct nfslockhashhead *, nfslockhash);
NFSD_VNET_DEFINE(struct nfssessionhash *, nfssessionhash);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
struct nfslayouthash *nfslayouthash;
volatile int nfsrv_dontlistlen = 0;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
static u_int32_t nfsrv_openpluslock = 0, nfsrv_delegatecnt = 0;
static int nfsrv_returnoldstateid = 0, nfsrv_clients = 0;
static int nfsrv_clienthighwater = NFSRV_CLIENTHIGHWATER;
static int nfsrv_nogsscallback = 0;
Avoid the overhead of acquiring a lock in nfsrv_checkgetattr() when there are no write delegations issued. manu@ reported on the freebsd-current@ mailing list that there was a significant performance hit in nfsrv_checkgetattr() caused by the acquisition/release of a state lock, even when there were no write delegations issued. This patch add a count of outstanding issued write delegations to the NFSv4 server. This count allows nfsrv_checkgetattr() to return without acquiring any lock when the count is 0, avoiding the performance hit for the case where no write delegations are issued. Reported by: manu Reviewed by: kib MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D13327
2017-12-04 21:50:27 +00:00
static volatile int nfsrv_writedelegcnt = 0;
Add a counter to limit the number of disabled DSs for a mirrored pNFS MDS. This patch adds a counter that limits the number of disabled mirrored DSs to mirror level - 1. It also makes a small change that keeps a Write that has failed with EACCES when attempted by a client to a DS from disabling the DS. This patch only affects the pNFS server.
2018-06-22 00:55:39 +00:00
static int nfsrv_faildscnt;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
NFSD_VNET_DEFINE_STATIC(time_t, nfsrvboottime);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/* local functions */
static void nfsrv_dumpaclient(struct nfsclient *clp,
struct nfsd_dumpclients *dumpp);
static void nfsrv_freeopenowner(struct nfsstate *stp, int cansleep,
NFSPROC_T *p);
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
static int nfsrv_freeopen(struct nfsstate *stp, vnode_t vp, int cansleep,
NFSPROC_T *p);
static void nfsrv_freelockowner(struct nfsstate *stp, vnode_t vp, int cansleep,
NFSPROC_T *p);
static void nfsrv_freeallnfslocks(struct nfsstate *stp, vnode_t vp,
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
int cansleep, NFSPROC_T *p);
static void nfsrv_freenfslock(struct nfslock *lop);
static void nfsrv_freenfslockfile(struct nfslockfile *lfp);
static void nfsrv_freedeleg(struct nfsstate *);
static int nfsrv_getstate(struct nfsclient *clp, nfsv4stateid_t *stateidp,
u_int32_t flags, struct nfsstate **stpp);
static void nfsrv_getowner(struct nfsstatehead *hp, struct nfsstate *new_stp,
struct nfsstate **stpp);
static int nfsrv_getlockfh(vnode_t vp, u_short flags,
Remove an unnecessary level of indirection for an argument. This simplifies the code and should avoid the clang sparc port from generating an abort() call. Requested by: rdivacky Submitted by: jhb MFC after: 2 weeks
2014-04-23 23:13:46 +00:00
struct nfslockfile *new_lfp, fhandle_t *nfhp, NFSPROC_T *p);
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
static int nfsrv_getlockfile(u_short flags, struct nfslockfile **new_lfpp,
struct nfslockfile **lfpp, fhandle_t *nfhp, int lockit);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
static void nfsrv_insertlock(struct nfslock *new_lop,
struct nfslock *insert_lop, struct nfsstate *stp, struct nfslockfile *lfp);
static void nfsrv_updatelock(struct nfsstate *stp, struct nfslock **new_lopp,
struct nfslock **other_lopp, struct nfslockfile *lfp);
static int nfsrv_getipnumber(u_char *cp);
static int nfsrv_checkrestart(nfsquad_t clientid, u_int32_t flags,
nfsv4stateid_t *stateidp, int specialid);
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
static int nfsrv_checkgrace(struct nfsrv_descript *nd, struct nfsclient *clp,
u_int32_t flags);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
static int nfsrv_docallback(struct nfsclient *clp, int procnum,
nfsv4stateid_t *stateidp, int trunc, fhandle_t *fhp,
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
struct nfsvattr *nap, nfsattrbit_t *attrbitp, int laytype, NFSPROC_T *p);
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
static int nfsrv_cbcallargs(struct nfsrv_descript *nd, struct nfsclient *clp,
nfsd: fix session slot handling for failed callbacks When the NFSv4.1/4.2 server does a callback to a client on the back channel, it will use a session slot in the back channel session. If the back channel has failed, the callback will fail and, without this patch, the session slot will not be released. As more callbacks are attempted, all session slots can become busy and then the nfsd thread gets stuck waiting for a back channel session slot. This patch frees the session slot upon callback failure to avoid this problem. Without this patch, the problem can be avoided by leaving delegations disabled in the NFS server. MFC after: 2 weeks
2021-04-23 22:24:47 +00:00
uint32_t callback, int op, const char *optag, struct nfsdsession **sepp,
int *slotposp);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
static u_int32_t nfsrv_nextclientindex(void);
static u_int32_t nfsrv_nextstateindex(struct nfsclient *clp);
static void nfsrv_markstable(struct nfsclient *clp);
End grace for the NFSv4 server if all mounts do ReclaimComplete. The NFSv4 protocol requires that the server only allow reclaim of state and not issue any new open/lock state for a grace period after booting. The NFSv4.0 protocol required this grace period to be greater than the lease duration (over 2minutes). For NFSv4.1, the client tells the server that it has done reclaiming state by doing a ReclaimComplete operation. If all NFSv4 clients are NFSv4.1, the grace period can end once all the clients have done ReclaimComplete, shortening the time period considerably. This patch does this. If there are any NFSv4.0 mounts, the grace period will still be over 2minutes. This change is only an optimization and does not affect correct operation. Tested by: andreas.nagy@frequentis.com MFC after: 2 months
2018-05-15 20:28:50 +00:00
static void nfsrv_markreclaim(struct nfsclient *clp);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
static int nfsrv_checkstable(struct nfsclient *clp);
static int nfsrv_clientconflict(struct nfsclient *clp, int *haslockp, struct
vnode *vp, NFSPROC_T *p);
static int nfsrv_delegconflict(struct nfsstate *stp, int *haslockp,
NFSPROC_T *p, vnode_t vp);
static int nfsrv_cleandeleg(vnode_t vp, struct nfslockfile *lfp,
struct nfsclient *clp, int *haslockp, NFSPROC_T *p);
nfsd: Add support for the SP4_MACH_CRED case in ExchangeID Commit f4179ad46fa4 added support for operation bitmaps for NFSv4.1/4.2. This commit uses those to implement the SP4_MACH_CRED case for the NFSv4.1/4.2 ExchangeID operation since the Linux NFSv4.1/4.2 client is now using this for Kerberized mounts. The Linux Kerberized NFSv4.1/4.2 mounts currently work without support for this because Linux will fall back to SP4_NONE, but there is no guarantee this fallback will work forever. This commit only affects Kerberized NFSv4.1/4.2 mounts from Linux at this time. MFC after: 3 months
2023-04-07 19:49:23 +00:00
static int nfsrv_notsamecredname(int op, struct nfsrv_descript *nd,
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
struct nfsclient *clp);
static time_t nfsrv_leaseexpiry(void);
static void nfsrv_delaydelegtimeout(struct nfsstate *stp);
static int nfsrv_checkseqid(struct nfsrv_descript *nd, u_int32_t seqid,
struct nfsstate *stp, struct nfsrvcache *op);
static int nfsrv_nootherstate(struct nfsstate *stp);
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
static int nfsrv_locallock(vnode_t vp, struct nfslockfile *lfp, int flags,
uint64_t first, uint64_t end, struct nfslockconflict *cfp, NFSPROC_T *p);
static void nfsrv_localunlock(vnode_t vp, struct nfslockfile *lfp,
uint64_t init_first, uint64_t init_end, NFSPROC_T *p);
static int nfsrv_dolocal(vnode_t vp, struct nfslockfile *lfp, int flags,
int oldflags, uint64_t first, uint64_t end, struct nfslockconflict *cfp,
NFSPROC_T *p);
static void nfsrv_locallock_rollback(vnode_t vp, struct nfslockfile *lfp,
NFSPROC_T *p);
static void nfsrv_locallock_commit(struct nfslockfile *lfp, int flags,
uint64_t first, uint64_t end);
static void nfsrv_locklf(struct nfslockfile *lfp);
static void nfsrv_unlocklf(struct nfslockfile *lfp);
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
static struct nfsdsession *nfsrv_findsession(uint8_t *sessionid);
nfsd: Add support for the SP4_MACH_CRED case in ExchangeID Commit f4179ad46fa4 added support for operation bitmaps for NFSv4.1/4.2. This commit uses those to implement the SP4_MACH_CRED case for the NFSv4.1/4.2 ExchangeID operation since the Linux NFSv4.1/4.2 client is now using this for Kerberized mounts. The Linux Kerberized NFSv4.1/4.2 mounts currently work without support for this because Linux will fall back to SP4_NONE, but there is no guarantee this fallback will work forever. This commit only affects Kerberized NFSv4.1/4.2 mounts from Linux at this time. MFC after: 3 months
2023-04-07 19:49:23 +00:00
static int nfsrv_freesession(struct nfsrv_descript *nd, struct nfsdsession *sep,
uint8_t *sessionid);
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
static int nfsv4_setcbsequence(struct nfsrv_descript *nd, struct nfsclient *clp,
nfsd: fix session slot handling for failed callbacks When the NFSv4.1/4.2 server does a callback to a client on the back channel, it will use a session slot in the back channel session. If the back channel has failed, the callback will fail and, without this patch, the session slot will not be released. As more callbacks are attempted, all session slots can become busy and then the nfsd thread gets stuck waiting for a back channel session slot. This patch frees the session slot upon callback failure to avoid this problem. Without this patch, the problem can be avoided by leaving delegations disabled in the NFS server. MFC after: 2 weeks
2021-04-23 22:24:47 +00:00
int dont_replycache, struct nfsdsession **sepp, int *slotposp);
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
static int nfsv4_getcbsession(struct nfsclient *clp, struct nfsdsession **sepp);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
static int nfsrv_addlayout(struct nfsrv_descript *nd, struct nfslayout **lypp,
nfsv4stateid_t *stateidp, char *layp, int *layoutlenp, NFSPROC_T *p);
static void nfsrv_freelayout(struct nfslayouthead *lhp, struct nfslayout *lyp);
static void nfsrv_freelayoutlist(nfsquad_t clientid);
static void nfsrv_freelayouts(nfsquad_t *clid, fsid_t *fs, int laytype,
int iomode);
static void nfsrv_freealllayouts(void);
static void nfsrv_freedevid(struct nfsdevice *ds);
Add an optional feature to the pNFS server. Without this patch, the pNFS server distributes the data storage files across all of the specified DSs. A tester noted that it would be nice if a system administrator could control which DSs are used to store the file data for a given exported MDS file system. This patch adds the kernel support to do this. It also makes a slight semantic change to nfsv4_findmirror(), since some uses of it no longer require that the DS being searched for have a current mirror. A patch that will be committed in a few minutes will modify the nfsd daemon to support this feature. The patch should only affect sites using the pNFS server (specified via the "-p" command line option for nfsd. Suggested by: james.rose@framestore.com
2018-07-02 19:21:33 +00:00
static int nfsrv_setdsserver(char *dspathp, char *mdspathp, NFSPROC_T *p,
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
struct nfsdevice **dsp);
static void nfsrv_deleteds(struct nfsdevice *fndds);
static void nfsrv_allocdevid(struct nfsdevice *ds, char *addr, char *dnshost);
static void nfsrv_freealldevids(void);
static void nfsrv_flexlayouterr(struct nfsrv_descript *nd, uint32_t *layp,
int maxcnt, NFSPROC_T *p);
static int nfsrv_recalllayout(nfsquad_t clid, nfsv4stateid_t *stateidp,
Assorted fixes to handling of LayoutRecall callbacks, mostly error handling. After a re-read of the appropriate section of RFC5661, I decided that a few things should be changed related to LayoutRecall callback handling. Here are the things fixed by this patch. - For two of the three cases that LayoutRecall is done, I now think setting the clora_changed argument false is correct. - All errors other than NFSERR_DELAY returned by LayoutRecall appear permanent, so don't retry for any of them. (NFSERR_DELAY is retried by newnfs_request(), so it is not affected by this patch.) - Instead of waiting "forever" (actually until the process is SIGTERM'd) for Layouts to be returned during a mirror copy, fail and return ENXIO after about 1minute. Waiting for a <ctrl>C made sense when pnfsdscopymr() was done by itself, but did not make sense when done via find(1). This patch only affects the pNFS server.
2018-08-08 20:21:45 +00:00
fhandle_t *fhp, struct nfslayout *lyp, int changed, int laytype,
NFSPROC_T *p);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
static int nfsrv_findlayout(nfsquad_t *clientidp, fhandle_t *fhp, int laytype,
NFSPROC_T *, struct nfslayout **lypp);
static int nfsrv_fndclid(nfsquad_t *clidvec, nfsquad_t clid, int clidcnt);
static struct nfslayout *nfsrv_filelayout(struct nfsrv_descript *nd, int iomode,
fhandle_t *fhp, fhandle_t *dsfhp, char *devid, fsid_t fs);
static struct nfslayout *nfsrv_flexlayout(struct nfsrv_descript *nd, int iomode,
int mirrorcnt, fhandle_t *fhp, fhandle_t *dsfhp, char *devid, fsid_t fs);
static int nfsrv_dontlayout(fhandle_t *fhp);
static int nfsrv_createdsfile(vnode_t vp, fhandle_t *fhp, struct pnfsdsfile *pf,
vnode_t dvp, struct nfsdevice *ds, struct ucred *cred, NFSPROC_T *p,
vnode_t *tvpp);
Add an optional feature to the pNFS server. Without this patch, the pNFS server distributes the data storage files across all of the specified DSs. A tester noted that it would be nice if a system administrator could control which DSs are used to store the file data for a given exported MDS file system. This patch adds the kernel support to do this. It also makes a slight semantic change to nfsv4_findmirror(), since some uses of it no longer require that the DS being searched for have a current mirror. A patch that will be committed in a few minutes will modify the nfsd daemon to support this feature. The patch should only affect sites using the pNFS server (specified via the "-p" command line option for nfsd. Suggested by: james.rose@framestore.com
2018-07-02 19:21:33 +00:00
static struct nfsdevice *nfsrv_findmirroredds(struct nfsmount *nmp);
nfsd: Add support for the SP4_MACH_CRED case in ExchangeID Commit f4179ad46fa4 added support for operation bitmaps for NFSv4.1/4.2. This commit uses those to implement the SP4_MACH_CRED case for the NFSv4.1/4.2 ExchangeID operation since the Linux NFSv4.1/4.2 client is now using this for Kerberized mounts. The Linux Kerberized NFSv4.1/4.2 mounts currently work without support for this because Linux will fall back to SP4_NONE, but there is no guarantee this fallback will work forever. This commit only affects Kerberized NFSv4.1/4.2 mounts from Linux at this time. MFC after: 3 months
2023-04-07 19:49:23 +00:00
static int nfsrv_checkmachcred(int op, struct nfsrv_descript *nd,
struct nfsclient *clp);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* Scan the client list for a match and either return the current one,
* create a new entry or return an error.
* If returning a non-error, the clp structure must either be linked into
* the client list or free'd.
*/
nfs: Remove APPLESTATIC macro It is no longer useful. Reviewed by: rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24811
2020-05-12 13:23:25 +00:00
int
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_setclient(struct nfsrv_descript *nd, struct nfsclient **new_clpp,
nfsquad_t *clientidp, nfsquad_t *confirmp, NFSPROC_T *p)
{
struct nfsclient *clp = NULL, *new_clp = *new_clpp;
Fix a slow leak of session structures in the NFSv4.1 server. For a fairly rare case of a client doing an ExchangeID after a hard reboot, the old confirmed clientid still exists, but some clients use a new co_verifier. For this case, the server was not freeing up the sessions on the old confirmed clientid. This patch fixes this case. It also adds two LIST_INIT() macros, which are actually no-ops, since the structure is malloc()d with M_ZERO so the pointer is already set to NULL. It should have minimal impact, since the only way I could exercise this code path was by doing a hard power cycle (pulling the plus) on a machine running Linux with a NFSv4.1 mount on the server. Originally spotted during testing of the ESXi 6.5 client. Tested by: andreas.nagy@frequentis.com MFC after: 2 months
2018-05-13 12:42:53 +00:00
int i, error = 0, ret;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
struct nfsstate *stp, *tstp;
Add support for INET6 addresses to the kernel code that dumps open/lock state. PR#223036 reported that INET6 callback addresses were not printed by nfsdumpstate(8). This kernel patch adds INET6 addresses to the dump structure, so that nfsdumpstate(8) can print them out, post-r346190. The patch also includes the addition of #ifdef INET, INET6 as requested by bz@. PR: 223036 Reviewed by: bz, rgrimes MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D19839
2019-04-13 22:00:09 +00:00
#ifdef INET
struct sockaddr_in *sin, *rin;
#endif
#ifdef INET6
struct sockaddr_in6 *sin6, *rin6;
#endif
Fix a slow leak of session structures in the NFSv4.1 server. For a fairly rare case of a client doing an ExchangeID after a hard reboot, the old confirmed clientid still exists, but some clients use a new co_verifier. For this case, the server was not freeing up the sessions on the old confirmed clientid. This patch fixes this case. It also adds two LIST_INIT() macros, which are actually no-ops, since the structure is malloc()d with M_ZERO so the pointer is already set to NULL. It should have minimal impact, since the only way I could exercise this code path was by doing a hard power cycle (pulling the plus) on a machine running Linux with a NFSv4.1 mount on the server. Originally spotted during testing of the ESXi 6.5 client. Tested by: andreas.nagy@frequentis.com MFC after: 2 months
2018-05-13 12:42:53 +00:00
struct nfsdsession *sep, *nsep;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
int zapit = 0, gotit, hasstate = 0, igotlock;
static u_int64_t confirm_index = 0;
/*
* Check for state resource limit exceeded.
*/
Make the size of the hash tables used by the NFSv4 server tunable. No appreciable change in performance was observed after increasing the sizes of these tables and then testing with a single client. However, there was an email that indicated high CPU overheads for a heavily loaded NFSv4 and it is hoped that increasing the sizes of the hash tables via these tunables might help. The tables remain the same size by default. Differential Revision: https://reviews.freebsd.org/D2596 MFC after: 2 weeks
2015-05-27 22:00:05 +00:00
if (nfsrv_openpluslock > nfsrv_v4statelimit) {
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
error = NFSERR_RESOURCE;
goto out;
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
Disable attempts to establish a callback connection from the experimental NFSv4 server to a NFSv4 client when delegations are not being issued, even if the client advertises a callback path. This avoids a problem where a Linux client advertises a callback path that doesn't work, due to a firewall, and then times out an Open attempt before the FreeBSD server gives up its callback connection attempt. (Suggested by drb at karlov.mff.cuni.cz to fix the Linux client problem that he reported on the fs-stable mailing list.) The server should probably have a 1sec timeout on callback connection attempts when there are no delegations issued to the client, but that patch will require changes to the krpc and this serves as a work around until then. Tested by: drb at karlov.mff.cuni.cz MFC after: 5 days
2010-12-09 19:02:23 +00:00
if (nfsrv_issuedelegs == 0 ||
((nd->nd_flag & ND_GSS) != 0 && nfsrv_nogsscallback != 0))
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
Disable attempts to establish a callback connection from the experimental NFSv4 server to a NFSv4 client when delegations are not being issued, even if the client advertises a callback path. This avoids a problem where a Linux client advertises a callback path that doesn't work, due to a firewall, and then times out an Open attempt before the FreeBSD server gives up its callback connection attempt. (Suggested by drb at karlov.mff.cuni.cz to fix the Linux client problem that he reported on the fs-stable mailing list.) The server should probably have a 1sec timeout on callback connection attempts when there are no delegations issued to the client, but that patch will require changes to the krpc and this serves as a work around until then. Tested by: drb at karlov.mff.cuni.cz MFC after: 5 days
2010-12-09 19:02:23 +00:00
* Don't do callbacks when delegations are disabled or
* for AUTH_GSS unless enabled via nfsrv_nogsscallback.
* If establishing a callback connection is attempted
* when a firewall is blocking the callback path, the
* server may wait too long for the connect attempt to
* succeed during the Open. Some clients, such as Linux,
* may timeout and give up on the Open before the server
* replies. Also, since AUTH_GSS callbacks are not
* yet interoperability tested, they might cause the
* server to crap out, if they get past the Init call to
* the client.
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
*/
new_clp->lc_program = 0;
/* Lock out other nfsd threads */
NFSLOCKV4ROOTMUTEX();
nfsv4_relref(&nfsv4rootfs_lock);
do {
igotlock = nfsv4_lock(&nfsv4rootfs_lock, 1, NULL,
Fix the new NFS client so that it handles NFSv4 state correctly during a forced dismount. This required that the exclusive and shared (refcnt) sleep lock functions check for MNTK_UMOUNTF before sleeping, so that they won't block while nfscl_umount() is getting rid of the state. As such, a "struct mount *" argument was added to the locking functions. I believe the only remaining case where a forced dismount can get hung in the kernel is when a thread is already attempting to do a TCP connect to a dead server when the krpc client structure called nr_client is NULL. This will only happen just after a "mount -u" with options that force a new TCP connection is done, so it shouldn't be a problem in practice. MFC after: 2 weeks
2011-05-27 22:05:10 +00:00
NFSV4ROOTLOCKMUTEXPTR, NULL);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
} while (!igotlock);
NFSUNLOCKV4ROOTMUTEX();
/*
* Search for a match in the client list.
*/
gotit = i = 0;
Make the size of the hash tables used by the NFSv4 server tunable. No appreciable change in performance was observed after increasing the sizes of these tables and then testing with a single client. However, there was an email that indicated high CPU overheads for a heavily loaded NFSv4 and it is hoped that increasing the sizes of the hash tables via these tunables might help. The tables remain the same size by default. Differential Revision: https://reviews.freebsd.org/D2596 MFC after: 2 weeks
2015-05-27 22:00:05 +00:00
while (i < nfsrv_clienthashsize && !gotit) {
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
LIST_FOREACH(clp, &NFSD_VNET(nfsclienthash)[i], lc_hash) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (new_clp->lc_idlen == clp->lc_idlen &&
!NFSBCMP(new_clp->lc_id, clp->lc_id, clp->lc_idlen)) {
gotit = 1;
break;
}
}
This patch fixes a problem where, if the NFSv4 server has a previous unconfirmed clientid structure for the same client on the last hash list, this old entry would not be removed/deleted. I do not think this bug would have caused serious problems, since the new entry would have been before the old one on the list. This old entry would have eventually been scavenged/removed. Detected while reading the code looking for another bug. MFC after: 3 days
2015-07-29 23:06:30 +00:00
if (gotit == 0)
i++;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
if (!gotit ||
(clp->lc_flags & (LCL_NEEDSCONFIRM | LCL_ADMINREVOKED))) {
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
if ((nd->nd_flag & ND_NFSV41) != 0 && confirmp->lval[1] != 0) {
/*
* For NFSv4.1, if confirmp->lval[1] is non-zero, the
* client is trying to update a confirmed clientid.
*/
NFSLOCKV4ROOTMUTEX();
nfsv4_unlock(&nfsv4rootfs_lock, 1);
NFSUNLOCKV4ROOTMUTEX();
confirmp->lval[1] = 0;
error = NFSERR_NOENT;
goto out;
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* Get rid of the old one.
*/
Make the size of the hash tables used by the NFSv4 server tunable. No appreciable change in performance was observed after increasing the sizes of these tables and then testing with a single client. However, there was an email that indicated high CPU overheads for a heavily loaded NFSv4 and it is hoped that increasing the sizes of the hash tables via these tunables might help. The tables remain the same size by default. Differential Revision: https://reviews.freebsd.org/D2596 MFC after: 2 weeks
2015-05-27 22:00:05 +00:00
if (i != nfsrv_clienthashsize) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
LIST_REMOVE(clp, lc_hash);
nfsrv_cleanclient(clp, p);
nfsrv_freedeleglist(&clp->lc_deleg);
nfsrv_freedeleglist(&clp->lc_olddeleg);
zapit = 1;
}
/*
* Add it after assigning a client id to it.
*/
new_clp->lc_flags |= LCL_NEEDSCONFIRM;
nfsd: Fix CreateSession for an established ClientID I mis-read the RFC w.r.t. handling of the sequenceid when a CreateSession is done after the initial one that confirms the ClientID. Fortunately this does not affect most extant NFSv4.1/4.2 clients, since they only acquire a single session for TCP for a ClientID (Solaris might be an exception?). This patch fixes the server to handle this case, where the RFC requires the sequenceid be incremented for each CreateSession and is required to reply to a retried CreateSession with a cached reply. It adds a field to nfsclient called lc_prevsess, which caches the sessionid, which is the only field in a CreateSession reply that will change for a retry, to implement this reply cache. The recent commits up to d4a11b3e3bdd that mark session slots bad when "intr" and/or "soft" mounts are used by the client needs this server patch. Without this patch, the client will do a full recovery, including a new ClientID, losing all byte range locks. However, prior to the recent client commits, the client would hang when all session slots were bad, so even without this patch it is not a regression. PR: 260011 MFC after: 2 weeks
2022-07-13 23:28:56 +00:00
if ((nd->nd_flag & ND_NFSV41) != 0) {
confirmp->lval[0] = ++confirm_index;
new_clp->lc_confirm.lval[0] = confirmp->lval[0] - 1;
} else
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
confirmp->qval = new_clp->lc_confirm.qval =
++confirm_index;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
clientidp->lval[0] = new_clp->lc_clientid.lval[0] =
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
NFSD_VNET(nfsrvboottime);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
clientidp->lval[1] = new_clp->lc_clientid.lval[1] =
nfsrv_nextclientindex();
new_clp->lc_stateindex = 0;
new_clp->lc_statemaxindex = 0;
nfsd: Fix CreateSession for an established ClientID I mis-read the RFC w.r.t. handling of the sequenceid when a CreateSession is done after the initial one that confirms the ClientID. Fortunately this does not affect most extant NFSv4.1/4.2 clients, since they only acquire a single session for TCP for a ClientID (Solaris might be an exception?). This patch fixes the server to handle this case, where the RFC requires the sequenceid be incremented for each CreateSession and is required to reply to a retried CreateSession with a cached reply. It adds a field to nfsclient called lc_prevsess, which caches the sessionid, which is the only field in a CreateSession reply that will change for a retry, to implement this reply cache. The recent commits up to d4a11b3e3bdd that mark session slots bad when "intr" and/or "soft" mounts are used by the client needs this server patch. Without this patch, the client will do a full recovery, including a new ClientID, losing all byte range locks. However, prior to the recent client commits, the client would hang when all session slots were bad, so even without this patch it is not a regression. PR: 260011 MFC after: 2 weeks
2022-07-13 23:28:56 +00:00
new_clp->lc_prevsess = 0;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
new_clp->lc_cbref = 0;
new_clp->lc_expiry = nfsrv_leaseexpiry();
LIST_INIT(&new_clp->lc_open);
LIST_INIT(&new_clp->lc_deleg);
LIST_INIT(&new_clp->lc_olddeleg);
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
LIST_INIT(&new_clp->lc_session);
Make the size of the hash tables used by the NFSv4 server tunable. No appreciable change in performance was observed after increasing the sizes of these tables and then testing with a single client. However, there was an email that indicated high CPU overheads for a heavily loaded NFSv4 and it is hoped that increasing the sizes of the hash tables via these tunables might help. The tables remain the same size by default. Differential Revision: https://reviews.freebsd.org/D2596 MFC after: 2 weeks
2015-05-27 22:00:05 +00:00
for (i = 0; i < nfsrv_statehashsize; i++)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
LIST_INIT(&new_clp->lc_stateid[i]);
LIST_INSERT_HEAD(NFSCLIENTHASH(new_clp->lc_clientid), new_clp,
lc_hash);
nfsd: Wrap nfsstatsv1_p in the NFSD_VNET() macro Commit 7344856e3a6d added a lot of macros that will front end vnet macros so that nfsd(8) can run in vnet prison. The nfsstatsv1_p variable got missed. This patch wraps all uses of nfsstatsv1_p with the NFSD_VNET() macro. The NFSD_VNET() macro is still a null macro. MFC after: 3 months
2023-02-16 01:39:07 +00:00
NFSD_VNET(nfsstatsv1_p)->srvclients++;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_openpluslock++;
nfsrv_clients++;
NFSLOCKV4ROOTMUTEX();
nfsv4_unlock(&nfsv4rootfs_lock, 1);
NFSUNLOCKV4ROOTMUTEX();
if (zapit)
nfsrv_zapclient(clp, p);
*new_clpp = NULL;
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* Now, handle the cases where the id is already issued.
*/
nfsd: Add support for the SP4_MACH_CRED case in ExchangeID Commit f4179ad46fa4 added support for operation bitmaps for NFSv4.1/4.2. This commit uses those to implement the SP4_MACH_CRED case for the NFSv4.1/4.2 ExchangeID operation since the Linux NFSv4.1/4.2 client is now using this for Kerberized mounts. The Linux Kerberized NFSv4.1/4.2 mounts currently work without support for this because Linux will fall back to SP4_NONE, but there is no guarantee this fallback will work forever. This commit only affects Kerberized NFSv4.1/4.2 mounts from Linux at this time. MFC after: 3 months
2023-04-07 19:49:23 +00:00
if (nfsrv_notsamecredname(NFSV4OP_EXCHANGEID, nd, clp)) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* Check to see if there is expired state that should go away.
*/
if (clp->lc_expiry < NFSD_MONOSEC &&
(!LIST_EMPTY(&clp->lc_open) || !LIST_EMPTY(&clp->lc_deleg))) {
nfsrv_cleanclient(clp, p);
nfsrv_freedeleglist(&clp->lc_deleg);
}
/*
* If there is outstanding state, then reply NFSERR_CLIDINUSE per
* RFC3530 Sec. 8.1.2 last para.
*/
if (!LIST_EMPTY(&clp->lc_deleg)) {
hasstate = 1;
} else if (LIST_EMPTY(&clp->lc_open)) {
hasstate = 0;
} else {
hasstate = 0;
/* Look for an Open on the OpenOwner */
LIST_FOREACH(stp, &clp->lc_open, ls_list) {
if (!LIST_EMPTY(&stp->ls_open)) {
hasstate = 1;
break;
}
}
}
if (hasstate) {
/*
* If the uid doesn't match, return NFSERR_CLIDINUSE after
* filling out the correct ipaddr and portnum.
*/
Add support for INET6 addresses to the kernel code that dumps open/lock state. PR#223036 reported that INET6 callback addresses were not printed by nfsdumpstate(8). This kernel patch adds INET6 addresses to the dump structure, so that nfsdumpstate(8) can print them out, post-r346190. The patch also includes the addition of #ifdef INET, INET6 as requested by bz@. PR: 223036 Reviewed by: bz, rgrimes MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D19839
2019-04-13 22:00:09 +00:00
switch (clp->lc_req.nr_nam->sa_family) {
#ifdef INET
case AF_INET:
sin = (struct sockaddr_in *)new_clp->lc_req.nr_nam;
rin = (struct sockaddr_in *)clp->lc_req.nr_nam;
sin->sin_addr.s_addr = rin->sin_addr.s_addr;
sin->sin_port = rin->sin_port;
break;
#endif
#ifdef INET6
case AF_INET6:
sin6 = (struct sockaddr_in6 *)new_clp->lc_req.nr_nam;
rin6 = (struct sockaddr_in6 *)clp->lc_req.nr_nam;
sin6->sin6_addr = rin6->sin6_addr;
sin6->sin6_port = rin6->sin6_port;
break;
#endif
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
NFSLOCKV4ROOTMUTEX();
nfsv4_unlock(&nfsv4rootfs_lock, 1);
NFSUNLOCKV4ROOTMUTEX();
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
error = NFSERR_CLIDINUSE;
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
}
if (NFSBCMP(new_clp->lc_verf, clp->lc_verf, NFSX_VERF)) {
/*
* If the verifier has changed, the client has rebooted
* and a new client id is issued. The old state info
* can be thrown away once the SETCLIENTID_CONFIRM occurs.
*/
LIST_REMOVE(clp, lc_hash);
Fix a slow leak of session structures in the NFSv4.1 server. For a fairly rare case of a client doing an ExchangeID after a hard reboot, the old confirmed clientid still exists, but some clients use a new co_verifier. For this case, the server was not freeing up the sessions on the old confirmed clientid. This patch fixes this case. It also adds two LIST_INIT() macros, which are actually no-ops, since the structure is malloc()d with M_ZERO so the pointer is already set to NULL. It should have minimal impact, since the only way I could exercise this code path was by doing a hard power cycle (pulling the plus) on a machine running Linux with a NFSv4.1 mount on the server. Originally spotted during testing of the ESXi 6.5 client. Tested by: andreas.nagy@frequentis.com MFC after: 2 months
2018-05-13 12:42:53 +00:00
/* Get rid of all sessions on this clientid. */
LIST_FOREACH_SAFE(sep, &clp->lc_session, sess_list, nsep) {
nfsd: Add support for the SP4_MACH_CRED case in ExchangeID Commit f4179ad46fa4 added support for operation bitmaps for NFSv4.1/4.2. This commit uses those to implement the SP4_MACH_CRED case for the NFSv4.1/4.2 ExchangeID operation since the Linux NFSv4.1/4.2 client is now using this for Kerberized mounts. The Linux Kerberized NFSv4.1/4.2 mounts currently work without support for this because Linux will fall back to SP4_NONE, but there is no guarantee this fallback will work forever. This commit only affects Kerberized NFSv4.1/4.2 mounts from Linux at this time. MFC after: 3 months
2023-04-07 19:49:23 +00:00
ret = nfsrv_freesession(NULL, sep, NULL);
Fix a slow leak of session structures in the NFSv4.1 server. For a fairly rare case of a client doing an ExchangeID after a hard reboot, the old confirmed clientid still exists, but some clients use a new co_verifier. For this case, the server was not freeing up the sessions on the old confirmed clientid. This patch fixes this case. It also adds two LIST_INIT() macros, which are actually no-ops, since the structure is malloc()d with M_ZERO so the pointer is already set to NULL. It should have minimal impact, since the only way I could exercise this code path was by doing a hard power cycle (pulling the plus) on a machine running Linux with a NFSv4.1 mount on the server. Originally spotted during testing of the ESXi 6.5 client. Tested by: andreas.nagy@frequentis.com MFC after: 2 months
2018-05-13 12:42:53 +00:00
if (ret != 0)
printf("nfsrv_setclient: verifier changed free"
" session failed=%d\n", ret);
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
new_clp->lc_flags |= LCL_NEEDSCONFIRM;
nfsd: Fix CreateSession for an established ClientID I mis-read the RFC w.r.t. handling of the sequenceid when a CreateSession is done after the initial one that confirms the ClientID. Fortunately this does not affect most extant NFSv4.1/4.2 clients, since they only acquire a single session for TCP for a ClientID (Solaris might be an exception?). This patch fixes the server to handle this case, where the RFC requires the sequenceid be incremented for each CreateSession and is required to reply to a retried CreateSession with a cached reply. It adds a field to nfsclient called lc_prevsess, which caches the sessionid, which is the only field in a CreateSession reply that will change for a retry, to implement this reply cache. The recent commits up to d4a11b3e3bdd that mark session slots bad when "intr" and/or "soft" mounts are used by the client needs this server patch. Without this patch, the client will do a full recovery, including a new ClientID, losing all byte range locks. However, prior to the recent client commits, the client would hang when all session slots were bad, so even without this patch it is not a regression. PR: 260011 MFC after: 2 weeks
2022-07-13 23:28:56 +00:00
if ((nd->nd_flag & ND_NFSV41) != 0) {
confirmp->lval[0] = ++confirm_index;
new_clp->lc_confirm.lval[0] = confirmp->lval[0] - 1;
} else
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
confirmp->qval = new_clp->lc_confirm.qval =
++confirm_index;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
clientidp->lval[0] = new_clp->lc_clientid.lval[0] =
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
NFSD_VNET(nfsrvboottime);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
clientidp->lval[1] = new_clp->lc_clientid.lval[1] =
nfsrv_nextclientindex();
new_clp->lc_stateindex = 0;
new_clp->lc_statemaxindex = 0;
nfsd: Fix CreateSession for an established ClientID I mis-read the RFC w.r.t. handling of the sequenceid when a CreateSession is done after the initial one that confirms the ClientID. Fortunately this does not affect most extant NFSv4.1/4.2 clients, since they only acquire a single session for TCP for a ClientID (Solaris might be an exception?). This patch fixes the server to handle this case, where the RFC requires the sequenceid be incremented for each CreateSession and is required to reply to a retried CreateSession with a cached reply. It adds a field to nfsclient called lc_prevsess, which caches the sessionid, which is the only field in a CreateSession reply that will change for a retry, to implement this reply cache. The recent commits up to d4a11b3e3bdd that mark session slots bad when "intr" and/or "soft" mounts are used by the client needs this server patch. Without this patch, the client will do a full recovery, including a new ClientID, losing all byte range locks. However, prior to the recent client commits, the client would hang when all session slots were bad, so even without this patch it is not a regression. PR: 260011 MFC after: 2 weeks
2022-07-13 23:28:56 +00:00
new_clp->lc_prevsess = 0;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
new_clp->lc_cbref = 0;
new_clp->lc_expiry = nfsrv_leaseexpiry();
/*
* Save the state until confirmed.
*/
LIST_NEWHEAD(&new_clp->lc_open, &clp->lc_open, ls_list);
LIST_FOREACH(tstp, &new_clp->lc_open, ls_list)
tstp->ls_clp = new_clp;
LIST_NEWHEAD(&new_clp->lc_deleg, &clp->lc_deleg, ls_list);
LIST_FOREACH(tstp, &new_clp->lc_deleg, ls_list)
tstp->ls_clp = new_clp;
LIST_NEWHEAD(&new_clp->lc_olddeleg, &clp->lc_olddeleg,
ls_list);
LIST_FOREACH(tstp, &new_clp->lc_olddeleg, ls_list)
tstp->ls_clp = new_clp;
Make the size of the hash tables used by the NFSv4 server tunable. No appreciable change in performance was observed after increasing the sizes of these tables and then testing with a single client. However, there was an email that indicated high CPU overheads for a heavily loaded NFSv4 and it is hoped that increasing the sizes of the hash tables via these tunables might help. The tables remain the same size by default. Differential Revision: https://reviews.freebsd.org/D2596 MFC after: 2 weeks
2015-05-27 22:00:05 +00:00
for (i = 0; i < nfsrv_statehashsize; i++) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
LIST_NEWHEAD(&new_clp->lc_stateid[i],
&clp->lc_stateid[i], ls_hash);
Tai Horgan reported via email that there were two places in the new NFSv4 server where the code follows the wrong list. Fortunately, for these fairly rare cases, the lc_stateid[] lists are normally empty. This patch fixes the code to follow the correct list. Reported by: tai.horgan at isilon.com Discussed with: zack MFC after: 2 weeks
2012-01-14 04:04:58 +00:00
LIST_FOREACH(tstp, &new_clp->lc_stateid[i], ls_hash)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
tstp->ls_clp = new_clp;
}
Fix a slow leak of session structures in the NFSv4.1 server. For a fairly rare case of a client doing an ExchangeID after a hard reboot, the old confirmed clientid still exists, but some clients use a new co_verifier. For this case, the server was not freeing up the sessions on the old confirmed clientid. This patch fixes this case. It also adds two LIST_INIT() macros, which are actually no-ops, since the structure is malloc()d with M_ZERO so the pointer is already set to NULL. It should have minimal impact, since the only way I could exercise this code path was by doing a hard power cycle (pulling the plus) on a machine running Linux with a NFSv4.1 mount on the server. Originally spotted during testing of the ESXi 6.5 client. Tested by: andreas.nagy@frequentis.com MFC after: 2 months
2018-05-13 12:42:53 +00:00
LIST_INIT(&new_clp->lc_session);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
LIST_INSERT_HEAD(NFSCLIENTHASH(new_clp->lc_clientid), new_clp,
lc_hash);
nfsd: Wrap nfsstatsv1_p in the NFSD_VNET() macro Commit 7344856e3a6d added a lot of macros that will front end vnet macros so that nfsd(8) can run in vnet prison. The nfsstatsv1_p variable got missed. This patch wraps all uses of nfsstatsv1_p with the NFSD_VNET() macro. The NFSD_VNET() macro is still a null macro. MFC after: 3 months
2023-02-16 01:39:07 +00:00
NFSD_VNET(nfsstatsv1_p)->srvclients++;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_openpluslock++;
nfsrv_clients++;
NFSLOCKV4ROOTMUTEX();
nfsv4_unlock(&nfsv4rootfs_lock, 1);
NFSUNLOCKV4ROOTMUTEX();
/*
* Must wait until any outstanding callback on the old clp
* completes.
*/
Fix two cases in the new NFS server where a tsleep() is used, when the code should actually protect the tested variable with a mutex. Since the tsleep()s had a 10sec timeout, the race would have only delayed the allocation of a new clientid for a client. The sleeps will also rarely occur, since having a callback in progress when a client acquires a new clientid, is unlikely. in practice, since having a callback in progress when a fresh clientid is being acquired by a client is unlikely. MFC after: 1 month
2012-05-12 22:20:55 +00:00
NFSLOCKSTATE();
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
while (clp->lc_cbref) {
clp->lc_flags |= LCL_WAKEUPWANTED;
Fix two cases in the new NFS server where a tsleep() is used, when the code should actually protect the tested variable with a mutex. Since the tsleep()s had a 10sec timeout, the race would have only delayed the allocation of a new clientid for a client. The sleeps will also rarely occur, since having a callback in progress when a client acquires a new clientid, is unlikely. in practice, since having a callback in progress when a fresh clientid is being acquired by a client is unlikely. MFC after: 1 month
2012-05-12 22:20:55 +00:00
(void)mtx_sleep(clp, NFSSTATEMUTEXPTR, PZERO - 1,
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
"nfsd clp", 10 * hz);
}
Fix two cases in the new NFS server where a tsleep() is used, when the code should actually protect the tested variable with a mutex. Since the tsleep()s had a 10sec timeout, the race would have only delayed the allocation of a new clientid for a client. The sleeps will also rarely occur, since having a callback in progress when a client acquires a new clientid, is unlikely. in practice, since having a callback in progress when a fresh clientid is being acquired by a client is unlikely. MFC after: 1 month
2012-05-12 22:20:55 +00:00
NFSUNLOCKSTATE();
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_zapclient(clp, p);
*new_clpp = NULL;
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
/* For NFSv4.1, mark that we found a confirmed clientid. */
For the case where an NFSv4.1 ExchangeID operation has the client identifier that already has a confirmed ClientID, the nfsrv_setclient() function would not fill in the clientidp being returned. As such, the value of ClientID returned would be whatever garbage was on the stack. An NFSv4.1 client would not normally do this, but it appears that it can happen for certain Linux clients. When it happens, the client persistently retries the ExchangeID and Create_session after Create_session fails when it uses the bogus clientid. With this patch, the correct clientid is replied. This problem was identified in a packet trace supplied by Ahmed Kamal via email. Reported by: email.ahmedkamal@googlemail.com MFC after: 2 weeks
2015-08-14 22:02:14 +00:00
if ((nd->nd_flag & ND_NFSV41) != 0) {
clientidp->lval[0] = clp->lc_clientid.lval[0];
clientidp->lval[1] = clp->lc_clientid.lval[1];
confirmp->lval[0] = 0; /* Ignored by client */
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
confirmp->lval[1] = 1;
For the case where an NFSv4.1 ExchangeID operation has the client identifier that already has a confirmed ClientID, the nfsrv_setclient() function would not fill in the clientidp being returned. As such, the value of ClientID returned would be whatever garbage was on the stack. An NFSv4.1 client would not normally do this, but it appears that it can happen for certain Linux clients. When it happens, the client persistently retries the ExchangeID and Create_session after Create_session fails when it uses the bogus clientid. With this patch, the correct clientid is replied. This problem was identified in a packet trace supplied by Ahmed Kamal via email. Reported by: email.ahmedkamal@googlemail.com MFC after: 2 weeks
2015-08-14 22:02:14 +00:00
} else {
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
/*
* id and verifier match, so update the net address info
* and get rid of any existing callback authentication
* handle, so a new one will be acquired.
*/
LIST_REMOVE(clp, lc_hash);
new_clp->lc_flags |= (LCL_NEEDSCONFIRM | LCL_DONTCLEAN);
new_clp->lc_expiry = nfsrv_leaseexpiry();
confirmp->qval = new_clp->lc_confirm.qval = ++confirm_index;
clientidp->lval[0] = new_clp->lc_clientid.lval[0] =
clp->lc_clientid.lval[0];
clientidp->lval[1] = new_clp->lc_clientid.lval[1] =
clp->lc_clientid.lval[1];
new_clp->lc_delegtime = clp->lc_delegtime;
new_clp->lc_stateindex = clp->lc_stateindex;
new_clp->lc_statemaxindex = clp->lc_statemaxindex;
new_clp->lc_cbref = 0;
LIST_NEWHEAD(&new_clp->lc_open, &clp->lc_open, ls_list);
LIST_FOREACH(tstp, &new_clp->lc_open, ls_list)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
tstp->ls_clp = new_clp;
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
LIST_NEWHEAD(&new_clp->lc_deleg, &clp->lc_deleg, ls_list);
LIST_FOREACH(tstp, &new_clp->lc_deleg, ls_list)
tstp->ls_clp = new_clp;
LIST_NEWHEAD(&new_clp->lc_olddeleg, &clp->lc_olddeleg, ls_list);
LIST_FOREACH(tstp, &new_clp->lc_olddeleg, ls_list)
tstp->ls_clp = new_clp;
Make the size of the hash tables used by the NFSv4 server tunable. No appreciable change in performance was observed after increasing the sizes of these tables and then testing with a single client. However, there was an email that indicated high CPU overheads for a heavily loaded NFSv4 and it is hoped that increasing the sizes of the hash tables via these tunables might help. The tables remain the same size by default. Differential Revision: https://reviews.freebsd.org/D2596 MFC after: 2 weeks
2015-05-27 22:00:05 +00:00
for (i = 0; i < nfsrv_statehashsize; i++) {
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
LIST_NEWHEAD(&new_clp->lc_stateid[i],
&clp->lc_stateid[i], ls_hash);
LIST_FOREACH(tstp, &new_clp->lc_stateid[i], ls_hash)
tstp->ls_clp = new_clp;
}
Fix a slow leak of session structures in the NFSv4.1 server. For a fairly rare case of a client doing an ExchangeID after a hard reboot, the old confirmed clientid still exists, but some clients use a new co_verifier. For this case, the server was not freeing up the sessions on the old confirmed clientid. This patch fixes this case. It also adds two LIST_INIT() macros, which are actually no-ops, since the structure is malloc()d with M_ZERO so the pointer is already set to NULL. It should have minimal impact, since the only way I could exercise this code path was by doing a hard power cycle (pulling the plus) on a machine running Linux with a NFSv4.1 mount on the server. Originally spotted during testing of the ESXi 6.5 client. Tested by: andreas.nagy@frequentis.com MFC after: 2 months
2018-05-13 12:42:53 +00:00
LIST_INIT(&new_clp->lc_session);
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
LIST_INSERT_HEAD(NFSCLIENTHASH(new_clp->lc_clientid), new_clp,
lc_hash);
nfsd: Wrap nfsstatsv1_p in the NFSD_VNET() macro Commit 7344856e3a6d added a lot of macros that will front end vnet macros so that nfsd(8) can run in vnet prison. The nfsstatsv1_p variable got missed. This patch wraps all uses of nfsstatsv1_p with the NFSD_VNET() macro. The NFSD_VNET() macro is still a null macro. MFC after: 3 months
2023-02-16 01:39:07 +00:00
NFSD_VNET(nfsstatsv1_p)->srvclients++;
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
nfsrv_openpluslock++;
nfsrv_clients++;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
NFSLOCKV4ROOTMUTEX();
nfsv4_unlock(&nfsv4rootfs_lock, 1);
NFSUNLOCKV4ROOTMUTEX();
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
if ((nd->nd_flag & ND_NFSV41) == 0) {
/*
* Must wait until any outstanding callback on the old clp
* completes.
*/
NFSLOCKSTATE();
while (clp->lc_cbref) {
clp->lc_flags |= LCL_WAKEUPWANTED;
(void)mtx_sleep(clp, NFSSTATEMUTEXPTR, PZERO - 1,
"nfsdclp", 10 * hz);
}
NFSUNLOCKSTATE();
nfsrv_zapclient(clp, p);
*new_clpp = NULL;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
out:
NFSEXITCODE2(error, nd);
return (error);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* Check to see if the client id exists and optionally confirm it.
*/
nfs: Remove APPLESTATIC macro It is no longer useful. Reviewed by: rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24811
2020-05-12 13:23:25 +00:00
int
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_getclient(nfsquad_t clientid, int opflags, struct nfsclient **clpp,
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
struct nfsdsession *nsep, nfsquad_t confirm, uint32_t cbprogram,
struct nfsrv_descript *nd, NFSPROC_T *p)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
{
struct nfsclient *clp;
struct nfsstate *stp;
int i;
struct nfsclienthashhead *hp;
int error = 0, igotlock, doneok;
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
struct nfssessionhash *shp;
struct nfsdsession *sep;
uint64_t sessid[2];
nfsd: Fix CreateSession for an established ClientID I mis-read the RFC w.r.t. handling of the sequenceid when a CreateSession is done after the initial one that confirms the ClientID. Fortunately this does not affect most extant NFSv4.1/4.2 clients, since they only acquire a single session for TCP for a ClientID (Solaris might be an exception?). This patch fixes the server to handle this case, where the RFC requires the sequenceid be incremented for each CreateSession and is required to reply to a retried CreateSession with a cached reply. It adds a field to nfsclient called lc_prevsess, which caches the sessionid, which is the only field in a CreateSession reply that will change for a retry, to implement this reply cache. The recent commits up to d4a11b3e3bdd that mark session slots bad when "intr" and/or "soft" mounts are used by the client needs this server patch. Without this patch, the client will do a full recovery, including a new ClientID, losing all byte range locks. However, prior to the recent client commits, the client would hang when all session slots were bad, so even without this patch it is not a regression. PR: 260011 MFC after: 2 weeks
2022-07-13 23:28:56 +00:00
bool sess_replay;
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
static uint64_t next_sess = 0;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (clpp)
*clpp = NULL;
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
if ((nd == NULL || (nd->nd_flag & ND_NFSV41) == 0 ||
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
opflags != CLOPS_RENEW) && NFSD_VNET(nfsrvboottime) !=
clientid.lval[0]) {
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
error = NFSERR_STALECLIENTID;
goto out;
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* If called with opflags == CLOPS_RENEW, the State Lock is
* already held. Otherwise, we need to get either that or,
* for the case of Confirm, lock out the nfsd threads.
*/
if (opflags & CLOPS_CONFIRM) {
NFSLOCKV4ROOTMUTEX();
nfsv4_relref(&nfsv4rootfs_lock);
do {
igotlock = nfsv4_lock(&nfsv4rootfs_lock, 1, NULL,
Fix the new NFS client so that it handles NFSv4 state correctly during a forced dismount. This required that the exclusive and shared (refcnt) sleep lock functions check for MNTK_UMOUNTF before sleeping, so that they won't block while nfscl_umount() is getting rid of the state. As such, a "struct mount *" argument was added to the locking functions. I believe the only remaining case where a forced dismount can get hung in the kernel is when a thread is already attempting to do a TCP connect to a dead server when the krpc client structure called nr_client is NULL. This will only happen just after a "mount -u" with options that force a new TCP connection is done, so it shouldn't be a problem in practice. MFC after: 2 weeks
2011-05-27 22:05:10 +00:00
NFSV4ROOTLOCKMUTEXPTR, NULL);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
} while (!igotlock);
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
/*
* Create a new sessionid here, since we need to do it where
* there is a mutex held to serialize update of next_sess.
*/
if ((nd->nd_flag & ND_NFSV41) != 0) {
sessid[0] = ++next_sess;
sessid[1] = clientid.qval;
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
NFSUNLOCKV4ROOTMUTEX();
} else if (opflags != CLOPS_RENEW) {
NFSLOCKSTATE();
}
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
/* For NFSv4.1, the clp is acquired from the associated session. */
if (nd != NULL && (nd->nd_flag & ND_NFSV41) != 0 &&
opflags == CLOPS_RENEW) {
clp = NULL;
if ((nd->nd_flag & ND_HASSEQUENCE) != 0) {
shp = NFSSESSIONHASH(nd->nd_sessionid);
NFSLOCKSESSION(shp);
sep = nfsrv_findsession(nd->nd_sessionid);
if (sep != NULL)
clp = sep->sess_clp;
NFSUNLOCKSESSION(shp);
}
} else {
hp = NFSCLIENTHASH(clientid);
LIST_FOREACH(clp, hp, lc_hash) {
if (clp->lc_clientid.lval[1] == clientid.lval[1])
break;
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
if (clp == NULL) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (opflags & CLOPS_CONFIRM)
error = NFSERR_STALECLIENTID;
else
error = NFSERR_EXPIRED;
} else if (clp->lc_flags & LCL_ADMINREVOKED) {
/*
* If marked admin revoked, just return the error.
*/
error = NFSERR_ADMINREVOKED;
}
if (error) {
if (opflags & CLOPS_CONFIRM) {
NFSLOCKV4ROOTMUTEX();
nfsv4_unlock(&nfsv4rootfs_lock, 1);
NFSUNLOCKV4ROOTMUTEX();
} else if (opflags != CLOPS_RENEW) {
NFSUNLOCKSTATE();
}
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* Perform any operations specified by the opflags.
*/
if (opflags & CLOPS_CONFIRM) {
nfsd: Fix CreateSession for an established ClientID I mis-read the RFC w.r.t. handling of the sequenceid when a CreateSession is done after the initial one that confirms the ClientID. Fortunately this does not affect most extant NFSv4.1/4.2 clients, since they only acquire a single session for TCP for a ClientID (Solaris might be an exception?). This patch fixes the server to handle this case, where the RFC requires the sequenceid be incremented for each CreateSession and is required to reply to a retried CreateSession with a cached reply. It adds a field to nfsclient called lc_prevsess, which caches the sessionid, which is the only field in a CreateSession reply that will change for a retry, to implement this reply cache. The recent commits up to d4a11b3e3bdd that mark session slots bad when "intr" and/or "soft" mounts are used by the client needs this server patch. Without this patch, the client will do a full recovery, including a new ClientID, losing all byte range locks. However, prior to the recent client commits, the client would hang when all session slots were bad, so even without this patch it is not a regression. PR: 260011 MFC after: 2 weeks
2022-07-13 23:28:56 +00:00
sess_replay = false;
if ((nd->nd_flag & ND_NFSV41) != 0) {
/*
* For the case where lc_confirm.lval[0] == confirm.lval[0],
* use the new session, but with the previous sessionid.
* This is not exactly what the RFC describes, but should
* result in the same reply as the previous CreateSession.
*/
if (clp->lc_confirm.lval[0] + 1 == confirm.lval[0]) {
clp->lc_confirm.lval[0] = confirm.lval[0];
clp->lc_prevsess = sessid[0];
} else if (clp->lc_confirm.lval[0] == confirm.lval[0]) {
if (clp->lc_prevsess == 0)
error = NFSERR_SEQMISORDERED;
else
sessid[0] = clp->lc_prevsess;
sess_replay = true;
} else
nfsd: Reply NFSERR_SEQMISORDERED for bogus seqid argument The ESXi NFSv4.1 client bogusly sends the wrong value for the csa_sequence argument for a Create_session operation. RFC8881 requires this value to be the same as the sequence reply from the ExchangeID operation most recently done for the client ID. Without this patch, the server replies NFSERR_STALECLIENTID, which is the correct response for an NFSv4.0 SetClientIDConfirm but is not the correct error for NFSv4.1/4.2, which is specified as NFSERR_SEQMISORDERED in RFC8881. This patch fixes this. This change does not fix the issue reported in the PR, where the ESXi client loops, attempting ExchangeID/Create_session repeatedly. Reported by: asomers Tested by: asomers PR: 261291 MFC after: 1 week
2022-02-09 23:17:50 +00:00
error = NFSERR_SEQMISORDERED;
nfsd: Fix CreateSession for an established ClientID I mis-read the RFC w.r.t. handling of the sequenceid when a CreateSession is done after the initial one that confirms the ClientID. Fortunately this does not affect most extant NFSv4.1/4.2 clients, since they only acquire a single session for TCP for a ClientID (Solaris might be an exception?). This patch fixes the server to handle this case, where the RFC requires the sequenceid be incremented for each CreateSession and is required to reply to a retried CreateSession with a cached reply. It adds a field to nfsclient called lc_prevsess, which caches the sessionid, which is the only field in a CreateSession reply that will change for a retry, to implement this reply cache. The recent commits up to d4a11b3e3bdd that mark session slots bad when "intr" and/or "soft" mounts are used by the client needs this server patch. Without this patch, the client will do a full recovery, including a new ClientID, losing all byte range locks. However, prior to the recent client commits, the client would hang when all session slots were bad, so even without this patch it is not a regression. PR: 260011 MFC after: 2 weeks
2022-07-13 23:28:56 +00:00
} else if ((nd->nd_flag & ND_NFSV41) == 0 &&
nfsd: Reply NFSERR_SEQMISORDERED for bogus seqid argument The ESXi NFSv4.1 client bogusly sends the wrong value for the csa_sequence argument for a Create_session operation. RFC8881 requires this value to be the same as the sequence reply from the ExchangeID operation most recently done for the client ID. Without this patch, the server replies NFSERR_STALECLIENTID, which is the correct response for an NFSv4.0 SetClientIDConfirm but is not the correct error for NFSv4.1/4.2, which is specified as NFSERR_SEQMISORDERED in RFC8881. This patch fixes this. This change does not fix the issue reported in the PR, where the ESXi client loops, attempting ExchangeID/Create_session repeatedly. Reported by: asomers Tested by: asomers PR: 261291 MFC after: 1 week
2022-02-09 23:17:50 +00:00
clp->lc_confirm.qval != confirm.qval)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
error = NFSERR_STALECLIENTID;
nfsd: Add support for the SP4_MACH_CRED case in ExchangeID Commit f4179ad46fa4 added support for operation bitmaps for NFSv4.1/4.2. This commit uses those to implement the SP4_MACH_CRED case for the NFSv4.1/4.2 ExchangeID operation since the Linux NFSv4.1/4.2 client is now using this for Kerberized mounts. The Linux Kerberized NFSv4.1/4.2 mounts currently work without support for this because Linux will fall back to SP4_NONE, but there is no guarantee this fallback will work forever. This commit only affects Kerberized NFSv4.1/4.2 mounts from Linux at this time. MFC after: 3 months
2023-04-07 19:49:23 +00:00
if (error == 0 && nfsrv_notsamecredname(NFSV4OP_CREATESESSION,
nd, clp))
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
error = NFSERR_CLIDINUSE;
if (!error) {
if ((clp->lc_flags & (LCL_NEEDSCONFIRM | LCL_DONTCLEAN)) ==
LCL_NEEDSCONFIRM) {
/*
* Hang onto the delegations (as old delegations)
* for an Open with CLAIM_DELEGATE_PREV unless in
* grace, but get rid of the rest of the state.
*/
nfsrv_cleanclient(clp, p);
nfsrv_freedeleglist(&clp->lc_olddeleg);
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
if (nfsrv_checkgrace(nd, clp, 0)) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/* In grace, so just delete delegations */
nfsrv_freedeleglist(&clp->lc_deleg);
} else {
LIST_FOREACH(stp, &clp->lc_deleg, ls_list)
stp->ls_flags |= NFSLCK_OLDDELEG;
clp->lc_delegtime = NFSD_MONOSEC +
nfsrv_lease + NFSRV_LEASEDELTA;
LIST_NEWHEAD(&clp->lc_olddeleg, &clp->lc_deleg,
ls_list);
}
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
if ((nd->nd_flag & ND_NFSV41) != 0)
clp->lc_program = cbprogram;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
clp->lc_flags &= ~(LCL_NEEDSCONFIRM | LCL_DONTCLEAN);
if (clp->lc_program)
clp->lc_flags |= LCL_NEEDSCBNULL;
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
/* For NFSv4.1, link the session onto the client. */
if (nsep != NULL) {
/* Hold a reference on the xprt for a backchannel. */
if ((nsep->sess_crflags & NFSV4CRSESS_CONNBACKCHAN)
nfsd: Fix CreateSession for an established ClientID I mis-read the RFC w.r.t. handling of the sequenceid when a CreateSession is done after the initial one that confirms the ClientID. Fortunately this does not affect most extant NFSv4.1/4.2 clients, since they only acquire a single session for TCP for a ClientID (Solaris might be an exception?). This patch fixes the server to handle this case, where the RFC requires the sequenceid be incremented for each CreateSession and is required to reply to a retried CreateSession with a cached reply. It adds a field to nfsclient called lc_prevsess, which caches the sessionid, which is the only field in a CreateSession reply that will change for a retry, to implement this reply cache. The recent commits up to d4a11b3e3bdd that mark session slots bad when "intr" and/or "soft" mounts are used by the client needs this server patch. Without this patch, the client will do a full recovery, including a new ClientID, losing all byte range locks. However, prior to the recent client commits, the client would hang when all session slots were bad, so even without this patch it is not a regression. PR: 260011 MFC after: 2 weeks
2022-07-13 23:28:56 +00:00
!= 0 && !sess_replay) {
Add the BindConnectiontoSession operation to the NFSv4.1 server. Under some fairly unusual circumstances, the Linux NFSv4.1 client is doing a BindConnectiontoSession operation for TCP connections. It is also used by the ESXi6.5 NFSv4.1 client. This patch adds this operation to the NFSv4.1 server. Reported by: andreas.nagy@frequentis.com Tested by: andreas.nagy@frequentis.com MFC after: 2 weeks
2018-06-01 19:47:41 +00:00
if (clp->lc_req.nr_client == NULL)
clp->lc_req.nr_client = (struct __rpc_client *)
clnt_bck_create(nd->nd_xprt->xp_socket,
cbprogram, NFSV4_CBVERS);
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
if (clp->lc_req.nr_client != NULL) {
SVC_ACQUIRE(nd->nd_xprt);
krpc: Acquire ref count of CLIENT for backchannel use Michael Dexter <editor@callfortesting.org> reported a crash in FreeNAS, where the first argument to clnt_bck_svccall() was no longer valid. This argument is a pointer to the callback CLIENT structure, which is free'd when the associated NFSv4 ClientID is free'd. This appears to have occurred because a callback reply was still in the socket receive queue when the CLIENT structure was free'd. This patch acquires a reference count on the CLIENT that is not CLNT_RELEASE()'d until the socket structure is destroyed. This should guarantee that the CLIENT structure is still valid when clnt_bck_svccall() is called. It also adds a check for closed or closing to clnt_bck_svccall() so that it will not process the callback RPC reply message after the ClientID is free'd. Comments by: mav MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D30153
2021-06-11 23:57:14 +00:00
CLNT_ACQUIRE(clp->lc_req.nr_client);
nd->nd_xprt->xp_p2 = clp->lc_req.nr_client;
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
/* Disable idle timeout. */
nd->nd_xprt->xp_idletimeout = 0;
nsep->sess_cbsess.nfsess_xprt = nd->nd_xprt;
} else
nsep->sess_crflags &= ~NFSV4CRSESS_CONNBACKCHAN;
}
NFSBCOPY(sessid, nsep->sess_sessionid,
NFSX_V4SESSIONID);
NFSBCOPY(sessid, nsep->sess_cbsess.nfsess_sessionid,
NFSX_V4SESSIONID);
nfsd: Fix CreateSession for an established ClientID I mis-read the RFC w.r.t. handling of the sequenceid when a CreateSession is done after the initial one that confirms the ClientID. Fortunately this does not affect most extant NFSv4.1/4.2 clients, since they only acquire a single session for TCP for a ClientID (Solaris might be an exception?). This patch fixes the server to handle this case, where the RFC requires the sequenceid be incremented for each CreateSession and is required to reply to a retried CreateSession with a cached reply. It adds a field to nfsclient called lc_prevsess, which caches the sessionid, which is the only field in a CreateSession reply that will change for a retry, to implement this reply cache. The recent commits up to d4a11b3e3bdd that mark session slots bad when "intr" and/or "soft" mounts are used by the client needs this server patch. Without this patch, the client will do a full recovery, including a new ClientID, losing all byte range locks. However, prior to the recent client commits, the client would hang when all session slots were bad, so even without this patch it is not a regression. PR: 260011 MFC after: 2 weeks
2022-07-13 23:28:56 +00:00
if (!sess_replay) {
shp = NFSSESSIONHASH(nsep->sess_sessionid);
NFSLOCKSTATE();
NFSLOCKSESSION(shp);
LIST_INSERT_HEAD(&shp->list, nsep, sess_hash);
LIST_INSERT_HEAD(&clp->lc_session, nsep, sess_list);
nsep->sess_clp = clp;
NFSUNLOCKSESSION(shp);
NFSUNLOCKSTATE();
}
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
} else if (clp->lc_flags & LCL_NEEDSCONFIRM) {
error = NFSERR_EXPIRED;
}
/*
* If called by the Renew Op, we must check the principal.
*/
if (!error && (opflags & CLOPS_RENEWOP)) {
nfsd: Add support for the SP4_MACH_CRED case in ExchangeID Commit f4179ad46fa4 added support for operation bitmaps for NFSv4.1/4.2. This commit uses those to implement the SP4_MACH_CRED case for the NFSv4.1/4.2 ExchangeID operation since the Linux NFSv4.1/4.2 client is now using this for Kerberized mounts. The Linux Kerberized NFSv4.1/4.2 mounts currently work without support for this because Linux will fall back to SP4_NONE, but there is no guarantee this fallback will work forever. This commit only affects Kerberized NFSv4.1/4.2 mounts from Linux at this time. MFC after: 3 months
2023-04-07 19:49:23 +00:00
if (nfsrv_notsamecredname(0, nd, clp)) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
doneok = 0;
Make the size of the hash tables used by the NFSv4 server tunable. No appreciable change in performance was observed after increasing the sizes of these tables and then testing with a single client. However, there was an email that indicated high CPU overheads for a heavily loaded NFSv4 and it is hoped that increasing the sizes of the hash tables via these tunables might help. The tables remain the same size by default. Differential Revision: https://reviews.freebsd.org/D2596 MFC after: 2 weeks
2015-05-27 22:00:05 +00:00
for (i = 0; i < nfsrv_statehashsize && doneok == 0; i++) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
LIST_FOREACH(stp, &clp->lc_stateid[i], ls_hash) {
if ((stp->ls_flags & NFSLCK_OPEN) &&
stp->ls_uid == nd->nd_cred->cr_uid) {
doneok = 1;
break;
}
}
}
if (!doneok)
error = NFSERR_ACCES;
}
if (!error && (clp->lc_flags & LCL_CBDOWN))
error = NFSERR_CBPATHDOWN;
}
if ((!error || error == NFSERR_CBPATHDOWN) &&
(opflags & CLOPS_RENEW)) {
clp->lc_expiry = nfsrv_leaseexpiry();
}
if (opflags & CLOPS_CONFIRM) {
NFSLOCKV4ROOTMUTEX();
nfsv4_unlock(&nfsv4rootfs_lock, 1);
NFSUNLOCKV4ROOTMUTEX();
} else if (opflags != CLOPS_RENEW) {
NFSUNLOCKSTATE();
}
if (clpp)
*clpp = clp;
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
out:
NFSEXITCODE2(error, nd);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
return (error);
}
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
/*
* Perform the NFSv4.1 destroy clientid.
*/
int
nfsd: Add support for the SP4_MACH_CRED case in ExchangeID Commit f4179ad46fa4 added support for operation bitmaps for NFSv4.1/4.2. This commit uses those to implement the SP4_MACH_CRED case for the NFSv4.1/4.2 ExchangeID operation since the Linux NFSv4.1/4.2 client is now using this for Kerberized mounts. The Linux Kerberized NFSv4.1/4.2 mounts currently work without support for this because Linux will fall back to SP4_NONE, but there is no guarantee this fallback will work forever. This commit only affects Kerberized NFSv4.1/4.2 mounts from Linux at this time. MFC after: 3 months
2023-04-07 19:49:23 +00:00
nfsrv_destroyclient(struct nfsrv_descript *nd, nfsquad_t clientid, NFSPROC_T *p)
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
{
struct nfsclient *clp;
struct nfsclienthashhead *hp;
int error = 0, i, igotlock;
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
if (NFSD_VNET(nfsrvboottime) != clientid.lval[0]) {
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
error = NFSERR_STALECLIENTID;
goto out;
}
/* Lock out other nfsd threads */
NFSLOCKV4ROOTMUTEX();
nfsv4_relref(&nfsv4rootfs_lock);
do {
igotlock = nfsv4_lock(&nfsv4rootfs_lock, 1, NULL,
NFSV4ROOTLOCKMUTEXPTR, NULL);
} while (igotlock == 0);
NFSUNLOCKV4ROOTMUTEX();
hp = NFSCLIENTHASH(clientid);
LIST_FOREACH(clp, hp, lc_hash) {
if (clp->lc_clientid.lval[1] == clientid.lval[1])
break;
}
if (clp == NULL) {
NFSLOCKV4ROOTMUTEX();
nfsv4_unlock(&nfsv4rootfs_lock, 1);
NFSUNLOCKV4ROOTMUTEX();
/* Just return ok, since it is gone. */
goto out;
}
nfsd: Add support for the SP4_MACH_CRED case in ExchangeID Commit f4179ad46fa4 added support for operation bitmaps for NFSv4.1/4.2. This commit uses those to implement the SP4_MACH_CRED case for the NFSv4.1/4.2 ExchangeID operation since the Linux NFSv4.1/4.2 client is now using this for Kerberized mounts. The Linux Kerberized NFSv4.1/4.2 mounts currently work without support for this because Linux will fall back to SP4_NONE, but there is no guarantee this fallback will work forever. This commit only affects Kerberized NFSv4.1/4.2 mounts from Linux at this time. MFC after: 3 months
2023-04-07 19:49:23 +00:00
/* Check for the SP4_MACH_CRED case. */
error = nfsrv_checkmachcred(NFSV4OP_DESTROYCLIENTID, nd, clp);
if (error != 0) {
NFSLOCKV4ROOTMUTEX();
nfsv4_unlock(&nfsv4rootfs_lock, 1);
NFSUNLOCKV4ROOTMUTEX();
goto out;
}
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
/*
* Free up all layouts on the clientid. Should the client return the
* layouts?
*/
nfsrv_freelayoutlist(clientid);
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
/* Scan for state on the clientid. */
Make the size of the hash tables used by the NFSv4 server tunable. No appreciable change in performance was observed after increasing the sizes of these tables and then testing with a single client. However, there was an email that indicated high CPU overheads for a heavily loaded NFSv4 and it is hoped that increasing the sizes of the hash tables via these tunables might help. The tables remain the same size by default. Differential Revision: https://reviews.freebsd.org/D2596 MFC after: 2 weeks
2015-05-27 22:00:05 +00:00
for (i = 0; i < nfsrv_statehashsize; i++)
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
if (!LIST_EMPTY(&clp->lc_stateid[i])) {
NFSLOCKV4ROOTMUTEX();
nfsv4_unlock(&nfsv4rootfs_lock, 1);
NFSUNLOCKV4ROOTMUTEX();
error = NFSERR_CLIENTIDBUSY;
goto out;
}
if (!LIST_EMPTY(&clp->lc_session) || !LIST_EMPTY(&clp->lc_deleg)) {
NFSLOCKV4ROOTMUTEX();
nfsv4_unlock(&nfsv4rootfs_lock, 1);
NFSUNLOCKV4ROOTMUTEX();
error = NFSERR_CLIENTIDBUSY;
goto out;
}
/* Destroy the clientid and return ok. */
nfsrv_cleanclient(clp, p);
nfsrv_freedeleglist(&clp->lc_deleg);
nfsrv_freedeleglist(&clp->lc_olddeleg);
LIST_REMOVE(clp, lc_hash);
NFSLOCKV4ROOTMUTEX();
nfsv4_unlock(&nfsv4rootfs_lock, 1);
NFSUNLOCKV4ROOTMUTEX();
nfsrv_zapclient(clp, p);
out:
NFSEXITCODE2(error, nd);
return (error);
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* Called from the new nfssvc syscall to admin revoke a clientid.
* Returns 0 for success, error otherwise.
*/
nfs: Remove APPLESTATIC macro It is no longer useful. Reviewed by: rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24811
2020-05-12 13:23:25 +00:00
int
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_adminrevoke(struct nfsd_clid *revokep, NFSPROC_T *p)
{
struct nfsclient *clp = NULL;
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
int i, error = 0;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
int gotit, igotlock;
/*
* First, lock out the nfsd so that state won't change while the
* revocation record is being written to the stable storage restart
* file.
*/
NFSLOCKV4ROOTMUTEX();
do {
igotlock = nfsv4_lock(&nfsv4rootfs_lock, 1, NULL,
Fix the new NFS client so that it handles NFSv4 state correctly during a forced dismount. This required that the exclusive and shared (refcnt) sleep lock functions check for MNTK_UMOUNTF before sleeping, so that they won't block while nfscl_umount() is getting rid of the state. As such, a "struct mount *" argument was added to the locking functions. I believe the only remaining case where a forced dismount can get hung in the kernel is when a thread is already attempting to do a TCP connect to a dead server when the krpc client structure called nr_client is NULL. This will only happen just after a "mount -u" with options that force a new TCP connection is done, so it shouldn't be a problem in practice. MFC after: 2 weeks
2011-05-27 22:05:10 +00:00
NFSV4ROOTLOCKMUTEXPTR, NULL);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
} while (!igotlock);
NFSUNLOCKV4ROOTMUTEX();
/*
* Search for a match in the client list.
*/
gotit = i = 0;
Make the size of the hash tables used by the NFSv4 server tunable. No appreciable change in performance was observed after increasing the sizes of these tables and then testing with a single client. However, there was an email that indicated high CPU overheads for a heavily loaded NFSv4 and it is hoped that increasing the sizes of the hash tables via these tunables might help. The tables remain the same size by default. Differential Revision: https://reviews.freebsd.org/D2596 MFC after: 2 weeks
2015-05-27 22:00:05 +00:00
while (i < nfsrv_clienthashsize && !gotit) {
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
LIST_FOREACH(clp, &NFSD_VNET(nfsclienthash)[i], lc_hash) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (revokep->nclid_idlen == clp->lc_idlen &&
!NFSBCMP(revokep->nclid_id, clp->lc_id, clp->lc_idlen)) {
gotit = 1;
break;
}
}
i++;
}
if (!gotit) {
NFSLOCKV4ROOTMUTEX();
nfsv4_unlock(&nfsv4rootfs_lock, 0);
NFSUNLOCKV4ROOTMUTEX();
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
error = EPERM;
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* Now, write out the revocation record
*/
nfsrv_writestable(clp->lc_id, clp->lc_idlen, NFSNST_REVOKE, p);
Modify the experimental NFSv4 server so that it posts a SIGUSR2 signal to the master nfsd daemon whenever the stable restart file has been modified. This will allow the master nfsd daemon to maintain an up to date backup copy of the file. This is enabled via the nfssvc() syscall, so that older nfsd daemons will not be signaled. Reviewed by: jhb MFC after: 1 week
2011-01-14 23:30:35 +00:00
nfsrv_backupstable();
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* and clear out the state, marking the clientid revoked.
*/
clp->lc_flags &= ~LCL_CALLBACKSON;
clp->lc_flags |= LCL_ADMINREVOKED;
nfsrv_cleanclient(clp, p);
nfsrv_freedeleglist(&clp->lc_deleg);
nfsrv_freedeleglist(&clp->lc_olddeleg);
NFSLOCKV4ROOTMUTEX();
nfsv4_unlock(&nfsv4rootfs_lock, 0);
NFSUNLOCKV4ROOTMUTEX();
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
out:
NFSEXITCODE(error);
return (error);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* Dump out stats for all clients. Called from nfssvc(2), that is used
Update the nfsstats structure to include the changes needed by the patch in D1626 plus changes so that it includes counts for NFSv4.1 (and the draft of NFSv4.2). Also, make all the counts uint64_t and add a vers field at the beginning, so that future revisions can easily be implemented. There is code in place to handle the old vesion of the nfsstats structure for backwards binary compatibility. Subsequent commits will update nfsstat(8) to use the new fields. Submitted by: will (earlier version) Reviewed by: ken MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D1626
2016-08-12 22:44:59 +00:00
* nfsstatsv1.
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
*/
nfs: Remove APPLESTATIC macro It is no longer useful. Reviewed by: rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24811
2020-05-12 13:23:25 +00:00
void
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_dumpclients(struct nfsd_dumpclients *dumpp, int maxcnt)
{
struct nfsclient *clp;
int i = 0, cnt = 0;
Patch the experimental NFSv4 server so that it acquires a reference count on nfsv4rootfs_lock when dumping state, since these functions are not called by nfsd threads. Without this reference count, it is possible for an nfsd thread to acquire an exclusive lock on nfsv4rootfs_lock while the dump is in progress and then change the lists, potentially causing a crash. Reported by: zack.kirsch at isilon.com MFC after: 2 weeks
2010-07-16 23:17:05 +00:00
/*
* First, get a reference on the nfsv4rootfs_lock so that an
* exclusive lock cannot be acquired while dumping the clients.
*/
NFSLOCKV4ROOTMUTEX();
Fix the new NFS client so that it handles NFSv4 state correctly during a forced dismount. This required that the exclusive and shared (refcnt) sleep lock functions check for MNTK_UMOUNTF before sleeping, so that they won't block while nfscl_umount() is getting rid of the state. As such, a "struct mount *" argument was added to the locking functions. I believe the only remaining case where a forced dismount can get hung in the kernel is when a thread is already attempting to do a TCP connect to a dead server when the krpc client structure called nr_client is NULL. This will only happen just after a "mount -u" with options that force a new TCP connection is done, so it shouldn't be a problem in practice. MFC after: 2 weeks
2011-05-27 22:05:10 +00:00
nfsv4_getref(&nfsv4rootfs_lock, NULL, NFSV4ROOTLOCKMUTEXPTR, NULL);
Patch the experimental NFSv4 server so that it acquires a reference count on nfsv4rootfs_lock when dumping state, since these functions are not called by nfsd threads. Without this reference count, it is possible for an nfsd thread to acquire an exclusive lock on nfsv4rootfs_lock while the dump is in progress and then change the lists, potentially causing a crash. Reported by: zack.kirsch at isilon.com MFC after: 2 weeks
2010-07-16 23:17:05 +00:00
NFSUNLOCKV4ROOTMUTEX();
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
NFSLOCKSTATE();
/*
* Rattle through the client lists until done.
*/
Make the size of the hash tables used by the NFSv4 server tunable. No appreciable change in performance was observed after increasing the sizes of these tables and then testing with a single client. However, there was an email that indicated high CPU overheads for a heavily loaded NFSv4 and it is hoped that increasing the sizes of the hash tables via these tunables might help. The tables remain the same size by default. Differential Revision: https://reviews.freebsd.org/D2596 MFC after: 2 weeks
2015-05-27 22:00:05 +00:00
while (i < nfsrv_clienthashsize && cnt < maxcnt) {
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
clp = LIST_FIRST(&NFSD_VNET(nfsclienthash)[i]);
while (clp != LIST_END(&NFSD_VNET(nfsclienthash)[i]) && cnt <
maxcnt) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_dumpaclient(clp, &dumpp[cnt]);
cnt++;
clp = LIST_NEXT(clp, lc_hash);
}
i++;
}
if (cnt < maxcnt)
dumpp[cnt].ndcl_clid.nclid_idlen = 0;
NFSUNLOCKSTATE();
Patch the experimental NFSv4 server so that it acquires a reference count on nfsv4rootfs_lock when dumping state, since these functions are not called by nfsd threads. Without this reference count, it is possible for an nfsd thread to acquire an exclusive lock on nfsv4rootfs_lock while the dump is in progress and then change the lists, potentially causing a crash. Reported by: zack.kirsch at isilon.com MFC after: 2 weeks
2010-07-16 23:17:05 +00:00
NFSLOCKV4ROOTMUTEX();
nfsv4_relref(&nfsv4rootfs_lock);
NFSUNLOCKV4ROOTMUTEX();
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* Dump stats for a client. Must be called with the NFSSTATELOCK and spl'd.
*/
static void
nfsrv_dumpaclient(struct nfsclient *clp, struct nfsd_dumpclients *dumpp)
{
struct nfsstate *stp, *openstp, *lckownstp;
struct nfslock *lop;
Add support for INET6 addresses to the kernel code that dumps open/lock state. PR#223036 reported that INET6 callback addresses were not printed by nfsdumpstate(8). This kernel patch adds INET6 addresses to the dump structure, so that nfsdumpstate(8) can print them out, post-r346190. The patch also includes the addition of #ifdef INET, INET6 as requested by bz@. PR: 223036 Reviewed by: bz, rgrimes MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D19839
2019-04-13 22:00:09 +00:00
sa_family_t af;
#ifdef INET
struct sockaddr_in *rin;
#endif
#ifdef INET6
struct sockaddr_in6 *rin6;
#endif
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
dumpp->ndcl_nopenowners = dumpp->ndcl_nlockowners = 0;
dumpp->ndcl_nopens = dumpp->ndcl_nlocks = 0;
dumpp->ndcl_ndelegs = dumpp->ndcl_nolddelegs = 0;
dumpp->ndcl_flags = clp->lc_flags;
dumpp->ndcl_clid.nclid_idlen = clp->lc_idlen;
NFSBCOPY(clp->lc_id, dumpp->ndcl_clid.nclid_id, clp->lc_idlen);
Add support for INET6 addresses to the kernel code that dumps open/lock state. PR#223036 reported that INET6 callback addresses were not printed by nfsdumpstate(8). This kernel patch adds INET6 addresses to the dump structure, so that nfsdumpstate(8) can print them out, post-r346190. The patch also includes the addition of #ifdef INET, INET6 as requested by bz@. PR: 223036 Reviewed by: bz, rgrimes MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D19839
2019-04-13 22:00:09 +00:00
af = clp->lc_req.nr_nam->sa_family;
dumpp->ndcl_addrfam = af;
switch (af) {
#ifdef INET
case AF_INET:
rin = (struct sockaddr_in *)clp->lc_req.nr_nam;
dumpp->ndcl_cbaddr.sin_addr = rin->sin_addr;
break;
#endif
#ifdef INET6
case AF_INET6:
rin6 = (struct sockaddr_in6 *)clp->lc_req.nr_nam;
dumpp->ndcl_cbaddr.sin6_addr = rin6->sin6_addr;
break;
#endif
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* Now, scan the state lists and total up the opens and locks.
*/
LIST_FOREACH(stp, &clp->lc_open, ls_list) {
dumpp->ndcl_nopenowners++;
LIST_FOREACH(openstp, &stp->ls_open, ls_list) {
dumpp->ndcl_nopens++;
LIST_FOREACH(lckownstp, &openstp->ls_open, ls_list) {
dumpp->ndcl_nlockowners++;
LIST_FOREACH(lop, &lckownstp->ls_lock, lo_lckowner) {
dumpp->ndcl_nlocks++;
}
}
}
}
/*
* and the delegation lists.
*/
LIST_FOREACH(stp, &clp->lc_deleg, ls_list) {
dumpp->ndcl_ndelegs++;
}
LIST_FOREACH(stp, &clp->lc_olddeleg, ls_list) {
dumpp->ndcl_nolddelegs++;
}
}
/*
* Dump out lock stats for a file.
*/
nfs: Remove APPLESTATIC macro It is no longer useful. Reviewed by: rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24811
2020-05-12 13:23:25 +00:00
void
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_dumplocks(vnode_t vp, struct nfsd_dumplocks *ldumpp, int maxcnt,
NFSPROC_T *p)
{
struct nfsstate *stp;
struct nfslock *lop;
int cnt = 0;
struct nfslockfile *lfp;
Add support for INET6 addresses to the kernel code that dumps open/lock state. PR#223036 reported that INET6 callback addresses were not printed by nfsdumpstate(8). This kernel patch adds INET6 addresses to the dump structure, so that nfsdumpstate(8) can print them out, post-r346190. The patch also includes the addition of #ifdef INET, INET6 as requested by bz@. PR: 223036 Reviewed by: bz, rgrimes MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D19839
2019-04-13 22:00:09 +00:00
sa_family_t af;
#ifdef INET
struct sockaddr_in *rin;
#endif
#ifdef INET6
struct sockaddr_in6 *rin6;
#endif
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
int ret;
fhandle_t nfh;
ret = nfsrv_getlockfh(vp, 0, NULL, &nfh, p);
Patch the experimental NFSv4 server so that it acquires a reference count on nfsv4rootfs_lock when dumping state, since these functions are not called by nfsd threads. Without this reference count, it is possible for an nfsd thread to acquire an exclusive lock on nfsv4rootfs_lock while the dump is in progress and then change the lists, potentially causing a crash. Reported by: zack.kirsch at isilon.com MFC after: 2 weeks
2010-07-16 23:17:05 +00:00
/*
* First, get a reference on the nfsv4rootfs_lock so that an
* exclusive lock on it cannot be acquired while dumping the locks.
*/
NFSLOCKV4ROOTMUTEX();
Fix the new NFS client so that it handles NFSv4 state correctly during a forced dismount. This required that the exclusive and shared (refcnt) sleep lock functions check for MNTK_UMOUNTF before sleeping, so that they won't block while nfscl_umount() is getting rid of the state. As such, a "struct mount *" argument was added to the locking functions. I believe the only remaining case where a forced dismount can get hung in the kernel is when a thread is already attempting to do a TCP connect to a dead server when the krpc client structure called nr_client is NULL. This will only happen just after a "mount -u" with options that force a new TCP connection is done, so it shouldn't be a problem in practice. MFC after: 2 weeks
2011-05-27 22:05:10 +00:00
nfsv4_getref(&nfsv4rootfs_lock, NULL, NFSV4ROOTLOCKMUTEXPTR, NULL);
Patch the experimental NFSv4 server so that it acquires a reference count on nfsv4rootfs_lock when dumping state, since these functions are not called by nfsd threads. Without this reference count, it is possible for an nfsd thread to acquire an exclusive lock on nfsv4rootfs_lock while the dump is in progress and then change the lists, potentially causing a crash. Reported by: zack.kirsch at isilon.com MFC after: 2 weeks
2010-07-16 23:17:05 +00:00
NFSUNLOCKV4ROOTMUTEX();
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
NFSLOCKSTATE();
if (!ret)
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
ret = nfsrv_getlockfile(0, NULL, &lfp, &nfh, 0);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (ret) {
ldumpp[0].ndlck_clid.nclid_idlen = 0;
NFSUNLOCKSTATE();
Patch the experimental NFSv4 server so that it acquires a reference count on nfsv4rootfs_lock when dumping state, since these functions are not called by nfsd threads. Without this reference count, it is possible for an nfsd thread to acquire an exclusive lock on nfsv4rootfs_lock while the dump is in progress and then change the lists, potentially causing a crash. Reported by: zack.kirsch at isilon.com MFC after: 2 weeks
2010-07-16 23:17:05 +00:00
NFSLOCKV4ROOTMUTEX();
nfsv4_relref(&nfsv4rootfs_lock);
NFSUNLOCKV4ROOTMUTEX();
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
return;
}
/*
* For each open share on file, dump it out.
*/
stp = LIST_FIRST(&lfp->lf_open);
while (stp != LIST_END(&lfp->lf_open) && cnt < maxcnt) {
ldumpp[cnt].ndlck_flags = stp->ls_flags;
ldumpp[cnt].ndlck_stateid.seqid = stp->ls_stateid.seqid;
ldumpp[cnt].ndlck_stateid.other[0] = stp->ls_stateid.other[0];
ldumpp[cnt].ndlck_stateid.other[1] = stp->ls_stateid.other[1];
ldumpp[cnt].ndlck_stateid.other[2] = stp->ls_stateid.other[2];
ldumpp[cnt].ndlck_owner.nclid_idlen =
stp->ls_openowner->ls_ownerlen;
NFSBCOPY(stp->ls_openowner->ls_owner,
ldumpp[cnt].ndlck_owner.nclid_id,
stp->ls_openowner->ls_ownerlen);
ldumpp[cnt].ndlck_clid.nclid_idlen = stp->ls_clp->lc_idlen;
NFSBCOPY(stp->ls_clp->lc_id, ldumpp[cnt].ndlck_clid.nclid_id,
stp->ls_clp->lc_idlen);
Add support for INET6 addresses to the kernel code that dumps open/lock state. PR#223036 reported that INET6 callback addresses were not printed by nfsdumpstate(8). This kernel patch adds INET6 addresses to the dump structure, so that nfsdumpstate(8) can print them out, post-r346190. The patch also includes the addition of #ifdef INET, INET6 as requested by bz@. PR: 223036 Reviewed by: bz, rgrimes MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D19839
2019-04-13 22:00:09 +00:00
af = stp->ls_clp->lc_req.nr_nam->sa_family;
ldumpp[cnt].ndlck_addrfam = af;
switch (af) {
#ifdef INET
case AF_INET:
rin = (struct sockaddr_in *)stp->ls_clp->lc_req.nr_nam;
ldumpp[cnt].ndlck_cbaddr.sin_addr = rin->sin_addr;
break;
#endif
#ifdef INET6
case AF_INET6:
rin6 = (struct sockaddr_in6 *)
stp->ls_clp->lc_req.nr_nam;
ldumpp[cnt].ndlck_cbaddr.sin6_addr = rin6->sin6_addr;
break;
#endif
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
stp = LIST_NEXT(stp, ls_file);
cnt++;
}
/*
* and all locks.
*/
lop = LIST_FIRST(&lfp->lf_lock);
while (lop != LIST_END(&lfp->lf_lock) && cnt < maxcnt) {
stp = lop->lo_stp;
ldumpp[cnt].ndlck_flags = lop->lo_flags;
ldumpp[cnt].ndlck_first = lop->lo_first;
ldumpp[cnt].ndlck_end = lop->lo_end;
ldumpp[cnt].ndlck_stateid.seqid = stp->ls_stateid.seqid;
ldumpp[cnt].ndlck_stateid.other[0] = stp->ls_stateid.other[0];
ldumpp[cnt].ndlck_stateid.other[1] = stp->ls_stateid.other[1];
ldumpp[cnt].ndlck_stateid.other[2] = stp->ls_stateid.other[2];
ldumpp[cnt].ndlck_owner.nclid_idlen = stp->ls_ownerlen;
NFSBCOPY(stp->ls_owner, ldumpp[cnt].ndlck_owner.nclid_id,
stp->ls_ownerlen);
ldumpp[cnt].ndlck_clid.nclid_idlen = stp->ls_clp->lc_idlen;
NFSBCOPY(stp->ls_clp->lc_id, ldumpp[cnt].ndlck_clid.nclid_id,
stp->ls_clp->lc_idlen);
Add support for INET6 addresses to the kernel code that dumps open/lock state. PR#223036 reported that INET6 callback addresses were not printed by nfsdumpstate(8). This kernel patch adds INET6 addresses to the dump structure, so that nfsdumpstate(8) can print them out, post-r346190. The patch also includes the addition of #ifdef INET, INET6 as requested by bz@. PR: 223036 Reviewed by: bz, rgrimes MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D19839
2019-04-13 22:00:09 +00:00
af = stp->ls_clp->lc_req.nr_nam->sa_family;
ldumpp[cnt].ndlck_addrfam = af;
switch (af) {
#ifdef INET
case AF_INET:
rin = (struct sockaddr_in *)stp->ls_clp->lc_req.nr_nam;
ldumpp[cnt].ndlck_cbaddr.sin_addr = rin->sin_addr;
break;
#endif
#ifdef INET6
case AF_INET6:
rin6 = (struct sockaddr_in6 *)
stp->ls_clp->lc_req.nr_nam;
ldumpp[cnt].ndlck_cbaddr.sin6_addr = rin6->sin6_addr;
break;
#endif
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
lop = LIST_NEXT(lop, lo_lckfile);
cnt++;
}
/*
* and the delegations.
*/
stp = LIST_FIRST(&lfp->lf_deleg);
while (stp != LIST_END(&lfp->lf_deleg) && cnt < maxcnt) {
ldumpp[cnt].ndlck_flags = stp->ls_flags;
ldumpp[cnt].ndlck_stateid.seqid = stp->ls_stateid.seqid;
ldumpp[cnt].ndlck_stateid.other[0] = stp->ls_stateid.other[0];
ldumpp[cnt].ndlck_stateid.other[1] = stp->ls_stateid.other[1];
ldumpp[cnt].ndlck_stateid.other[2] = stp->ls_stateid.other[2];
ldumpp[cnt].ndlck_owner.nclid_idlen = 0;
ldumpp[cnt].ndlck_clid.nclid_idlen = stp->ls_clp->lc_idlen;
NFSBCOPY(stp->ls_clp->lc_id, ldumpp[cnt].ndlck_clid.nclid_id,
stp->ls_clp->lc_idlen);
Add support for INET6 addresses to the kernel code that dumps open/lock state. PR#223036 reported that INET6 callback addresses were not printed by nfsdumpstate(8). This kernel patch adds INET6 addresses to the dump structure, so that nfsdumpstate(8) can print them out, post-r346190. The patch also includes the addition of #ifdef INET, INET6 as requested by bz@. PR: 223036 Reviewed by: bz, rgrimes MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D19839
2019-04-13 22:00:09 +00:00
af = stp->ls_clp->lc_req.nr_nam->sa_family;
ldumpp[cnt].ndlck_addrfam = af;
switch (af) {
#ifdef INET
case AF_INET:
rin = (struct sockaddr_in *)stp->ls_clp->lc_req.nr_nam;
ldumpp[cnt].ndlck_cbaddr.sin_addr = rin->sin_addr;
break;
#endif
#ifdef INET6
case AF_INET6:
rin6 = (struct sockaddr_in6 *)
stp->ls_clp->lc_req.nr_nam;
ldumpp[cnt].ndlck_cbaddr.sin6_addr = rin6->sin6_addr;
break;
#endif
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
stp = LIST_NEXT(stp, ls_file);
cnt++;
}
/*
* If list isn't full, mark end of list by setting the client name
* to zero length.
*/
if (cnt < maxcnt)
ldumpp[cnt].ndlck_clid.nclid_idlen = 0;
NFSUNLOCKSTATE();
Patch the experimental NFSv4 server so that it acquires a reference count on nfsv4rootfs_lock when dumping state, since these functions are not called by nfsd threads. Without this reference count, it is possible for an nfsd thread to acquire an exclusive lock on nfsv4rootfs_lock while the dump is in progress and then change the lists, potentially causing a crash. Reported by: zack.kirsch at isilon.com MFC after: 2 weeks
2010-07-16 23:17:05 +00:00
NFSLOCKV4ROOTMUTEX();
nfsv4_relref(&nfsv4rootfs_lock);
NFSUNLOCKV4ROOTMUTEX();
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* Server timer routine. It can scan any linked list, so long
The timer routine in the experimental NFS server did not acquire the correct mutex when checking nfsv4root_lock. Although this could be fixed by adding mutex lock/unlock calls, zack.kirsch at isilon.com suggested a better fix that uses a non-blocking acquisition of a reference count on nfsv4root_lock. This fix allows the weird NFSLOCKSTATE(); NFSUNLOCKSTATE(); synchronization to be deleted. This patch applies this fix. Tested by: zack.kirsch at isilon.com MFC after: 2 weeks
2010-08-28 21:41:18 +00:00
* as it holds the spin/mutex lock and there is no exclusive lock on
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
* nfsv4rootfs_lock.
* (For OpenBSD, a kthread is ok. For FreeBSD, I think it is ok
* to do this from a callout, since the spin locks work. For
* Darwin, I'm not sure what will work correctly yet.)
* Should be called once per second.
*/
nfs: Remove APPLESTATIC macro It is no longer useful. Reviewed by: rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24811
2020-05-12 13:23:25 +00:00
void
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
nfsrv_servertimer(void *arg __unused)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
{
struct nfsclient *clp, *nclp;
struct nfsstate *stp, *nstp;
The timer routine in the experimental NFS server did not acquire the correct mutex when checking nfsv4root_lock. Although this could be fixed by adding mutex lock/unlock calls, zack.kirsch at isilon.com suggested a better fix that uses a non-blocking acquisition of a reference count on nfsv4root_lock. This fix allows the weird NFSLOCKSTATE(); NFSUNLOCKSTATE(); synchronization to be deleted. This patch applies this fix. Tested by: zack.kirsch at isilon.com MFC after: 2 weeks
2010-08-28 21:41:18 +00:00
int got_ref, i;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* Make sure nfsboottime is set. This is used by V3 as well
* as V4. Note that nfsboottime is not nfsrvboottime, which is
* only used by the V4 server for leases.
*/
if (nfsboottime.tv_sec == 0)
NFSSETBOOTTIME(nfsboottime);
/*
* If server hasn't started yet, just return.
*/
NFSLOCKSTATE();
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
if (NFSD_VNET(nfsrv_stablefirst).nsf_eograce == 0) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
NFSUNLOCKSTATE();
return;
}
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
if (!(NFSD_VNET(nfsrv_stablefirst).nsf_flags & NFSNSF_UPDATEDONE)) {
if (!(NFSD_VNET(nfsrv_stablefirst).nsf_flags &
NFSNSF_GRACEOVER) &&
NFSD_MONOSEC > NFSD_VNET(nfsrv_stablefirst).nsf_eograce)
NFSD_VNET(nfsrv_stablefirst).nsf_flags |=
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
(NFSNSF_GRACEOVER | NFSNSF_NEEDLOCK);
NFSUNLOCKSTATE();
return;
}
/*
The timer routine in the experimental NFS server did not acquire the correct mutex when checking nfsv4root_lock. Although this could be fixed by adding mutex lock/unlock calls, zack.kirsch at isilon.com suggested a better fix that uses a non-blocking acquisition of a reference count on nfsv4root_lock. This fix allows the weird NFSLOCKSTATE(); NFSUNLOCKSTATE(); synchronization to be deleted. This patch applies this fix. Tested by: zack.kirsch at isilon.com MFC after: 2 weeks
2010-08-28 21:41:18 +00:00
* Try and get a reference count on the nfsv4rootfs_lock so that
* no nfsd thread can acquire an exclusive lock on it before this
* call is done. If it is already exclusively locked, just return.
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
*/
The timer routine in the experimental NFS server did not acquire the correct mutex when checking nfsv4root_lock. Although this could be fixed by adding mutex lock/unlock calls, zack.kirsch at isilon.com suggested a better fix that uses a non-blocking acquisition of a reference count on nfsv4root_lock. This fix allows the weird NFSLOCKSTATE(); NFSUNLOCKSTATE(); synchronization to be deleted. This patch applies this fix. Tested by: zack.kirsch at isilon.com MFC after: 2 weeks
2010-08-28 21:41:18 +00:00
NFSLOCKV4ROOTMUTEX();
got_ref = nfsv4_getref_nonblock(&nfsv4rootfs_lock);
NFSUNLOCKV4ROOTMUTEX();
if (got_ref == 0) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
NFSUNLOCKSTATE();
return;
}
/*
* For each client...
*/
Make the size of the hash tables used by the NFSv4 server tunable. No appreciable change in performance was observed after increasing the sizes of these tables and then testing with a single client. However, there was an email that indicated high CPU overheads for a heavily loaded NFSv4 and it is hoped that increasing the sizes of the hash tables via these tunables might help. The tables remain the same size by default. Differential Revision: https://reviews.freebsd.org/D2596 MFC after: 2 weeks
2015-05-27 22:00:05 +00:00
for (i = 0; i < nfsrv_clienthashsize; i++) {
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
clp = LIST_FIRST(&NFSD_VNET(nfsclienthash)[i]);
while (clp != LIST_END(&NFSD_VNET(nfsclienthash)[i])) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nclp = LIST_NEXT(clp, lc_hash);
if (!(clp->lc_flags & LCL_EXPIREIT)) {
if (((clp->lc_expiry + NFSRV_STALELEASE) < NFSD_MONOSEC
&& ((LIST_EMPTY(&clp->lc_deleg)
&& LIST_EMPTY(&clp->lc_open)) ||
nfsrv_clients > nfsrv_clienthighwater)) ||
(clp->lc_expiry + NFSRV_MOULDYLEASE) < NFSD_MONOSEC ||
(clp->lc_expiry < NFSD_MONOSEC &&
Make the size of the hash tables used by the NFSv4 server tunable. No appreciable change in performance was observed after increasing the sizes of these tables and then testing with a single client. However, there was an email that indicated high CPU overheads for a heavily loaded NFSv4 and it is hoped that increasing the sizes of the hash tables via these tunables might help. The tables remain the same size by default. Differential Revision: https://reviews.freebsd.org/D2596 MFC after: 2 weeks
2015-05-27 22:00:05 +00:00
(nfsrv_openpluslock * 10 / 9) > nfsrv_v4statelimit)) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* Lease has expired several nfsrv_lease times ago:
* PLUS
* - no state is associated with it
* OR
* - above high water mark for number of clients
* (nfsrv_clienthighwater should be large enough
* that this only occurs when clients fail to
* use the same nfs_client_id4.id. Maybe somewhat
* higher that the maximum number of clients that
* will mount this server?)
* OR
* Lease has expired a very long time ago
* OR
* Lease has expired PLUS the number of opens + locks
* has exceeded 90% of capacity
*
* --> Mark for expiry. The actual expiry will be done
* by an nfsd sometime soon.
*/
clp->lc_flags |= LCL_EXPIREIT;
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
NFSD_VNET(nfsrv_stablefirst).nsf_flags |=
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
(NFSNSF_NEEDLOCK | NFSNSF_EXPIREDCLIENT);
} else {
/*
* If there are no opens, increment no open tick cnt
* If time exceeds NFSNOOPEN, mark it to be thrown away
* otherwise, if there is an open, reset no open time
* Hopefully, this will avoid excessive re-creation
* of open owners and subsequent open confirms.
*/
stp = LIST_FIRST(&clp->lc_open);
while (stp != LIST_END(&clp->lc_open)) {
nstp = LIST_NEXT(stp, ls_list);
if (LIST_EMPTY(&stp->ls_open)) {
stp->ls_noopens++;
if (stp->ls_noopens > NFSNOOPEN ||
(nfsrv_openpluslock * 2) >
Make the size of the hash tables used by the NFSv4 server tunable. No appreciable change in performance was observed after increasing the sizes of these tables and then testing with a single client. However, there was an email that indicated high CPU overheads for a heavily loaded NFSv4 and it is hoped that increasing the sizes of the hash tables via these tunables might help. The tables remain the same size by default. Differential Revision: https://reviews.freebsd.org/D2596 MFC after: 2 weeks
2015-05-27 22:00:05 +00:00
nfsrv_v4statelimit)
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
NFSD_VNET(nfsrv_stablefirst).nsf_flags |=
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
NFSNSF_NOOPENS;
} else {
stp->ls_noopens = 0;
}
stp = nstp;
}
}
}
clp = nclp;
}
}
NFSUNLOCKSTATE();
The timer routine in the experimental NFS server did not acquire the correct mutex when checking nfsv4root_lock. Although this could be fixed by adding mutex lock/unlock calls, zack.kirsch at isilon.com suggested a better fix that uses a non-blocking acquisition of a reference count on nfsv4root_lock. This fix allows the weird NFSLOCKSTATE(); NFSUNLOCKSTATE(); synchronization to be deleted. This patch applies this fix. Tested by: zack.kirsch at isilon.com MFC after: 2 weeks
2010-08-28 21:41:18 +00:00
NFSLOCKV4ROOTMUTEX();
nfsv4_relref(&nfsv4rootfs_lock);
NFSUNLOCKV4ROOTMUTEX();
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* The following set of functions free up the various data structures.
*/
/*
* Clear out all open/lock state related to this nfsclient.
* Caller must hold an exclusive lock on nfsv4rootfs_lock, so that
* there are no other active nfsd threads.
*/
nfs: Remove APPLESTATIC macro It is no longer useful. Reviewed by: rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24811
2020-05-12 13:23:25 +00:00
void
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_cleanclient(struct nfsclient *clp, NFSPROC_T *p)
{
struct nfsstate *stp, *nstp;
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
struct nfsdsession *sep, *nsep;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
LIST_FOREACH_SAFE(stp, &clp->lc_open, ls_list, nstp)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_freeopenowner(stp, 1, p);
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
if ((clp->lc_flags & LCL_ADMINREVOKED) == 0)
LIST_FOREACH_SAFE(sep, &clp->lc_session, sess_list, nsep)
nfsd: Add support for the SP4_MACH_CRED case in ExchangeID Commit f4179ad46fa4 added support for operation bitmaps for NFSv4.1/4.2. This commit uses those to implement the SP4_MACH_CRED case for the NFSv4.1/4.2 ExchangeID operation since the Linux NFSv4.1/4.2 client is now using this for Kerberized mounts. The Linux Kerberized NFSv4.1/4.2 mounts currently work without support for this because Linux will fall back to SP4_NONE, but there is no guarantee this fallback will work forever. This commit only affects Kerberized NFSv4.1/4.2 mounts from Linux at this time. MFC after: 3 months
2023-04-07 19:49:23 +00:00
(void)nfsrv_freesession(NULL, sep, NULL);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* Free a client that has been cleaned. It should also already have been
* removed from the lists.
* (Just to be safe w.r.t. newnfs_disconnect(), call this function when
* softclock interrupts are enabled.)
*/
nfs: Remove APPLESTATIC macro It is no longer useful. Reviewed by: rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24811
2020-05-12 13:23:25 +00:00
void
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_zapclient(struct nfsclient *clp, NFSPROC_T *p)
{
#ifdef notyet
if ((clp->lc_flags & (LCL_GSS | LCL_CALLBACKSON)) ==
(LCL_GSS | LCL_CALLBACKSON) &&
(clp->lc_hand.nfsh_flag & NFSG_COMPLETE) &&
clp->lc_handlelen > 0) {
clp->lc_hand.nfsh_flag &= ~NFSG_COMPLETE;
clp->lc_hand.nfsh_flag |= NFSG_DESTROYED;
(void) nfsrv_docallback(clp, NFSV4PROC_CBNULL,
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
NULL, 0, NULL, NULL, NULL, 0, p);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
#endif
nfscl: Add a Linux compatible "nconnect" mount option Linux has had an "nconnect" NFS mount option for some time. It specifies that N (up to 16) TCP connections are to created for a mount, instead of just one TCP connection. A discussion on freebsd-net@ indicated that this could improve client<-->server network bandwidth, if either the client or server have one of the following: - multiple network ports aggregated to-gether with lagg/lacp. - a fast NIC that is using multiple queues It does result in using more IP port#s and might increase server peak load for a client. One difference from the Linux implementation is that this implementation uses the first TCP connection for all RPCs composed of small messages and uses the additional TCP connections for RPCs that normally have large messages (Read/Readdir/Write). The Linux implementation spreads all RPCs across all TCP connections in a round robin fashion, whereas this implementation spreads Read/Readdir/Write across the additional TCP connections in a round robin fashion. Reviewed by: markj MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D30970
2021-07-09 00:39:04 +00:00
newnfs_disconnect(NULL, &clp->lc_req);
nfs: Remove NFSSOCKADDRALLOC, NFSSOCKADDRFREE macros They were just thin wrappers over malloc(9) w/ M_ZERO and free(9). Discussed with: rmacklem, markj Sponsored by: Dell EMC Isilon
2018-01-25 22:38:39 +00:00
free(clp->lc_req.nr_nam, M_SONAME);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
NFSFREEMUTEX(&clp->lc_req.nr_mtx);
Make the size of the hash tables used by the NFSv4 server tunable. No appreciable change in performance was observed after increasing the sizes of these tables and then testing with a single client. However, there was an email that indicated high CPU overheads for a heavily loaded NFSv4 and it is hoped that increasing the sizes of the hash tables via these tunables might help. The tables remain the same size by default. Differential Revision: https://reviews.freebsd.org/D2596 MFC after: 2 weeks
2015-05-27 22:00:05 +00:00
free(clp->lc_stateid, M_NFSDCLIENT);
free(clp, M_NFSDCLIENT);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
NFSLOCKSTATE();
nfsd: Wrap nfsstatsv1_p in the NFSD_VNET() macro Commit 7344856e3a6d added a lot of macros that will front end vnet macros so that nfsd(8) can run in vnet prison. The nfsstatsv1_p variable got missed. This patch wraps all uses of nfsstatsv1_p with the NFSD_VNET() macro. The NFSD_VNET() macro is still a null macro. MFC after: 3 months
2023-02-16 01:39:07 +00:00
NFSD_VNET(nfsstatsv1_p)->srvclients--;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_openpluslock--;
nfsrv_clients--;
NFSUNLOCKSTATE();
}
/*
* Free a list of delegation state structures.
* (This function will also free all nfslockfile structures that no
* longer have associated state.)
*/
nfs: Remove APPLESTATIC macro It is no longer useful. Reviewed by: rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24811
2020-05-12 13:23:25 +00:00
void
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_freedeleglist(struct nfsstatehead *sthp)
{
struct nfsstate *stp, *nstp;
LIST_FOREACH_SAFE(stp, sthp, ls_list, nstp) {
nfsrv_freedeleg(stp);
}
LIST_INIT(sthp);
}
/*
* Free up a delegation.
*/
static void
nfsrv_freedeleg(struct nfsstate *stp)
{
struct nfslockfile *lfp;
LIST_REMOVE(stp, ls_hash);
LIST_REMOVE(stp, ls_list);
LIST_REMOVE(stp, ls_file);
Avoid the overhead of acquiring a lock in nfsrv_checkgetattr() when there are no write delegations issued. manu@ reported on the freebsd-current@ mailing list that there was a significant performance hit in nfsrv_checkgetattr() caused by the acquisition/release of a state lock, even when there were no write delegations issued. This patch add a count of outstanding issued write delegations to the NFSv4 server. This count allows nfsrv_checkgetattr() to return without acquiring any lock when the count is 0, avoiding the performance hit for the case where no write delegations are issued. Reported by: manu Reviewed by: kib MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D13327
2017-12-04 21:50:27 +00:00
if ((stp->ls_flags & NFSLCK_DELEGWRITE) != 0)
nfsrv_writedelegcnt--;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
lfp = stp->ls_lfp;
if (LIST_EMPTY(&lfp->lf_open) &&
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
LIST_EMPTY(&lfp->lf_lock) && LIST_EMPTY(&lfp->lf_deleg) &&
LIST_EMPTY(&lfp->lf_locallock) && LIST_EMPTY(&lfp->lf_rollback) &&
lfp->lf_usecount == 0 &&
nfsv4_testlock(&lfp->lf_locallock_lck) == 0)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_freenfslockfile(lfp);
style: Remove remaining deprecated MALLOC/FREE macros Mechanically replace uses of MALLOC/FREE with appropriate invocations of malloc(9) / free(9) (a series of sed expressions). Something like: * MALLOC(a, b, ... -> a = malloc(... * FREE( -> free( * free((caddr_t) -> free( No functional change. For now, punt on modifying contrib ipfilter code, leaving a definition of the macro in its KMALLOC(). Reported by: jhb Reviewed by: cy, imp, markj, rmacklem Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D14035
2018-01-25 22:25:13 +00:00
free(stp, M_NFSDSTATE);
nfsd: Wrap nfsstatsv1_p in the NFSD_VNET() macro Commit 7344856e3a6d added a lot of macros that will front end vnet macros so that nfsd(8) can run in vnet prison. The nfsstatsv1_p variable got missed. This patch wraps all uses of nfsstatsv1_p with the NFSD_VNET() macro. The NFSD_VNET() macro is still a null macro. MFC after: 3 months
2023-02-16 01:39:07 +00:00
NFSD_VNET(nfsstatsv1_p)->srvdelegates--;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_openpluslock--;
nfsrv_delegatecnt--;
}
/*
* This function frees an open owner and all associated opens.
*/
static void
nfsrv_freeopenowner(struct nfsstate *stp, int cansleep, NFSPROC_T *p)
{
struct nfsstate *nstp, *tstp;
LIST_REMOVE(stp, ls_list);
/*
* Now, free all associated opens.
*/
nstp = LIST_FIRST(&stp->ls_open);
while (nstp != LIST_END(&stp->ls_open)) {
tstp = nstp;
nstp = LIST_NEXT(nstp, ls_list);
(void) nfsrv_freeopen(tstp, NULL, cansleep, p);
}
if (stp->ls_op)
nfsrvd_derefcache(stp->ls_op);
style: Remove remaining deprecated MALLOC/FREE macros Mechanically replace uses of MALLOC/FREE with appropriate invocations of malloc(9) / free(9) (a series of sed expressions). Something like: * MALLOC(a, b, ... -> a = malloc(... * FREE( -> free( * free((caddr_t) -> free( No functional change. For now, punt on modifying contrib ipfilter code, leaving a definition of the macro in its KMALLOC(). Reported by: jhb Reviewed by: cy, imp, markj, rmacklem Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D14035
2018-01-25 22:25:13 +00:00
free(stp, M_NFSDSTATE);
nfsd: Wrap nfsstatsv1_p in the NFSD_VNET() macro Commit 7344856e3a6d added a lot of macros that will front end vnet macros so that nfsd(8) can run in vnet prison. The nfsstatsv1_p variable got missed. This patch wraps all uses of nfsstatsv1_p with the NFSD_VNET() macro. The NFSD_VNET() macro is still a null macro. MFC after: 3 months
2023-02-16 01:39:07 +00:00
NFSD_VNET(nfsstatsv1_p)->srvopenowners--;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_openpluslock--;
}
/*
* This function frees an open (nfsstate open structure) with all associated
* lock_owners and locks. It also frees the nfslockfile structure iff there
* are no other opens on the file.
* Returns 1 if it free'd the nfslockfile, 0 otherwise.
*/
static int
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
nfsrv_freeopen(struct nfsstate *stp, vnode_t vp, int cansleep, NFSPROC_T *p)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
{
struct nfsstate *nstp, *tstp;
struct nfslockfile *lfp;
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
int ret;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
LIST_REMOVE(stp, ls_hash);
LIST_REMOVE(stp, ls_list);
LIST_REMOVE(stp, ls_file);
lfp = stp->ls_lfp;
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
/*
* Now, free all lockowners associated with this open.
*/
LIST_FOREACH_SAFE(tstp, &stp->ls_open, ls_list, nstp)
nfsrv_freelockowner(tstp, vp, cansleep, p);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* The nfslockfile is freed here if there are no locks
* associated with the open.
* If there are locks associated with the open, the
* nfslockfile structure can be freed via nfsrv_freelockowner().
This patch applies one of the two fixes suggested by zack.kirsch at isilon.com for a race between nfsrv_freeopen() and nfsrv_getlockfile() in the experimental NFS server that he found during testing. Although nfsrv_freeopen() holds a sleep lock on the lock file structure when called with cansleep != 0, nfsrv_getlockfile() could still search the list, once it acquired the NFSLOCKSTATE() mutex. I believe that acquiring the mutex in nfsrv_freeopen() fixes the race. MFC after: 2 weeks
2010-09-10 23:49:33 +00:00
* Acquire the state mutex to avoid races with calls to
* nfsrv_getlockfile().
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
*/
This patch applies one of the two fixes suggested by zack.kirsch at isilon.com for a race between nfsrv_freeopen() and nfsrv_getlockfile() in the experimental NFS server that he found during testing. Although nfsrv_freeopen() holds a sleep lock on the lock file structure when called with cansleep != 0, nfsrv_getlockfile() could still search the list, once it acquired the NFSLOCKSTATE() mutex. I believe that acquiring the mutex in nfsrv_freeopen() fixes the race. MFC after: 2 weeks
2010-09-10 23:49:33 +00:00
if (cansleep != 0)
NFSLOCKSTATE();
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
if (lfp != NULL && LIST_EMPTY(&lfp->lf_open) &&
LIST_EMPTY(&lfp->lf_deleg) && LIST_EMPTY(&lfp->lf_lock) &&
LIST_EMPTY(&lfp->lf_locallock) && LIST_EMPTY(&lfp->lf_rollback) &&
lfp->lf_usecount == 0 &&
(cansleep != 0 || nfsv4_testlock(&lfp->lf_locallock_lck) == 0)) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_freenfslockfile(lfp);
ret = 1;
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
} else
ret = 0;
This patch applies one of the two fixes suggested by zack.kirsch at isilon.com for a race between nfsrv_freeopen() and nfsrv_getlockfile() in the experimental NFS server that he found during testing. Although nfsrv_freeopen() holds a sleep lock on the lock file structure when called with cansleep != 0, nfsrv_getlockfile() could still search the list, once it acquired the NFSLOCKSTATE() mutex. I believe that acquiring the mutex in nfsrv_freeopen() fixes the race. MFC after: 2 weeks
2010-09-10 23:49:33 +00:00
if (cansleep != 0)
NFSUNLOCKSTATE();
style: Remove remaining deprecated MALLOC/FREE macros Mechanically replace uses of MALLOC/FREE with appropriate invocations of malloc(9) / free(9) (a series of sed expressions). Something like: * MALLOC(a, b, ... -> a = malloc(... * FREE( -> free( * free((caddr_t) -> free( No functional change. For now, punt on modifying contrib ipfilter code, leaving a definition of the macro in its KMALLOC(). Reported by: jhb Reviewed by: cy, imp, markj, rmacklem Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D14035
2018-01-25 22:25:13 +00:00
free(stp, M_NFSDSTATE);
nfsd: Wrap nfsstatsv1_p in the NFSD_VNET() macro Commit 7344856e3a6d added a lot of macros that will front end vnet macros so that nfsd(8) can run in vnet prison. The nfsstatsv1_p variable got missed. This patch wraps all uses of nfsstatsv1_p with the NFSD_VNET() macro. The NFSD_VNET() macro is still a null macro. MFC after: 3 months
2023-02-16 01:39:07 +00:00
NFSD_VNET(nfsstatsv1_p)->srvopens--;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_openpluslock--;
return (ret);
}
/*
* Frees a lockowner and all associated locks.
*/
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
static void
nfsrv_freelockowner(struct nfsstate *stp, vnode_t vp, int cansleep,
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
NFSPROC_T *p)
{
LIST_REMOVE(stp, ls_hash);
LIST_REMOVE(stp, ls_list);
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
nfsrv_freeallnfslocks(stp, vp, cansleep, p);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (stp->ls_op)
nfsrvd_derefcache(stp->ls_op);
style: Remove remaining deprecated MALLOC/FREE macros Mechanically replace uses of MALLOC/FREE with appropriate invocations of malloc(9) / free(9) (a series of sed expressions). Something like: * MALLOC(a, b, ... -> a = malloc(... * FREE( -> free( * free((caddr_t) -> free( No functional change. For now, punt on modifying contrib ipfilter code, leaving a definition of the macro in its KMALLOC(). Reported by: jhb Reviewed by: cy, imp, markj, rmacklem Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D14035
2018-01-25 22:25:13 +00:00
free(stp, M_NFSDSTATE);
nfsd: Wrap nfsstatsv1_p in the NFSD_VNET() macro Commit 7344856e3a6d added a lot of macros that will front end vnet macros so that nfsd(8) can run in vnet prison. The nfsstatsv1_p variable got missed. This patch wraps all uses of nfsstatsv1_p with the NFSD_VNET() macro. The NFSD_VNET() macro is still a null macro. MFC after: 3 months
2023-02-16 01:39:07 +00:00
NFSD_VNET(nfsstatsv1_p)->srvlockowners--;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_openpluslock--;
}
/*
* Free all the nfs locks on a lockowner.
*/
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
static void
nfsrv_freeallnfslocks(struct nfsstate *stp, vnode_t vp, int cansleep,
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
NFSPROC_T *p)
{
struct nfslock *lop, *nlop;
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
struct nfsrollback *rlp, *nrlp;
struct nfslockfile *lfp = NULL;
int gottvp = 0;
vnode_t tvp = NULL;
Fix nfsrv_freeallnfslocks() in the experimental NFSv4 server so that it frees local locks correctly upon close. In order for nfsrv_localunlock() to work correctly, the lock can no longer be in the lockowner's stateid list. As such, nfsrv_freenfslock() has to be called before nfsrv_localunlock(), to get rid of the lock structure on the lockowner's stateid list. This only affected operation when local locks (vfs.newnfs.enable_locallocks=1) are enabled, which is not the default at this time. MFC after: 1 week
2010-09-19 01:18:03 +00:00
uint64_t first, end;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
if (vp != NULL)
ASSERT_VOP_UNLOCKED(vp, "nfsrv_freeallnfslocks: vnode locked");
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
lop = LIST_FIRST(&stp->ls_lock);
while (lop != LIST_END(&stp->ls_lock)) {
nlop = LIST_NEXT(lop, lo_lckowner);
/*
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
* Since all locks should be for the same file, lfp should
* not change.
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
*/
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
if (lfp == NULL)
lfp = lop->lo_lfp;
else if (lfp != lop->lo_lfp)
panic("allnfslocks");
/*
* If vp is NULL and cansleep != 0, a vnode must be acquired
* from the file handle. This only occurs when called from
* nfsrv_cleanclient().
*/
if (gottvp == 0) {
if (nfsrv_dolocallocks == 0)
tvp = NULL;
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
else if (vp == NULL && cansleep != 0) {
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
tvp = nfsvno_getvp(&lfp->lf_fh);
Fix a crash in the NFSv4 server. The PR reported a crash that occurred when a file was removed while client(s) were actively doing lock operations on it. Since nfsvno_getvp() will return NULL when the file does not exist, the bug was obvious and easy to fix via this patch. It is a little surprising that this wasn't found sooner, but I guess the above case rarely occurs. Tested by: iron.udjin@gmail.com PR: 242768 Reported by: iron.udjin@gmail.com MFC after: 2 weeks
2020-01-26 17:59:05 +00:00
if (tvp != NULL)
NFSVOPUNLOCK(tvp);
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
} else
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
tvp = vp;
gottvp = 1;
}
if (tvp != NULL) {
if (cansleep == 0)
panic("allnfs2");
Fix nfsrv_freeallnfslocks() in the experimental NFSv4 server so that it frees local locks correctly upon close. In order for nfsrv_localunlock() to work correctly, the lock can no longer be in the lockowner's stateid list. As such, nfsrv_freenfslock() has to be called before nfsrv_localunlock(), to get rid of the lock structure on the lockowner's stateid list. This only affected operation when local locks (vfs.newnfs.enable_locallocks=1) are enabled, which is not the default at this time. MFC after: 1 week
2010-09-19 01:18:03 +00:00
first = lop->lo_first;
end = lop->lo_end;
nfsrv_freenfslock(lop);
nfsrv_localunlock(tvp, lfp, first, end, p);
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
LIST_FOREACH_SAFE(rlp, &lfp->lf_rollback, rlck_list,
nrlp)
free(rlp, M_NFSDROLLBACK);
LIST_INIT(&lfp->lf_rollback);
Fix nfsrv_freeallnfslocks() in the experimental NFSv4 server so that it frees local locks correctly upon close. In order for nfsrv_localunlock() to work correctly, the lock can no longer be in the lockowner's stateid list. As such, nfsrv_freenfslock() has to be called before nfsrv_localunlock(), to get rid of the lock structure on the lockowner's stateid list. This only affected operation when local locks (vfs.newnfs.enable_locallocks=1) are enabled, which is not the default at this time. MFC after: 1 week
2010-09-19 01:18:03 +00:00
} else
nfsrv_freenfslock(lop);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
lop = nlop;
}
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
if (vp == NULL && tvp != NULL)
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
vrele(tvp);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* Free an nfslock structure.
*/
static void
nfsrv_freenfslock(struct nfslock *lop)
{
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
if (lop->lo_lckfile.le_prev != NULL) {
LIST_REMOVE(lop, lo_lckfile);
nfsd: Wrap nfsstatsv1_p in the NFSD_VNET() macro Commit 7344856e3a6d added a lot of macros that will front end vnet macros so that nfsd(8) can run in vnet prison. The nfsstatsv1_p variable got missed. This patch wraps all uses of nfsstatsv1_p with the NFSD_VNET() macro. The NFSD_VNET() macro is still a null macro. MFC after: 3 months
2023-02-16 01:39:07 +00:00
NFSD_VNET(nfsstatsv1_p)->srvlocks--;
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
nfsrv_openpluslock--;
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
LIST_REMOVE(lop, lo_lckowner);
style: Remove remaining deprecated MALLOC/FREE macros Mechanically replace uses of MALLOC/FREE with appropriate invocations of malloc(9) / free(9) (a series of sed expressions). Something like: * MALLOC(a, b, ... -> a = malloc(... * FREE( -> free( * free((caddr_t) -> free( No functional change. For now, punt on modifying contrib ipfilter code, leaving a definition of the macro in its KMALLOC(). Reported by: jhb Reviewed by: cy, imp, markj, rmacklem Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D14035
2018-01-25 22:25:13 +00:00
free(lop, M_NFSDLOCK);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* This function frees an nfslockfile structure.
*/
static void
nfsrv_freenfslockfile(struct nfslockfile *lfp)
{
LIST_REMOVE(lfp, lf_hash);
style: Remove remaining deprecated MALLOC/FREE macros Mechanically replace uses of MALLOC/FREE with appropriate invocations of malloc(9) / free(9) (a series of sed expressions). Something like: * MALLOC(a, b, ... -> a = malloc(... * FREE( -> free( * free((caddr_t) -> free( No functional change. For now, punt on modifying contrib ipfilter code, leaving a definition of the macro in its KMALLOC(). Reported by: jhb Reviewed by: cy, imp, markj, rmacklem Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D14035
2018-01-25 22:25:13 +00:00
free(lfp, M_NFSDLOCKFILE);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* This function looks up an nfsstate structure via stateid.
*/
static int
nfsrv_getstate(struct nfsclient *clp, nfsv4stateid_t *stateidp, __unused u_int32_t flags,
struct nfsstate **stpp)
{
struct nfsstate *stp;
struct nfsstatehead *hp;
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
int error = 0;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
*stpp = NULL;
hp = NFSSTATEHASH(clp, *stateidp);
LIST_FOREACH(stp, hp, ls_hash) {
if (!NFSBCMP(stp->ls_stateid.other, stateidp->other,
NFSX_STATEIDOTHER))
break;
}
/*
* If no state id in list, return NFSERR_BADSTATEID.
*/
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
if (stp == LIST_END(hp)) {
error = NFSERR_BADSTATEID;
goto out;
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
*stpp = stp;
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
out:
NFSEXITCODE(error);
return (error);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* This function gets an nfsstate structure via owner string.
*/
static void
nfsrv_getowner(struct nfsstatehead *hp, struct nfsstate *new_stp,
struct nfsstate **stpp)
{
struct nfsstate *stp;
*stpp = NULL;
LIST_FOREACH(stp, hp, ls_list) {
if (new_stp->ls_ownerlen == stp->ls_ownerlen &&
!NFSBCMP(new_stp->ls_owner,stp->ls_owner,stp->ls_ownerlen)) {
*stpp = stp;
return;
}
}
}
/*
* Lock control function called to update lock status.
* Returns 0 upon success, -1 if there is no lock and the flags indicate
* that one isn't to be created and an NFSERR_xxx for other errors.
* The structures new_stp and new_lop are passed in as pointers that should
* be set to NULL if the structure is used and shouldn't be free'd.
* For the NFSLCK_TEST and NFSLCK_CHECK cases, the structures are
* never used and can safely be allocated on the stack. For all other
* cases, *new_stpp and *new_lopp should be malloc'd before the call,
* in case they are used.
*/
nfs: Remove APPLESTATIC macro It is no longer useful. Reviewed by: rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24811
2020-05-12 13:23:25 +00:00
int
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_lockctrl(vnode_t vp, struct nfsstate **new_stpp,
struct nfslock **new_lopp, struct nfslockconflict *cfp,
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
nfsquad_t clientid, nfsv4stateid_t *stateidp,
__unused struct nfsexstuff *exp,
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
struct nfsrv_descript *nd, NFSPROC_T *p)
{
struct nfslock *lop;
struct nfsstate *new_stp = *new_stpp;
struct nfslock *new_lop = *new_lopp;
struct nfsstate *tstp, *mystp, *nstp;
int specialid = 0;
struct nfslockfile *lfp;
struct nfslock *other_lop = NULL;
struct nfsstate *stp, *lckstp = NULL;
struct nfsclient *clp = NULL;
u_int32_t bits;
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
int error = 0, haslock = 0, ret, reterr;
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
int getlckret, delegation = 0, filestruct_locked, vnode_unlocked = 0;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
fhandle_t nfh;
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
uint64_t first, end;
uint32_t lock_flags;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (new_stp->ls_flags & (NFSLCK_CHECK | NFSLCK_SETATTR)) {
/*
* Note the special cases of "all 1s" or "all 0s" stateids and
* let reads with all 1s go ahead.
*/
if (new_stp->ls_stateid.seqid == 0x0 &&
new_stp->ls_stateid.other[0] == 0x0 &&
new_stp->ls_stateid.other[1] == 0x0 &&
new_stp->ls_stateid.other[2] == 0x0)
specialid = 1;
else if (new_stp->ls_stateid.seqid == 0xffffffff &&
new_stp->ls_stateid.other[0] == 0xffffffff &&
new_stp->ls_stateid.other[1] == 0xffffffff &&
new_stp->ls_stateid.other[2] == 0xffffffff)
specialid = 2;
}
/*
* Check for restart conditions (client and server).
*/
error = nfsrv_checkrestart(clientid, new_stp->ls_flags,
&new_stp->ls_stateid, specialid);
if (error)
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* Check for state resource limit exceeded.
*/
if ((new_stp->ls_flags & NFSLCK_LOCK) &&
Make the size of the hash tables used by the NFSv4 server tunable. No appreciable change in performance was observed after increasing the sizes of these tables and then testing with a single client. However, there was an email that indicated high CPU overheads for a heavily loaded NFSv4 and it is hoped that increasing the sizes of the hash tables via these tunables might help. The tables remain the same size by default. Differential Revision: https://reviews.freebsd.org/D2596 MFC after: 2 weeks
2015-05-27 22:00:05 +00:00
nfsrv_openpluslock > nfsrv_v4statelimit) {
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
error = NFSERR_RESOURCE;
goto out;
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* For the lock case, get another nfslock structure,
* just in case we need it.
* Malloc now, before we start sifting through the linked lists,
* in case we have to wait for memory.
*/
tryagain:
if (new_stp->ls_flags & NFSLCK_LOCK)
style: Remove remaining deprecated MALLOC/FREE macros Mechanically replace uses of MALLOC/FREE with appropriate invocations of malloc(9) / free(9) (a series of sed expressions). Something like: * MALLOC(a, b, ... -> a = malloc(... * FREE( -> free( * free((caddr_t) -> free( No functional change. For now, punt on modifying contrib ipfilter code, leaving a definition of the macro in its KMALLOC(). Reported by: jhb Reviewed by: cy, imp, markj, rmacklem Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D14035
2018-01-25 22:25:13 +00:00
other_lop = malloc(sizeof (struct nfslock),
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
M_NFSDLOCK, M_WAITOK);
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
filestruct_locked = 0;
reterr = 0;
lfp = NULL;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* Get the lockfile structure for CFH now, so we can do a sanity
* check against the stateid, before incrementing the seqid#, since
* we want to return NFSERR_BADSTATEID on failure and the seqid#
* shouldn't be incremented for this case.
* If nfsrv_getlockfile() returns -1, it means "not found", which
* will be handled later.
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
* If we are doing Lock/LockU and local locking is enabled, sleep
* lock the nfslockfile structure.
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
*/
getlckret = nfsrv_getlockfh(vp, new_stp->ls_flags, NULL, &nfh, p);
NFSLOCKSTATE();
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
if (getlckret == 0) {
if ((new_stp->ls_flags & (NFSLCK_LOCK | NFSLCK_UNLOCK)) != 0 &&
nfsrv_dolocallocks != 0 && nd->nd_repstat == 0) {
getlckret = nfsrv_getlockfile(new_stp->ls_flags, NULL,
&lfp, &nfh, 1);
if (getlckret == 0)
filestruct_locked = 1;
} else
getlckret = nfsrv_getlockfile(new_stp->ls_flags, NULL,
&lfp, &nfh, 0);
}
if (getlckret != 0 && getlckret != -1)
reterr = getlckret;
if (filestruct_locked != 0) {
LIST_INIT(&lfp->lf_rollback);
if ((new_stp->ls_flags & NFSLCK_LOCK)) {
/*
* For local locking, do the advisory locking now, so
* that any conflict can be detected. A failure later
* can be rolled back locally. If an error is returned,
* struct nfslockfile has been unlocked and any local
* locking rolled back.
*/
NFSUNLOCKSTATE();
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
if (vnode_unlocked == 0) {
ASSERT_VOP_ELOCKED(vp, "nfsrv_lockctrl1");
vnode_unlocked = 1;
vfs: drop the mostly unused flags argument from VOP_UNLOCK Filesystems which want to use it in limited capacity can employ the VOP_UNLOCK_FLAGS macro. Reviewed by: kib (previous version) Differential Revision: https://reviews.freebsd.org/D21427
2020-01-03 22:29:58 +00:00
NFSVOPUNLOCK(vp);
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
}
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
reterr = nfsrv_locallock(vp, lfp,
(new_lop->lo_flags & (NFSLCK_READ | NFSLCK_WRITE)),
new_lop->lo_first, new_lop->lo_end, cfp, p);
NFSLOCKSTATE();
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
}
if (specialid == 0) {
if (new_stp->ls_flags & NFSLCK_TEST) {
/*
* RFC 3530 does not list LockT as an op that renews a
NFS: spelling fixes on comments. No funcional change.
2016-04-29 16:07:25 +00:00
* lease, but the consensus seems to be that it is ok
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
* for a server to do so.
*/
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
error = nfsrv_getclient(clientid, CLOPS_RENEW, &clp, NULL,
(nfsquad_t)((u_quad_t)0), 0, nd, p);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* Since NFSERR_EXPIRED, NFSERR_ADMINREVOKED are not valid
* error returns for LockT, just go ahead and test for a lock,
* since there are no locks for this client, but other locks
* can conflict. (ie. same client will always be false)
*/
if (error == NFSERR_EXPIRED || error == NFSERR_ADMINREVOKED)
error = 0;
lckstp = new_stp;
} else {
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
error = nfsrv_getclient(clientid, CLOPS_RENEW, &clp, NULL,
(nfsquad_t)((u_quad_t)0), 0, nd, p);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (error == 0)
/*
* Look up the stateid
*/
error = nfsrv_getstate(clp, &new_stp->ls_stateid,
new_stp->ls_flags, &stp);
/*
* do some sanity checks for an unconfirmed open or a
* stateid that refers to the wrong file, for an open stateid
*/
if (error == 0 && (stp->ls_flags & NFSLCK_OPEN) &&
((stp->ls_openowner->ls_flags & NFSLCK_NEEDSCONFIRM) ||
Fix failures and warnings reported by newpynfs20090424 test tool. This fix addresses only issues with the pynfs reports, none of these issues are know to create problems for extant real clients. Submitted by: Bart Hsiao <bart.hsiao@gmail.com> Reworked by: myself Reviewed by: rmacklem Approved by: rmacklem Sponsored by: QNAP Systems Inc.
2014-10-03 02:24:41 +00:00
(getlckret == 0 && stp->ls_lfp != lfp))){
/*
* NFSLCK_SETATTR should return OK rather than NFSERR_BADSTATEID
* The only exception is using SETATTR with SIZE.
* */
if ((new_stp->ls_flags &
(NFSLCK_SETATTR | NFSLCK_CHECK)) != NFSLCK_SETATTR)
error = NFSERR_BADSTATEID;
}
if (error == 0 &&
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
(stp->ls_flags & (NFSLCK_DELEGREAD | NFSLCK_DELEGWRITE)) &&
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
getlckret == 0 && stp->ls_lfp != lfp)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
error = NFSERR_BADSTATEID;
/*
* If the lockowner stateid doesn't refer to the same file,
* I believe that is considered ok, since some clients will
* only create a single lockowner and use that for all locks
* on all files.
* For now, log it as a diagnostic, instead of considering it
* a BadStateid.
*/
if (error == 0 && (stp->ls_flags &
(NFSLCK_OPEN | NFSLCK_DELEGREAD | NFSLCK_DELEGWRITE)) == 0 &&
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
getlckret == 0 && stp->ls_lfp != lfp) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
#ifdef DIAGNOSTIC
printf("Got a lock statid for different file open\n");
#endif
/*
error = NFSERR_BADSTATEID;
*/
}
if (error == 0) {
if (new_stp->ls_flags & NFSLCK_OPENTOLOCK) {
/*
* If haslock set, we've already checked the seqid.
*/
if (!haslock) {
if (stp->ls_flags & NFSLCK_OPEN)
error = nfsrv_checkseqid(nd, new_stp->ls_seq,
stp->ls_openowner, new_stp->ls_op);
else
error = NFSERR_BADSTATEID;
}
if (!error)
nfsrv_getowner(&stp->ls_open, new_stp, &lckstp);
Bjorn reported a problem where the Linux NFSv4.1 client is using an open_to_lock_owner4 when that lock_owner4 has already been created by a previous open_to_lock_owner4. This caused the NFS server to reply NFSERR_INVAL. For NFSv4.0, this is an error, although the updated NFSv4.0 RFC7530 notes that the correct error reply is NFSERR_BADSEQID (RFC3530 did not specify what error to return). For NFSv4.1, it is not obvious whether or not this is allowed by RFC5661, but the NFSv4.1 server can handle this case without error. This patch changes the NFSv4.1 (and NFSv4.2) server to handle multiple uses of the same lock_owner in open_to_lock_owner so that it now correctly interoperates with the Linux NFS client. It also changes the error returned for NFSv4.0 to be NFSERR_BADSEQID. Thanks go to Bjorn for diagnosing this and testing the patch. He also provided a program that I could use to reproduce the problem. Tested by: bj@cebitec.uni-bielefeld.de (Bjorn Fischer) PR: 249567 Reported by: bj@cebitec.uni-bielefeld.de (Bjorn Fischer) MFC after: 3 days
2020-09-26 23:05:38 +00:00
if (lckstp) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
Bjorn reported a problem where the Linux NFSv4.1 client is using an open_to_lock_owner4 when that lock_owner4 has already been created by a previous open_to_lock_owner4. This caused the NFS server to reply NFSERR_INVAL. For NFSv4.0, this is an error, although the updated NFSv4.0 RFC7530 notes that the correct error reply is NFSERR_BADSEQID (RFC3530 did not specify what error to return). For NFSv4.1, it is not obvious whether or not this is allowed by RFC5661, but the NFSv4.1 server can handle this case without error. This patch changes the NFSv4.1 (and NFSv4.2) server to handle multiple uses of the same lock_owner in open_to_lock_owner so that it now correctly interoperates with the Linux NFS client. It also changes the error returned for NFSv4.0 to be NFSERR_BADSEQID. Thanks go to Bjorn for diagnosing this and testing the patch. He also provided a program that I could use to reproduce the problem. Tested by: bj@cebitec.uni-bielefeld.de (Bjorn Fischer) PR: 249567 Reported by: bj@cebitec.uni-bielefeld.de (Bjorn Fischer) MFC after: 3 days
2020-09-26 23:05:38 +00:00
* For NFSv4.1 and NFSv4.2 allow an
* open_to_lock_owner when the lock_owner already
* exists. Just clear NFSLCK_OPENTOLOCK so that
* a new lock_owner will not be created.
* RFC7530 states that the error for NFSv4.0
* is NFS4ERR_BAD_SEQID.
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
*/
Bjorn reported a problem where the Linux NFSv4.1 client is using an open_to_lock_owner4 when that lock_owner4 has already been created by a previous open_to_lock_owner4. This caused the NFS server to reply NFSERR_INVAL. For NFSv4.0, this is an error, although the updated NFSv4.0 RFC7530 notes that the correct error reply is NFSERR_BADSEQID (RFC3530 did not specify what error to return). For NFSv4.1, it is not obvious whether or not this is allowed by RFC5661, but the NFSv4.1 server can handle this case without error. This patch changes the NFSv4.1 (and NFSv4.2) server to handle multiple uses of the same lock_owner in open_to_lock_owner so that it now correctly interoperates with the Linux NFS client. It also changes the error returned for NFSv4.0 to be NFSERR_BADSEQID. Thanks go to Bjorn for diagnosing this and testing the patch. He also provided a program that I could use to reproduce the problem. Tested by: bj@cebitec.uni-bielefeld.de (Bjorn Fischer) PR: 249567 Reported by: bj@cebitec.uni-bielefeld.de (Bjorn Fischer) MFC after: 3 days
2020-09-26 23:05:38 +00:00
if ((nd->nd_flag & ND_NFSV41) != 0)
new_stp->ls_flags &= ~NFSLCK_OPENTOLOCK;
else
error = NFSERR_BADSEQID;
} else
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
lckstp = new_stp;
} else if (new_stp->ls_flags&(NFSLCK_LOCK|NFSLCK_UNLOCK)) {
/*
* If haslock set, ditto above.
*/
if (!haslock) {
if (stp->ls_flags & NFSLCK_OPEN)
error = NFSERR_BADSTATEID;
else
error = nfsrv_checkseqid(nd, new_stp->ls_seq,
stp, new_stp->ls_op);
}
lckstp = stp;
} else {
lckstp = stp;
}
}
/*
* If the seqid part of the stateid isn't the same, return
* NFSERR_OLDSTATEID for cases other than I/O Ops.
* For I/O Ops, only return NFSERR_OLDSTATEID if
NFS: spelling fixes on comments. No funcional change.
2016-04-29 16:07:25 +00:00
* nfsrv_returnoldstateid is set. (The consensus on the email
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
* list was that most clients would prefer to not receive
* NFSERR_OLDSTATEID for I/O Ops, but the RFC suggests that that
* is what will happen, so I use the nfsrv_returnoldstateid to
* allow for either server configuration.)
*/
if (!error && stp->ls_stateid.seqid!=new_stp->ls_stateid.seqid &&
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
(((nd->nd_flag & ND_NFSV41) == 0 &&
(!(new_stp->ls_flags & NFSLCK_CHECK) ||
nfsrv_returnoldstateid)) ||
((nd->nd_flag & ND_NFSV41) != 0 &&
new_stp->ls_stateid.seqid != 0)))
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
error = NFSERR_OLDSTATEID;
}
}
/*
* Now we can check for grace.
*/
if (!error)
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
error = nfsrv_checkgrace(nd, clp, new_stp->ls_flags);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if ((new_stp->ls_flags & NFSLCK_RECLAIM) && !error &&
nfsrv_checkstable(clp))
error = NFSERR_NOGRACE;
/*
* If we successfully Reclaimed state, note that.
*/
if ((new_stp->ls_flags & NFSLCK_RECLAIM) && !error)
nfsrv_markstable(clp);
/*
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
* At this point, either error == NFSERR_BADSTATEID or the
* seqid# has been updated, so we can return any error.
* If error == 0, there may be an error in:
* nd_repstat - Set by the calling function.
* reterr - Set above, if getting the nfslockfile structure
* or acquiring the local lock failed.
* (If both of these are set, nd_repstat should probably be
* returned, since that error was detected before this
* function call.)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
*/
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
if (error != 0 || nd->nd_repstat != 0 || reterr != 0) {
if (error == 0) {
if (nd->nd_repstat != 0)
error = nd->nd_repstat;
else
error = reterr;
}
if (filestruct_locked != 0) {
/* Roll back local locks. */
NFSUNLOCKSTATE();
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
if (vnode_unlocked == 0) {
ASSERT_VOP_ELOCKED(vp, "nfsrv_lockctrl2");
vnode_unlocked = 1;
vfs: drop the mostly unused flags argument from VOP_UNLOCK Filesystems which want to use it in limited capacity can employ the VOP_UNLOCK_FLAGS macro. Reviewed by: kib (previous version) Differential Revision: https://reviews.freebsd.org/D21427
2020-01-03 22:29:58 +00:00
NFSVOPUNLOCK(vp);
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
}
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
nfsrv_locallock_rollback(vp, lfp, p);
NFSLOCKSTATE();
nfsrv_unlocklf(lfp);
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
NFSUNLOCKSTATE();
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* Check the nfsrv_getlockfile return.
* Returned -1 if no structure found.
*/
if (getlckret == -1) {
error = NFSERR_EXPIRED;
/*
* Called from lockt, so no lock is OK.
*/
if (new_stp->ls_flags & NFSLCK_TEST) {
error = 0;
} else if (new_stp->ls_flags &
(NFSLCK_CHECK | NFSLCK_SETATTR)) {
/*
* Called to check for a lock, OK if the stateid is all
* 1s or all 0s, but there should be an nfsstate
* otherwise.
* (ie. If there is no open, I'll assume no share
* deny bits.)
*/
if (specialid)
error = 0;
else
error = NFSERR_BADSTATEID;
}
NFSUNLOCKSTATE();
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* For NFSLCK_CHECK and NFSLCK_LOCK, test for a share conflict.
* For NFSLCK_CHECK, allow a read if write access is granted,
* but check for a deny. For NFSLCK_LOCK, require correct access,
* which implies a conflicting deny can't exist.
*/
if (new_stp->ls_flags & (NFSLCK_CHECK | NFSLCK_LOCK)) {
/*
* Four kinds of state id:
* - specialid (all 0s or all 1s), only for NFSLCK_CHECK
* - stateid for an open
* - stateid for a delegation
* - stateid for a lock owner
*/
if (!specialid) {
if (stp->ls_flags & (NFSLCK_DELEGREAD | NFSLCK_DELEGWRITE)) {
delegation = 1;
mystp = stp;
nfsrv_delaydelegtimeout(stp);
} else if (stp->ls_flags & NFSLCK_OPEN) {
mystp = stp;
} else {
mystp = stp->ls_openstp;
}
/*
* If locking or checking, require correct access
* bit set.
*/
if (((new_stp->ls_flags & NFSLCK_LOCK) &&
!((new_lop->lo_flags >> NFSLCK_LOCKSHIFT) &
mystp->ls_flags & NFSLCK_ACCESSBITS)) ||
((new_stp->ls_flags & (NFSLCK_CHECK|NFSLCK_READACCESS)) ==
(NFSLCK_CHECK | NFSLCK_READACCESS) &&
Allow use of a write open stateid for reading in the NFSv4 server. The NFSv4 RFCs give a server the option of allowing the use of an open stateid for write access to be used for a Read operation. This patch enables this by default and adds a sysctl to disable it, for anyone who does not want this capability. Allowing this is particularily useful for a pNFS Data Server (DS), since they are not permitted to allow the use of special stateids. Discovered during recent testing of the pNFS server under development. MFC after: 2 weeks
2017-04-24 20:46:19 +00:00
!(mystp->ls_flags & NFSLCK_READACCESS) &&
nfsrv_allowreadforwriteopen == 0) ||
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
((new_stp->ls_flags & (NFSLCK_CHECK|NFSLCK_WRITEACCESS)) ==
(NFSLCK_CHECK | NFSLCK_WRITEACCESS) &&
!(mystp->ls_flags & NFSLCK_WRITEACCESS))) {
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
if (filestruct_locked != 0) {
/* Roll back local locks. */
NFSUNLOCKSTATE();
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
if (vnode_unlocked == 0) {
ASSERT_VOP_ELOCKED(vp,
"nfsrv_lockctrl3");
vnode_unlocked = 1;
vfs: drop the mostly unused flags argument from VOP_UNLOCK Filesystems which want to use it in limited capacity can employ the VOP_UNLOCK_FLAGS macro. Reviewed by: kib (previous version) Differential Revision: https://reviews.freebsd.org/D21427
2020-01-03 22:29:58 +00:00
NFSVOPUNLOCK(vp);
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
}
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
nfsrv_locallock_rollback(vp, lfp, p);
NFSLOCKSTATE();
nfsrv_unlocklf(lfp);
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
NFSUNLOCKSTATE();
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
error = NFSERR_OPENMODE;
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
} else
mystp = NULL;
if ((new_stp->ls_flags & NFSLCK_CHECK) && !delegation) {
/*
* Check for a conflicting deny bit.
*/
LIST_FOREACH(tstp, &lfp->lf_open, ls_file) {
if (tstp != mystp) {
bits = tstp->ls_flags;
bits >>= NFSLCK_SHIFT;
if (new_stp->ls_flags & bits & NFSLCK_ACCESSBITS) {
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
KASSERT(vnode_unlocked == 0,
("nfsrv_lockctrl: vnode unlocked1"));
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
ret = nfsrv_clientconflict(tstp->ls_clp, &haslock,
vp, p);
Add checks for VI_DOOMED and vn_lock() failures to the experimental NFS server, to handle the case where an exported file system is forced dismounted while an RPC is in progress. Further commits will fix the cases where a mount point is used when the associated vnode isn't locked. Reviewed by: kib MFC after: 2 weeks
2011-01-02 19:58:39 +00:00
if (ret == 1) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* nfsrv_clientconflict unlocks state
* when it returns non-zero.
*/
lckstp = NULL;
goto tryagain;
}
Add checks for VI_DOOMED and vn_lock() failures to the experimental NFS server, to handle the case where an exported file system is forced dismounted while an RPC is in progress. Further commits will fix the cases where a mount point is used when the associated vnode isn't locked. Reviewed by: kib MFC after: 2 weeks
2011-01-02 19:58:39 +00:00
if (ret == 0)
NFSUNLOCKSTATE();
if (ret == 2)
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
error = NFSERR_PERM;
Add checks for VI_DOOMED and vn_lock() failures to the experimental NFS server, to handle the case where an exported file system is forced dismounted while an RPC is in progress. Further commits will fix the cases where a mount point is used when the associated vnode isn't locked. Reviewed by: kib MFC after: 2 weeks
2011-01-02 19:58:39 +00:00
else
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
error = NFSERR_OPENMODE;
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
}
}
/* We're outta here */
NFSUNLOCKSTATE();
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
}
/*
* For setattr, just get rid of all the Delegations for other clients.
*/
if (new_stp->ls_flags & NFSLCK_SETATTR) {
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
KASSERT(vnode_unlocked == 0,
("nfsrv_lockctrl: vnode unlocked2"));
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
ret = nfsrv_cleandeleg(vp, lfp, clp, &haslock, p);
if (ret) {
/*
* nfsrv_cleandeleg() unlocks state when it
* returns non-zero.
*/
if (ret == -1) {
lckstp = NULL;
goto tryagain;
}
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
error = ret;
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
if (!(new_stp->ls_flags & NFSLCK_CHECK) ||
(LIST_EMPTY(&lfp->lf_open) && LIST_EMPTY(&lfp->lf_lock) &&
LIST_EMPTY(&lfp->lf_deleg))) {
NFSUNLOCKSTATE();
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
}
/*
* Check for a conflicting delegation. If one is found, call
* nfsrv_delegconflict() to handle it. If the v4root lock hasn't
* been set yet, it will get the lock. Otherwise, it will recall
* the delegation. Then, we try try again...
* I currently believe the conflict algorithm to be:
* For Lock Ops (Lock/LockT/LockU)
* - there is a conflict iff a different client has a write delegation
* For Reading (Read Op)
* - there is a conflict iff a different client has a write delegation
* (the specialids are always a different client)
* For Writing (Write/Setattr of size)
* - there is a conflict if a different client has any delegation
* - there is a conflict if the same client has a read delegation
* (I don't understand why this isn't allowed, but that seems to be
NFS: spelling fixes on comments. No funcional change.
2016-04-29 16:07:25 +00:00
* the current consensus?)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
*/
tstp = LIST_FIRST(&lfp->lf_deleg);
while (tstp != LIST_END(&lfp->lf_deleg)) {
nstp = LIST_NEXT(tstp, ls_file);
if ((((new_stp->ls_flags&(NFSLCK_LOCK|NFSLCK_UNLOCK|NFSLCK_TEST))||
((new_stp->ls_flags & NFSLCK_CHECK) &&
(new_lop->lo_flags & NFSLCK_READ))) &&
clp != tstp->ls_clp &&
(tstp->ls_flags & NFSLCK_DELEGWRITE)) ||
((new_stp->ls_flags & NFSLCK_CHECK) &&
(new_lop->lo_flags & NFSLCK_WRITE) &&
(clp != tstp->ls_clp ||
(tstp->ls_flags & NFSLCK_DELEGREAD)))) {
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
ret = 0;
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
if (filestruct_locked != 0) {
/* Roll back local locks. */
NFSUNLOCKSTATE();
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
if (vnode_unlocked == 0) {
ASSERT_VOP_ELOCKED(vp, "nfsrv_lockctrl4");
vfs: drop the mostly unused flags argument from VOP_UNLOCK Filesystems which want to use it in limited capacity can employ the VOP_UNLOCK_FLAGS macro. Reviewed by: kib (previous version) Differential Revision: https://reviews.freebsd.org/D21427
2020-01-03 22:29:58 +00:00
NFSVOPUNLOCK(vp);
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
}
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
nfsrv_locallock_rollback(vp, lfp, p);
NFSLOCKSTATE();
nfsrv_unlocklf(lfp);
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
NFSUNLOCKSTATE();
NFSVOPLOCK(vp, LK_EXCLUSIVE | LK_RETRY);
vnode_unlocked = 0;
vfs: introduce v_irflag and make v_type smaller The current vnode layout is not smp-friendly by having frequently read data avoidably sharing cachelines with very frequently modified fields. In particular v_iflag inspected for VI_DOOMED can be found in the same line with v_usecount. Instead make it available in the same cacheline as the v_op, v_data and v_type which all get read all the time. v_type is avoidably 4 bytes while the necessary data will easily fit in 1. Shrinking it frees up 3 bytes, 2 of which get used here to introduce a new flag field with a new value: VIRF_DOOMED. Reviewed by: kib, jeff Differential Revision: https://reviews.freebsd.org/D22715
2019-12-08 21:30:04 +00:00
if (VN_IS_DOOMED(vp))
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
ret = NFSERR_SERVERFAULT;
NFSLOCKSTATE();
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
}
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
if (ret == 0)
ret = nfsrv_delegconflict(tstp, &haslock, p, vp);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (ret) {
/*
* nfsrv_delegconflict unlocks state when it
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
* returns non-zero, which it always does.
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
*/
if (other_lop) {
style: Remove remaining deprecated MALLOC/FREE macros Mechanically replace uses of MALLOC/FREE with appropriate invocations of malloc(9) / free(9) (a series of sed expressions). Something like: * MALLOC(a, b, ... -> a = malloc(... * FREE( -> free( * free((caddr_t) -> free( No functional change. For now, punt on modifying contrib ipfilter code, leaving a definition of the macro in its KMALLOC(). Reported by: jhb Reviewed by: cy, imp, markj, rmacklem Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D14035
2018-01-25 22:25:13 +00:00
free(other_lop, M_NFSDLOCK);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
other_lop = NULL;
}
if (ret == -1) {
lckstp = NULL;
goto tryagain;
}
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
error = ret;
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
/* Never gets here. */
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
tstp = nstp;
}
/*
* Handle the unlock case by calling nfsrv_updatelock().
* (Should I have done some access checking above for unlock? For now,
* just let it happen.)
*/
if (new_stp->ls_flags & NFSLCK_UNLOCK) {
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
first = new_lop->lo_first;
end = new_lop->lo_end;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_updatelock(stp, new_lopp, &other_lop, lfp);
stateidp->seqid = ++(stp->ls_stateid.seqid);
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
if ((nd->nd_flag & ND_NFSV41) != 0 && stateidp->seqid == 0)
stateidp->seqid = stp->ls_stateid.seqid = 1;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
stateidp->other[0] = stp->ls_stateid.other[0];
stateidp->other[1] = stp->ls_stateid.other[1];
stateidp->other[2] = stp->ls_stateid.other[2];
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
if (filestruct_locked != 0) {
NFSUNLOCKSTATE();
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
if (vnode_unlocked == 0) {
ASSERT_VOP_ELOCKED(vp, "nfsrv_lockctrl5");
vnode_unlocked = 1;
vfs: drop the mostly unused flags argument from VOP_UNLOCK Filesystems which want to use it in limited capacity can employ the VOP_UNLOCK_FLAGS macro. Reviewed by: kib (previous version) Differential Revision: https://reviews.freebsd.org/D21427
2020-01-03 22:29:58 +00:00
NFSVOPUNLOCK(vp);
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
}
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
/* Update the local locks. */
nfsrv_localunlock(vp, lfp, first, end, p);
NFSLOCKSTATE();
nfsrv_unlocklf(lfp);
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
NFSUNLOCKSTATE();
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* Search for a conflicting lock. A lock conflicts if:
* - the lock range overlaps and
* - at least one lock is a write lock and
* - it is not owned by the same lock owner
*/
if (!delegation) {
LIST_FOREACH(lop, &lfp->lf_lock, lo_lckfile) {
if (new_lop->lo_end > lop->lo_first &&
new_lop->lo_first < lop->lo_end &&
(new_lop->lo_flags == NFSLCK_WRITE ||
lop->lo_flags == NFSLCK_WRITE) &&
lckstp != lop->lo_stp &&
The test for "same client" for the experimental nfs server over NFSv4 was broken w.r.t. byte range lock conflicts when it was the same client and the request used the open_to_lock_owner4 case, since lckstp->ls_clp was not set. This patch fixes it by using "clp" instead of "lckstp->ls_clp". MFC after: 2 weeks
2010-01-03 20:08:10 +00:00
(clp != lop->lo_stp->ls_clp ||
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
lckstp->ls_ownerlen != lop->lo_stp->ls_ownerlen ||
NFSBCMP(lckstp->ls_owner, lop->lo_stp->ls_owner,
lckstp->ls_ownerlen))) {
if (other_lop) {
style: Remove remaining deprecated MALLOC/FREE macros Mechanically replace uses of MALLOC/FREE with appropriate invocations of malloc(9) / free(9) (a series of sed expressions). Something like: * MALLOC(a, b, ... -> a = malloc(... * FREE( -> free( * free((caddr_t) -> free( No functional change. For now, punt on modifying contrib ipfilter code, leaving a definition of the macro in its KMALLOC(). Reported by: jhb Reviewed by: cy, imp, markj, rmacklem Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D14035
2018-01-25 22:25:13 +00:00
free(other_lop, M_NFSDLOCK);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
other_lop = NULL;
}
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
if (vnode_unlocked != 0)
ret = nfsrv_clientconflict(lop->lo_stp->ls_clp, &haslock,
NULL, p);
else
ret = nfsrv_clientconflict(lop->lo_stp->ls_clp, &haslock,
vp, p);
Add checks for VI_DOOMED and vn_lock() failures to the experimental NFS server, to handle the case where an exported file system is forced dismounted while an RPC is in progress. Further commits will fix the cases where a mount point is used when the associated vnode isn't locked. Reviewed by: kib MFC after: 2 weeks
2011-01-02 19:58:39 +00:00
if (ret == 1) {
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
if (filestruct_locked != 0) {
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
if (vnode_unlocked == 0) {
ASSERT_VOP_ELOCKED(vp, "nfsrv_lockctrl6");
vfs: drop the mostly unused flags argument from VOP_UNLOCK Filesystems which want to use it in limited capacity can employ the VOP_UNLOCK_FLAGS macro. Reviewed by: kib (previous version) Differential Revision: https://reviews.freebsd.org/D21427
2020-01-03 22:29:58 +00:00
NFSVOPUNLOCK(vp);
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
}
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
/* Roll back local locks. */
nfsrv_locallock_rollback(vp, lfp, p);
NFSLOCKSTATE();
nfsrv_unlocklf(lfp);
NFSUNLOCKSTATE();
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
NFSVOPLOCK(vp, LK_EXCLUSIVE | LK_RETRY);
vnode_unlocked = 0;
vfs: introduce v_irflag and make v_type smaller The current vnode layout is not smp-friendly by having frequently read data avoidably sharing cachelines with very frequently modified fields. In particular v_iflag inspected for VI_DOOMED can be found in the same line with v_usecount. Instead make it available in the same cacheline as the v_op, v_data and v_type which all get read all the time. v_type is avoidably 4 bytes while the necessary data will easily fit in 1. Shrinking it frees up 3 bytes, 2 of which get used here to introduce a new flag field with a new value: VIRF_DOOMED. Reviewed by: kib, jeff Differential Revision: https://reviews.freebsd.org/D22715
2019-12-08 21:30:04 +00:00
if (VN_IS_DOOMED(vp)) {
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
error = NFSERR_SERVERFAULT;
goto out;
}
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* nfsrv_clientconflict() unlocks state when it
* returns non-zero.
*/
lckstp = NULL;
goto tryagain;
}
/*
* Found a conflicting lock, so record the conflict and
* return the error.
*/
Add checks for VI_DOOMED and vn_lock() failures to the experimental NFS server, to handle the case where an exported file system is forced dismounted while an RPC is in progress. Further commits will fix the cases where a mount point is used when the associated vnode isn't locked. Reviewed by: kib MFC after: 2 weeks
2011-01-02 19:58:39 +00:00
if (cfp != NULL && ret == 0) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
cfp->cl_clientid.lval[0]=lop->lo_stp->ls_stateid.other[0];
cfp->cl_clientid.lval[1]=lop->lo_stp->ls_stateid.other[1];
cfp->cl_first = lop->lo_first;
cfp->cl_end = lop->lo_end;
cfp->cl_flags = lop->lo_flags;
cfp->cl_ownerlen = lop->lo_stp->ls_ownerlen;
NFSBCOPY(lop->lo_stp->ls_owner, cfp->cl_owner,
cfp->cl_ownerlen);
}
Add checks for VI_DOOMED and vn_lock() failures to the experimental NFS server, to handle the case where an exported file system is forced dismounted while an RPC is in progress. Further commits will fix the cases where a mount point is used when the associated vnode isn't locked. Reviewed by: kib MFC after: 2 weeks
2011-01-02 19:58:39 +00:00
if (ret == 2)
error = NFSERR_PERM;
else if (new_stp->ls_flags & NFSLCK_RECLAIM)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
error = NFSERR_RECLAIMCONFLICT;
else if (new_stp->ls_flags & NFSLCK_CHECK)
error = NFSERR_LOCKED;
else
error = NFSERR_DENIED;
Add checks for VI_DOOMED and vn_lock() failures to the experimental NFS server, to handle the case where an exported file system is forced dismounted while an RPC is in progress. Further commits will fix the cases where a mount point is used when the associated vnode isn't locked. Reviewed by: kib MFC after: 2 weeks
2011-01-02 19:58:39 +00:00
if (filestruct_locked != 0 && ret == 0) {
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
/* Roll back local locks. */
NFSUNLOCKSTATE();
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
if (vnode_unlocked == 0) {
ASSERT_VOP_ELOCKED(vp, "nfsrv_lockctrl7");
vnode_unlocked = 1;
vfs: drop the mostly unused flags argument from VOP_UNLOCK Filesystems which want to use it in limited capacity can employ the VOP_UNLOCK_FLAGS macro. Reviewed by: kib (previous version) Differential Revision: https://reviews.freebsd.org/D21427
2020-01-03 22:29:58 +00:00
NFSVOPUNLOCK(vp);
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
}
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
nfsrv_locallock_rollback(vp, lfp, p);
NFSLOCKSTATE();
nfsrv_unlocklf(lfp);
}
Add checks for VI_DOOMED and vn_lock() failures to the experimental NFS server, to handle the case where an exported file system is forced dismounted while an RPC is in progress. Further commits will fix the cases where a mount point is used when the associated vnode isn't locked. Reviewed by: kib MFC after: 2 weeks
2011-01-02 19:58:39 +00:00
if (ret == 0)
NFSUNLOCKSTATE();
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
}
}
/*
* We only get here if there was no lock that conflicted.
*/
if (new_stp->ls_flags & (NFSLCK_TEST | NFSLCK_CHECK)) {
NFSUNLOCKSTATE();
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* We only get here when we are creating or modifying a lock.
* There are two variants:
* - exist_lock_owner where lock_owner exists
* - open_to_lock_owner with new lock_owner
*/
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
first = new_lop->lo_first;
end = new_lop->lo_end;
lock_flags = new_lop->lo_flags;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (!(new_stp->ls_flags & NFSLCK_OPENTOLOCK)) {
nfsrv_updatelock(lckstp, new_lopp, &other_lop, lfp);
stateidp->seqid = ++(lckstp->ls_stateid.seqid);
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
if ((nd->nd_flag & ND_NFSV41) != 0 && stateidp->seqid == 0)
stateidp->seqid = lckstp->ls_stateid.seqid = 1;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
stateidp->other[0] = lckstp->ls_stateid.other[0];
stateidp->other[1] = lckstp->ls_stateid.other[1];
stateidp->other[2] = lckstp->ls_stateid.other[2];
} else {
/*
* The new open_to_lock_owner case.
* Link the new nfsstate into the lists.
*/
new_stp->ls_seq = new_stp->ls_opentolockseq;
nfsrvd_refcache(new_stp->ls_op);
In the experimental NFS server, when converting an open-owner to a lock-owner, start at sequence id 1 instead of 0, to match up with both Solaris and Linux. Reviewed by: rmacklem Approved by: zml (mentor)
2011-01-12 23:46:12 +00:00
stateidp->seqid = new_stp->ls_stateid.seqid = 1;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
stateidp->other[0] = new_stp->ls_stateid.other[0] =
clp->lc_clientid.lval[0];
stateidp->other[1] = new_stp->ls_stateid.other[1] =
clp->lc_clientid.lval[1];
stateidp->other[2] = new_stp->ls_stateid.other[2] =
nfsrv_nextstateindex(clp);
new_stp->ls_clp = clp;
LIST_INIT(&new_stp->ls_lock);
new_stp->ls_openstp = stp;
new_stp->ls_lfp = lfp;
nfsrv_insertlock(new_lop, (struct nfslock *)new_stp, new_stp,
lfp);
LIST_INSERT_HEAD(NFSSTATEHASH(clp, new_stp->ls_stateid),
new_stp, ls_hash);
LIST_INSERT_HEAD(&stp->ls_open, new_stp, ls_list);
*new_lopp = NULL;
*new_stpp = NULL;
nfsd: Wrap nfsstatsv1_p in the NFSD_VNET() macro Commit 7344856e3a6d added a lot of macros that will front end vnet macros so that nfsd(8) can run in vnet prison. The nfsstatsv1_p variable got missed. This patch wraps all uses of nfsstatsv1_p with the NFSD_VNET() macro. The NFSD_VNET() macro is still a null macro. MFC after: 3 months
2023-02-16 01:39:07 +00:00
NFSD_VNET(nfsstatsv1_p)->srvlockowners++;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_openpluslock++;
}
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
if (filestruct_locked != 0) {
NFSUNLOCKSTATE();
nfsrv_locallock_commit(lfp, lock_flags, first, end);
NFSLOCKSTATE();
nfsrv_unlocklf(lfp);
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
NFSUNLOCKSTATE();
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
out:
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (haslock) {
NFSLOCKV4ROOTMUTEX();
nfsv4_unlock(&nfsv4rootfs_lock, 1);
NFSUNLOCKV4ROOTMUTEX();
}
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
if (vnode_unlocked != 0) {
NFSVOPLOCK(vp, LK_EXCLUSIVE | LK_RETRY);
vfs: introduce v_irflag and make v_type smaller The current vnode layout is not smp-friendly by having frequently read data avoidably sharing cachelines with very frequently modified fields. In particular v_iflag inspected for VI_DOOMED can be found in the same line with v_usecount. Instead make it available in the same cacheline as the v_op, v_data and v_type which all get read all the time. v_type is avoidably 4 bytes while the necessary data will easily fit in 1. Shrinking it frees up 3 bytes, 2 of which get used here to introduce a new flag field with a new value: VIRF_DOOMED. Reviewed by: kib, jeff Differential Revision: https://reviews.freebsd.org/D22715
2019-12-08 21:30:04 +00:00
if (error == 0 && VN_IS_DOOMED(vp))
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
error = NFSERR_SERVERFAULT;
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (other_lop)
style: Remove remaining deprecated MALLOC/FREE macros Mechanically replace uses of MALLOC/FREE with appropriate invocations of malloc(9) / free(9) (a series of sed expressions). Something like: * MALLOC(a, b, ... -> a = malloc(... * FREE( -> free( * free((caddr_t) -> free( No functional change. For now, punt on modifying contrib ipfilter code, leaving a definition of the macro in its KMALLOC(). Reported by: jhb Reviewed by: cy, imp, markj, rmacklem Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D14035
2018-01-25 22:25:13 +00:00
free(other_lop, M_NFSDLOCK);
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
NFSEXITCODE2(error, nd);
return (error);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* Check for state errors for Open.
* repstat is passed back out as an error if more critical errors
* are not detected.
*/
nfs: Remove APPLESTATIC macro It is no longer useful. Reviewed by: rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24811
2020-05-12 13:23:25 +00:00
int
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_opencheck(nfsquad_t clientid, nfsv4stateid_t *stateidp,
struct nfsstate *new_stp, vnode_t vp, struct nfsrv_descript *nd,
NFSPROC_T *p, int repstat)
{
struct nfsstate *stp, *nstp;
struct nfsclient *clp;
struct nfsstate *ownerstp;
struct nfslockfile *lfp, *new_lfp;
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
int error = 0, haslock = 0, ret, readonly = 0, getfhret = 0;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if ((new_stp->ls_flags & NFSLCK_SHAREBITS) == NFSLCK_READACCESS)
readonly = 1;
/*
* Check for restart conditions (client and server).
*/
error = nfsrv_checkrestart(clientid, new_stp->ls_flags,
&new_stp->ls_stateid, 0);
if (error)
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* Check for state resource limit exceeded.
* Technically this should be SMP protected, but the worst
* case error is "out by one or two" on the count when it
* returns NFSERR_RESOURCE and the limit is just a rather
* arbitrary high water mark, so no harm is done.
*/
Make the size of the hash tables used by the NFSv4 server tunable. No appreciable change in performance was observed after increasing the sizes of these tables and then testing with a single client. However, there was an email that indicated high CPU overheads for a heavily loaded NFSv4 and it is hoped that increasing the sizes of the hash tables via these tunables might help. The tables remain the same size by default. Differential Revision: https://reviews.freebsd.org/D2596 MFC after: 2 weeks
2015-05-27 22:00:05 +00:00
if (nfsrv_openpluslock > nfsrv_v4statelimit) {
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
error = NFSERR_RESOURCE;
goto out;
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
tryagain:
style: Remove remaining deprecated MALLOC/FREE macros Mechanically replace uses of MALLOC/FREE with appropriate invocations of malloc(9) / free(9) (a series of sed expressions). Something like: * MALLOC(a, b, ... -> a = malloc(... * FREE( -> free( * free((caddr_t) -> free( No functional change. For now, punt on modifying contrib ipfilter code, leaving a definition of the macro in its KMALLOC(). Reported by: jhb Reviewed by: cy, imp, markj, rmacklem Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D14035
2018-01-25 22:25:13 +00:00
new_lfp = malloc(sizeof (struct nfslockfile),
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
M_NFSDLOCKFILE, M_WAITOK);
if (vp)
Remove an unnecessary level of indirection for an argument. This simplifies the code and should avoid the clang sparc port from generating an abort() call. Requested by: rdivacky Submitted by: jhb MFC after: 2 weeks
2014-04-23 23:13:46 +00:00
getfhret = nfsrv_getlockfh(vp, new_stp->ls_flags, new_lfp,
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
NULL, p);
NFSLOCKSTATE();
/*
* Get the nfsclient structure.
*/
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
error = nfsrv_getclient(clientid, CLOPS_RENEW, &clp, NULL,
(nfsquad_t)((u_quad_t)0), 0, nd, p);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* Look up the open owner. See if it needs confirmation and
* check the seq#, as required.
*/
if (!error)
nfsrv_getowner(&clp->lc_open, new_stp, &ownerstp);
if (!error && ownerstp) {
error = nfsrv_checkseqid(nd, new_stp->ls_seq, ownerstp,
new_stp->ls_op);
/*
* If the OpenOwner hasn't been confirmed, assume the
* old one was a replay and this one is ok.
* See: RFC3530 Sec. 14.2.18.
*/
if (error == NFSERR_BADSEQID &&
(ownerstp->ls_flags & NFSLCK_NEEDSCONFIRM))
error = 0;
}
/*
* Check for grace.
*/
if (!error)
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
error = nfsrv_checkgrace(nd, clp, new_stp->ls_flags);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if ((new_stp->ls_flags & NFSLCK_RECLAIM) && !error &&
nfsrv_checkstable(clp))
error = NFSERR_NOGRACE;
/*
* If none of the above errors occurred, let repstat be
* returned.
*/
if (repstat && !error)
error = repstat;
if (error) {
NFSUNLOCKSTATE();
if (haslock) {
NFSLOCKV4ROOTMUTEX();
nfsv4_unlock(&nfsv4rootfs_lock, 1);
NFSUNLOCKV4ROOTMUTEX();
}
style: Remove remaining deprecated MALLOC/FREE macros Mechanically replace uses of MALLOC/FREE with appropriate invocations of malloc(9) / free(9) (a series of sed expressions). Something like: * MALLOC(a, b, ... -> a = malloc(... * FREE( -> free( * free((caddr_t) -> free( No functional change. For now, punt on modifying contrib ipfilter code, leaving a definition of the macro in its KMALLOC(). Reported by: jhb Reviewed by: cy, imp, markj, rmacklem Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D14035
2018-01-25 22:25:13 +00:00
free(new_lfp, M_NFSDLOCKFILE);
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* If vp == NULL, the file doesn't exist yet, so return ok.
* (This always happens on the first pass, so haslock must be 0.)
*/
if (vp == NULL) {
NFSUNLOCKSTATE();
style: Remove remaining deprecated MALLOC/FREE macros Mechanically replace uses of MALLOC/FREE with appropriate invocations of malloc(9) / free(9) (a series of sed expressions). Something like: * MALLOC(a, b, ... -> a = malloc(... * FREE( -> free( * free((caddr_t) -> free( No functional change. For now, punt on modifying contrib ipfilter code, leaving a definition of the macro in its KMALLOC(). Reported by: jhb Reviewed by: cy, imp, markj, rmacklem Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D14035
2018-01-25 22:25:13 +00:00
free(new_lfp, M_NFSDLOCKFILE);
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* Get the structure for the underlying file.
*/
if (getfhret)
error = getfhret;
else
error = nfsrv_getlockfile(new_stp->ls_flags, &new_lfp, &lfp,
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
NULL, 0);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (new_lfp)
style: Remove remaining deprecated MALLOC/FREE macros Mechanically replace uses of MALLOC/FREE with appropriate invocations of malloc(9) / free(9) (a series of sed expressions). Something like: * MALLOC(a, b, ... -> a = malloc(... * FREE( -> free( * free((caddr_t) -> free( No functional change. For now, punt on modifying contrib ipfilter code, leaving a definition of the macro in its KMALLOC(). Reported by: jhb Reviewed by: cy, imp, markj, rmacklem Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D14035
2018-01-25 22:25:13 +00:00
free(new_lfp, M_NFSDLOCKFILE);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (error) {
NFSUNLOCKSTATE();
if (haslock) {
NFSLOCKV4ROOTMUTEX();
nfsv4_unlock(&nfsv4rootfs_lock, 1);
NFSUNLOCKV4ROOTMUTEX();
}
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* Search for a conflicting open/share.
*/
if (new_stp->ls_flags & NFSLCK_DELEGCUR) {
/*
* For Delegate_Cur, search for the matching Delegation,
* which indicates no conflict.
* An old delegation should have been recovered by the
* client doing a Claim_DELEGATE_Prev, so I won't let
* it match and return NFSERR_EXPIRED. Should I let it
* match?
*/
LIST_FOREACH(stp, &lfp->lf_deleg, ls_file) {
if (!(stp->ls_flags & NFSLCK_OLDDELEG) &&
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
(((nd->nd_flag & ND_NFSV41) != 0 &&
stateidp->seqid == 0) ||
stateidp->seqid == stp->ls_stateid.seqid) &&
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
!NFSBCMP(stateidp->other, stp->ls_stateid.other,
NFSX_STATEIDOTHER))
break;
}
if (stp == LIST_END(&lfp->lf_deleg) ||
((new_stp->ls_flags & NFSLCK_WRITEACCESS) &&
(stp->ls_flags & NFSLCK_DELEGREAD))) {
NFSUNLOCKSTATE();
if (haslock) {
NFSLOCKV4ROOTMUTEX();
nfsv4_unlock(&nfsv4rootfs_lock, 1);
NFSUNLOCKV4ROOTMUTEX();
}
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
error = NFSERR_EXPIRED;
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
}
/*
* Check for access/deny bit conflicts. I check for the same
* owner as well, in case the client didn't bother.
*/
LIST_FOREACH(stp, &lfp->lf_open, ls_file) {
if (!(new_stp->ls_flags & NFSLCK_DELEGCUR) &&
(((new_stp->ls_flags & NFSLCK_ACCESSBITS) &
((stp->ls_flags>>NFSLCK_SHIFT) & NFSLCK_ACCESSBITS))||
((stp->ls_flags & NFSLCK_ACCESSBITS) &
((new_stp->ls_flags>>NFSLCK_SHIFT)&NFSLCK_ACCESSBITS)))){
ret = nfsrv_clientconflict(stp->ls_clp,&haslock,vp,p);
Add checks for VI_DOOMED and vn_lock() failures to the experimental NFS server, to handle the case where an exported file system is forced dismounted while an RPC is in progress. Further commits will fix the cases where a mount point is used when the associated vnode isn't locked. Reviewed by: kib MFC after: 2 weeks
2011-01-02 19:58:39 +00:00
if (ret == 1) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* nfsrv_clientconflict() unlocks
* state when it returns non-zero.
*/
goto tryagain;
}
Add checks for VI_DOOMED and vn_lock() failures to the experimental NFS server, to handle the case where an exported file system is forced dismounted while an RPC is in progress. Further commits will fix the cases where a mount point is used when the associated vnode isn't locked. Reviewed by: kib MFC after: 2 weeks
2011-01-02 19:58:39 +00:00
if (ret == 2)
error = NFSERR_PERM;
else if (new_stp->ls_flags & NFSLCK_RECLAIM)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
error = NFSERR_RECLAIMCONFLICT;
else
error = NFSERR_SHAREDENIED;
Add checks for VI_DOOMED and vn_lock() failures to the experimental NFS server, to handle the case where an exported file system is forced dismounted while an RPC is in progress. Further commits will fix the cases where a mount point is used when the associated vnode isn't locked. Reviewed by: kib MFC after: 2 weeks
2011-01-02 19:58:39 +00:00
if (ret == 0)
NFSUNLOCKSTATE();
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (haslock) {
NFSLOCKV4ROOTMUTEX();
nfsv4_unlock(&nfsv4rootfs_lock, 1);
NFSUNLOCKV4ROOTMUTEX();
}
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
}
/*
* Check for a conflicting delegation. If one is found, call
* nfsrv_delegconflict() to handle it. If the v4root lock hasn't
* been set yet, it will get the lock. Otherwise, it will recall
* the delegation. Then, we try try again...
* (If NFSLCK_DELEGCUR is set, it has a delegation, so there
* isn't a conflict.)
* I currently believe the conflict algorithm to be:
* For Open with Read Access and Deny None
* - there is a conflict iff a different client has a write delegation
* For Open with other Write Access or any Deny except None
* - there is a conflict if a different client has any delegation
* - there is a conflict if the same client has a read delegation
NFS: spelling fixes on comments. No funcional change.
2016-04-29 16:07:25 +00:00
* (The current consensus is that this last case should be
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
* considered a conflict since the client with a read delegation
* could have done an Open with ReadAccess and WriteDeny
* locally and then not have checked for the WriteDeny.)
* Don't check for a Reclaim, since that will be dealt with
* by nfsrv_openctrl().
*/
if (!(new_stp->ls_flags &
(NFSLCK_DELEGPREV | NFSLCK_DELEGCUR | NFSLCK_RECLAIM))) {
stp = LIST_FIRST(&lfp->lf_deleg);
while (stp != LIST_END(&lfp->lf_deleg)) {
nstp = LIST_NEXT(stp, ls_file);
if ((readonly && stp->ls_clp != clp &&
(stp->ls_flags & NFSLCK_DELEGWRITE)) ||
(!readonly && (stp->ls_clp != clp ||
(stp->ls_flags & NFSLCK_DELEGREAD)))) {
ret = nfsrv_delegconflict(stp, &haslock, p, vp);
if (ret) {
/*
* nfsrv_delegconflict() unlocks state
* when it returns non-zero.
*/
if (ret == -1)
goto tryagain;
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
error = ret;
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
}
stp = nstp;
}
}
NFSUNLOCKSTATE();
if (haslock) {
NFSLOCKV4ROOTMUTEX();
nfsv4_unlock(&nfsv4rootfs_lock, 1);
NFSUNLOCKV4ROOTMUTEX();
}
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
out:
NFSEXITCODE2(error, nd);
return (error);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* Open control function to create/update open state for an open.
*/
nfs: Remove APPLESTATIC macro It is no longer useful. Reviewed by: rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24811
2020-05-12 13:23:25 +00:00
int
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_openctrl(struct nfsrv_descript *nd, vnode_t vp,
struct nfsstate **new_stpp, nfsquad_t clientid, nfsv4stateid_t *stateidp,
nfsv4stateid_t *delegstateidp, u_int32_t *rflagsp, struct nfsexstuff *exp,
NFSPROC_T *p, u_quad_t filerev)
{
struct nfsstate *new_stp = *new_stpp;
struct nfsstate *stp, *nstp;
struct nfsstate *openstp = NULL, *new_open, *ownerstp, *new_deleg;
struct nfslockfile *lfp, *new_lfp;
struct nfsclient *clp;
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
int error = 0, haslock = 0, ret, delegate = 1, writedeleg = 1;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
int readonly = 0, cbret = 1, getfhret = 0;
Fix the NFSv4.1 server for Open reclaim after a reboot. The NFSv4.1 server failed to update the nfs-stablerestart file for a client when the client was issued its first Open. As such, recovery of Opens after a server reboot failed with NFSERR_NOGRACE. This patch fixes this. It also changes the code so that it malloc()'s the 1024 byte array instead of allocating it on the kernel stack for both NFSv4.0 and NFSv4.1. Note that this bug only affected NFSv4.1 and only when clients attempted to reclaim Opens after a server reboot. MFC after: 2 weeks
2016-12-05 22:36:25 +00:00
int gotstate = 0, len = 0;
u_char *clidp = NULL;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if ((new_stp->ls_flags & NFSLCK_SHAREBITS) == NFSLCK_READACCESS)
readonly = 1;
/*
* Check for restart conditions (client and server).
* (Paranoia, should have been detected by nfsrv_opencheck().)
* If an error does show up, return NFSERR_EXPIRED, since the
* the seqid# has already been incremented.
*/
error = nfsrv_checkrestart(clientid, new_stp->ls_flags,
&new_stp->ls_stateid, 0);
if (error) {
printf("Nfsd: openctrl unexpected restart err=%d\n",
error);
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
error = NFSERR_EXPIRED;
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
Fix the NFSv4.1 server for Open reclaim after a reboot. The NFSv4.1 server failed to update the nfs-stablerestart file for a client when the client was issued its first Open. As such, recovery of Opens after a server reboot failed with NFSERR_NOGRACE. This patch fixes this. It also changes the code so that it malloc()'s the 1024 byte array instead of allocating it on the kernel stack for both NFSv4.0 and NFSv4.1. Note that this bug only affected NFSv4.1 and only when clients attempted to reclaim Opens after a server reboot. MFC after: 2 weeks
2016-12-05 22:36:25 +00:00
clidp = malloc(NFSV4_OPAQUELIMIT, M_TEMP, M_WAITOK);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
tryagain:
style: Remove remaining deprecated MALLOC/FREE macros Mechanically replace uses of MALLOC/FREE with appropriate invocations of malloc(9) / free(9) (a series of sed expressions). Something like: * MALLOC(a, b, ... -> a = malloc(... * FREE( -> free( * free((caddr_t) -> free( No functional change. For now, punt on modifying contrib ipfilter code, leaving a definition of the macro in its KMALLOC(). Reported by: jhb Reviewed by: cy, imp, markj, rmacklem Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D14035
2018-01-25 22:25:13 +00:00
new_lfp = malloc(sizeof (struct nfslockfile),
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
M_NFSDLOCKFILE, M_WAITOK);
style: Remove remaining deprecated MALLOC/FREE macros Mechanically replace uses of MALLOC/FREE with appropriate invocations of malloc(9) / free(9) (a series of sed expressions). Something like: * MALLOC(a, b, ... -> a = malloc(... * FREE( -> free( * free((caddr_t) -> free( No functional change. For now, punt on modifying contrib ipfilter code, leaving a definition of the macro in its KMALLOC(). Reported by: jhb Reviewed by: cy, imp, markj, rmacklem Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D14035
2018-01-25 22:25:13 +00:00
new_open = malloc(sizeof (struct nfsstate),
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
M_NFSDSTATE, M_WAITOK);
style: Remove remaining deprecated MALLOC/FREE macros Mechanically replace uses of MALLOC/FREE with appropriate invocations of malloc(9) / free(9) (a series of sed expressions). Something like: * MALLOC(a, b, ... -> a = malloc(... * FREE( -> free( * free((caddr_t) -> free( No functional change. For now, punt on modifying contrib ipfilter code, leaving a definition of the macro in its KMALLOC(). Reported by: jhb Reviewed by: cy, imp, markj, rmacklem Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D14035
2018-01-25 22:25:13 +00:00
new_deleg = malloc(sizeof (struct nfsstate),
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
M_NFSDSTATE, M_WAITOK);
Remove an unnecessary level of indirection for an argument. This simplifies the code and should avoid the clang sparc port from generating an abort() call. Requested by: rdivacky Submitted by: jhb MFC after: 2 weeks
2014-04-23 23:13:46 +00:00
getfhret = nfsrv_getlockfh(vp, new_stp->ls_flags, new_lfp,
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
NULL, p);
NFSLOCKSTATE();
/*
* Get the client structure. Since the linked lists could be changed
* by other nfsd processes if this process does a tsleep(), one of
* two things must be done.
* 1 - don't tsleep()
* or
* 2 - get the nfsv4_lock() { indicated by haslock == 1 }
* before using the lists, since this lock stops the other
* nfsd. This should only be used for rare cases, since it
* essentially single threads the nfsd.
* At this time, it is only done for cases where the stable
* storage file must be written prior to completion of state
* expiration.
*/
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
error = nfsrv_getclient(clientid, CLOPS_RENEW, &clp, NULL,
(nfsquad_t)((u_quad_t)0), 0, nd, p);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (!error && (clp->lc_flags & LCL_NEEDSCBNULL) &&
clp->lc_program) {
/*
* This happens on the first open for a client
* that supports callbacks.
*/
NFSUNLOCKSTATE();
/*
* Although nfsrv_docallback() will sleep, clp won't
* go away, since they are only removed when the
* nfsv4_lock() has blocked the nfsd threads. The
* fields in clp can change, but having multiple
* threads do this Null callback RPC should be
* harmless.
*/
cbret = nfsrv_docallback(clp, NFSV4PROC_CBNULL,
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
NULL, 0, NULL, NULL, NULL, 0, p);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
NFSLOCKSTATE();
clp->lc_flags &= ~LCL_NEEDSCBNULL;
if (!cbret)
clp->lc_flags |= LCL_CALLBACKSON;
}
/*
* Look up the open owner. See if it needs confirmation and
* check the seq#, as required.
*/
if (!error)
nfsrv_getowner(&clp->lc_open, new_stp, &ownerstp);
if (error) {
NFSUNLOCKSTATE();
printf("Nfsd: openctrl unexpected state err=%d\n",
error);
style: Remove remaining deprecated MALLOC/FREE macros Mechanically replace uses of MALLOC/FREE with appropriate invocations of malloc(9) / free(9) (a series of sed expressions). Something like: * MALLOC(a, b, ... -> a = malloc(... * FREE( -> free( * free((caddr_t) -> free( No functional change. For now, punt on modifying contrib ipfilter code, leaving a definition of the macro in its KMALLOC(). Reported by: jhb Reviewed by: cy, imp, markj, rmacklem Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D14035
2018-01-25 22:25:13 +00:00
free(new_lfp, M_NFSDLOCKFILE);
free(new_open, M_NFSDSTATE);
free(new_deleg, M_NFSDSTATE);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (haslock) {
NFSLOCKV4ROOTMUTEX();
nfsv4_unlock(&nfsv4rootfs_lock, 1);
NFSUNLOCKV4ROOTMUTEX();
}
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
error = NFSERR_EXPIRED;
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
if (new_stp->ls_flags & NFSLCK_RECLAIM)
nfsrv_markstable(clp);
/*
* Get the structure for the underlying file.
*/
if (getfhret)
error = getfhret;
else
error = nfsrv_getlockfile(new_stp->ls_flags, &new_lfp, &lfp,
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
NULL, 0);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (new_lfp)
style: Remove remaining deprecated MALLOC/FREE macros Mechanically replace uses of MALLOC/FREE with appropriate invocations of malloc(9) / free(9) (a series of sed expressions). Something like: * MALLOC(a, b, ... -> a = malloc(... * FREE( -> free( * free((caddr_t) -> free( No functional change. For now, punt on modifying contrib ipfilter code, leaving a definition of the macro in its KMALLOC(). Reported by: jhb Reviewed by: cy, imp, markj, rmacklem Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D14035
2018-01-25 22:25:13 +00:00
free(new_lfp, M_NFSDLOCKFILE);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (error) {
NFSUNLOCKSTATE();
printf("Nfsd openctrl unexpected getlockfile err=%d\n",
error);
style: Remove remaining deprecated MALLOC/FREE macros Mechanically replace uses of MALLOC/FREE with appropriate invocations of malloc(9) / free(9) (a series of sed expressions). Something like: * MALLOC(a, b, ... -> a = malloc(... * FREE( -> free( * free((caddr_t) -> free( No functional change. For now, punt on modifying contrib ipfilter code, leaving a definition of the macro in its KMALLOC(). Reported by: jhb Reviewed by: cy, imp, markj, rmacklem Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D14035
2018-01-25 22:25:13 +00:00
free(new_open, M_NFSDSTATE);
free(new_deleg, M_NFSDSTATE);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (haslock) {
NFSLOCKV4ROOTMUTEX();
nfsv4_unlock(&nfsv4rootfs_lock, 1);
NFSUNLOCKV4ROOTMUTEX();
}
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* Search for a conflicting open/share.
*/
if (new_stp->ls_flags & NFSLCK_DELEGCUR) {
/*
* For Delegate_Cur, search for the matching Delegation,
* which indicates no conflict.
* An old delegation should have been recovered by the
* client doing a Claim_DELEGATE_Prev, so I won't let
* it match and return NFSERR_EXPIRED. Should I let it
* match?
*/
LIST_FOREACH(stp, &lfp->lf_deleg, ls_file) {
if (!(stp->ls_flags & NFSLCK_OLDDELEG) &&
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
(((nd->nd_flag & ND_NFSV41) != 0 &&
stateidp->seqid == 0) ||
stateidp->seqid == stp->ls_stateid.seqid) &&
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
!NFSBCMP(stateidp->other, stp->ls_stateid.other,
NFSX_STATEIDOTHER))
break;
}
if (stp == LIST_END(&lfp->lf_deleg) ||
((new_stp->ls_flags & NFSLCK_WRITEACCESS) &&
(stp->ls_flags & NFSLCK_DELEGREAD))) {
NFSUNLOCKSTATE();
printf("Nfsd openctrl unexpected expiry\n");
style: Remove remaining deprecated MALLOC/FREE macros Mechanically replace uses of MALLOC/FREE with appropriate invocations of malloc(9) / free(9) (a series of sed expressions). Something like: * MALLOC(a, b, ... -> a = malloc(... * FREE( -> free( * free((caddr_t) -> free( No functional change. For now, punt on modifying contrib ipfilter code, leaving a definition of the macro in its KMALLOC(). Reported by: jhb Reviewed by: cy, imp, markj, rmacklem Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D14035
2018-01-25 22:25:13 +00:00
free(new_open, M_NFSDSTATE);
free(new_deleg, M_NFSDSTATE);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (haslock) {
NFSLOCKV4ROOTMUTEX();
nfsv4_unlock(&nfsv4rootfs_lock, 1);
NFSUNLOCKV4ROOTMUTEX();
}
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
error = NFSERR_EXPIRED;
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* Don't issue a Delegation, since one already exists and
* delay delegation timeout, as required.
*/
delegate = 0;
nfsrv_delaydelegtimeout(stp);
}
/*
* Check for access/deny bit conflicts. I also check for the
* same owner, since the client might not have bothered to check.
* Also, note an open for the same file and owner, if found,
* which is all we do here for Delegate_Cur, since conflict
* checking is already done.
*/
LIST_FOREACH(stp, &lfp->lf_open, ls_file) {
if (ownerstp && stp->ls_openowner == ownerstp)
openstp = stp;
if (!(new_stp->ls_flags & NFSLCK_DELEGCUR)) {
/*
* If another client has the file open, the only
* delegation that can be issued is a Read delegation
* and only if it is a Read open with Deny none.
*/
if (clp != stp->ls_clp) {
if ((stp->ls_flags & NFSLCK_SHAREBITS) ==
NFSLCK_READACCESS)
writedeleg = 0;
else
delegate = 0;
}
if(((new_stp->ls_flags & NFSLCK_ACCESSBITS) &
((stp->ls_flags>>NFSLCK_SHIFT) & NFSLCK_ACCESSBITS))||
((stp->ls_flags & NFSLCK_ACCESSBITS) &
((new_stp->ls_flags>>NFSLCK_SHIFT)&NFSLCK_ACCESSBITS))){
ret = nfsrv_clientconflict(stp->ls_clp,&haslock,vp,p);
Add checks for VI_DOOMED and vn_lock() failures to the experimental NFS server, to handle the case where an exported file system is forced dismounted while an RPC is in progress. Further commits will fix the cases where a mount point is used when the associated vnode isn't locked. Reviewed by: kib MFC after: 2 weeks
2011-01-02 19:58:39 +00:00
if (ret == 1) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* nfsrv_clientconflict() unlocks state
* when it returns non-zero.
*/
style: Remove remaining deprecated MALLOC/FREE macros Mechanically replace uses of MALLOC/FREE with appropriate invocations of malloc(9) / free(9) (a series of sed expressions). Something like: * MALLOC(a, b, ... -> a = malloc(... * FREE( -> free( * free((caddr_t) -> free( No functional change. For now, punt on modifying contrib ipfilter code, leaving a definition of the macro in its KMALLOC(). Reported by: jhb Reviewed by: cy, imp, markj, rmacklem Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D14035
2018-01-25 22:25:13 +00:00
free(new_open, M_NFSDSTATE);
free(new_deleg, M_NFSDSTATE);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
openstp = NULL;
goto tryagain;
}
Add checks for VI_DOOMED and vn_lock() failures to the experimental NFS server, to handle the case where an exported file system is forced dismounted while an RPC is in progress. Further commits will fix the cases where a mount point is used when the associated vnode isn't locked. Reviewed by: kib MFC after: 2 weeks
2011-01-02 19:58:39 +00:00
if (ret == 2)
error = NFSERR_PERM;
else if (new_stp->ls_flags & NFSLCK_RECLAIM)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
error = NFSERR_RECLAIMCONFLICT;
else
error = NFSERR_SHAREDENIED;
Add checks for VI_DOOMED and vn_lock() failures to the experimental NFS server, to handle the case where an exported file system is forced dismounted while an RPC is in progress. Further commits will fix the cases where a mount point is used when the associated vnode isn't locked. Reviewed by: kib MFC after: 2 weeks
2011-01-02 19:58:39 +00:00
if (ret == 0)
NFSUNLOCKSTATE();
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (haslock) {
NFSLOCKV4ROOTMUTEX();
nfsv4_unlock(&nfsv4rootfs_lock, 1);
NFSUNLOCKV4ROOTMUTEX();
}
style: Remove remaining deprecated MALLOC/FREE macros Mechanically replace uses of MALLOC/FREE with appropriate invocations of malloc(9) / free(9) (a series of sed expressions). Something like: * MALLOC(a, b, ... -> a = malloc(... * FREE( -> free( * free((caddr_t) -> free( No functional change. For now, punt on modifying contrib ipfilter code, leaving a definition of the macro in its KMALLOC(). Reported by: jhb Reviewed by: cy, imp, markj, rmacklem Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D14035
2018-01-25 22:25:13 +00:00
free(new_open, M_NFSDSTATE);
free(new_deleg, M_NFSDSTATE);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
printf("nfsd openctrl unexpected client cnfl\n");
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
}
}
/*
* Check for a conflicting delegation. If one is found, call
* nfsrv_delegconflict() to handle it. If the v4root lock hasn't
* been set yet, it will get the lock. Otherwise, it will recall
* the delegation. Then, we try try again...
* (If NFSLCK_DELEGCUR is set, it has a delegation, so there
* isn't a conflict.)
* I currently believe the conflict algorithm to be:
* For Open with Read Access and Deny None
* - there is a conflict iff a different client has a write delegation
* For Open with other Write Access or any Deny except None
* - there is a conflict if a different client has any delegation
* - there is a conflict if the same client has a read delegation
NFS: spelling fixes on comments. No funcional change.
2016-04-29 16:07:25 +00:00
* (The current consensus is that this last case should be
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
* considered a conflict since the client with a read delegation
* could have done an Open with ReadAccess and WriteDeny
* locally and then not have checked for the WriteDeny.)
*/
if (!(new_stp->ls_flags & (NFSLCK_DELEGPREV | NFSLCK_DELEGCUR))) {
stp = LIST_FIRST(&lfp->lf_deleg);
while (stp != LIST_END(&lfp->lf_deleg)) {
nstp = LIST_NEXT(stp, ls_file);
if (stp->ls_clp != clp && (stp->ls_flags & NFSLCK_DELEGREAD))
writedeleg = 0;
else
delegate = 0;
if ((readonly && stp->ls_clp != clp &&
(stp->ls_flags & NFSLCK_DELEGWRITE)) ||
(!readonly && (stp->ls_clp != clp ||
(stp->ls_flags & NFSLCK_DELEGREAD)))) {
if (new_stp->ls_flags & NFSLCK_RECLAIM) {
delegate = 2;
} else {
ret = nfsrv_delegconflict(stp, &haslock, p, vp);
if (ret) {
/*
* nfsrv_delegconflict() unlocks state
* when it returns non-zero.
*/
printf("Nfsd openctrl unexpected deleg cnfl\n");
style: Remove remaining deprecated MALLOC/FREE macros Mechanically replace uses of MALLOC/FREE with appropriate invocations of malloc(9) / free(9) (a series of sed expressions). Something like: * MALLOC(a, b, ... -> a = malloc(... * FREE( -> free( * free((caddr_t) -> free( No functional change. For now, punt on modifying contrib ipfilter code, leaving a definition of the macro in its KMALLOC(). Reported by: jhb Reviewed by: cy, imp, markj, rmacklem Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D14035
2018-01-25 22:25:13 +00:00
free(new_open, M_NFSDSTATE);
free(new_deleg, M_NFSDSTATE);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (ret == -1) {
openstp = NULL;
goto tryagain;
}
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
error = ret;
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
}
}
stp = nstp;
}
}
/*
* We only get here if there was no open that conflicted.
* If an open for the owner exists, or in the access/deny bits.
* Otherwise it is a new open. If the open_owner hasn't been
* confirmed, replace the open with the new one needing confirmation,
* otherwise add the open.
*/
if (new_stp->ls_flags & NFSLCK_DELEGPREV) {
/*
* Handle NFSLCK_DELEGPREV by searching the old delegations for
* a match. If found, just move the old delegation to the current
* delegation list and issue open. If not found, return
* NFSERR_EXPIRED.
*/
LIST_FOREACH(stp, &clp->lc_olddeleg, ls_list) {
if (stp->ls_lfp == lfp) {
/* Found it */
if (stp->ls_clp != clp)
panic("olddeleg clp");
LIST_REMOVE(stp, ls_list);
LIST_REMOVE(stp, ls_hash);
stp->ls_flags &= ~NFSLCK_OLDDELEG;
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
stp->ls_stateid.seqid = delegstateidp->seqid = 1;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
stp->ls_stateid.other[0] = delegstateidp->other[0] =
clp->lc_clientid.lval[0];
stp->ls_stateid.other[1] = delegstateidp->other[1] =
clp->lc_clientid.lval[1];
stp->ls_stateid.other[2] = delegstateidp->other[2] =
nfsrv_nextstateindex(clp);
stp->ls_compref = nd->nd_compref;
LIST_INSERT_HEAD(&clp->lc_deleg, stp, ls_list);
LIST_INSERT_HEAD(NFSSTATEHASH(clp,
stp->ls_stateid), stp, ls_hash);
if (stp->ls_flags & NFSLCK_DELEGWRITE)
*rflagsp |= NFSV4OPEN_WRITEDELEGATE;
else
*rflagsp |= NFSV4OPEN_READDELEGATE;
clp->lc_delegtime = NFSD_MONOSEC +
nfsrv_lease + NFSRV_LEASEDELTA;
/*
* Now, do the associated open.
*/
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
new_open->ls_stateid.seqid = 1;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
new_open->ls_stateid.other[0] = clp->lc_clientid.lval[0];
new_open->ls_stateid.other[1] = clp->lc_clientid.lval[1];
new_open->ls_stateid.other[2] = nfsrv_nextstateindex(clp);
new_open->ls_flags = (new_stp->ls_flags&NFSLCK_DENYBITS)|
NFSLCK_OPEN;
if (stp->ls_flags & NFSLCK_DELEGWRITE)
new_open->ls_flags |= (NFSLCK_READACCESS |
NFSLCK_WRITEACCESS);
else
new_open->ls_flags |= NFSLCK_READACCESS;
new_open->ls_uid = new_stp->ls_uid;
new_open->ls_lfp = lfp;
new_open->ls_clp = clp;
LIST_INIT(&new_open->ls_open);
LIST_INSERT_HEAD(&lfp->lf_open, new_open, ls_file);
LIST_INSERT_HEAD(NFSSTATEHASH(clp, new_open->ls_stateid),
new_open, ls_hash);
/*
* and handle the open owner
*/
if (ownerstp) {
new_open->ls_openowner = ownerstp;
LIST_INSERT_HEAD(&ownerstp->ls_open,new_open,ls_list);
} else {
new_open->ls_openowner = new_stp;
new_stp->ls_flags = 0;
nfsrvd_refcache(new_stp->ls_op);
new_stp->ls_noopens = 0;
LIST_INIT(&new_stp->ls_open);
LIST_INSERT_HEAD(&new_stp->ls_open, new_open, ls_list);
LIST_INSERT_HEAD(&clp->lc_open, new_stp, ls_list);
*new_stpp = NULL;
nfsd: Wrap nfsstatsv1_p in the NFSD_VNET() macro Commit 7344856e3a6d added a lot of macros that will front end vnet macros so that nfsd(8) can run in vnet prison. The nfsstatsv1_p variable got missed. This patch wraps all uses of nfsstatsv1_p with the NFSD_VNET() macro. The NFSD_VNET() macro is still a null macro. MFC after: 3 months
2023-02-16 01:39:07 +00:00
NFSD_VNET(nfsstatsv1_p)->srvopenowners++;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_openpluslock++;
}
openstp = new_open;
new_open = NULL;
nfsd: Wrap nfsstatsv1_p in the NFSD_VNET() macro Commit 7344856e3a6d added a lot of macros that will front end vnet macros so that nfsd(8) can run in vnet prison. The nfsstatsv1_p variable got missed. This patch wraps all uses of nfsstatsv1_p with the NFSD_VNET() macro. The NFSD_VNET() macro is still a null macro. MFC after: 3 months
2023-02-16 01:39:07 +00:00
NFSD_VNET(nfsstatsv1_p)->srvopens++;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_openpluslock++;
break;
}
}
if (stp == LIST_END(&clp->lc_olddeleg))
error = NFSERR_EXPIRED;
} else if (new_stp->ls_flags & (NFSLCK_DELEGREAD | NFSLCK_DELEGWRITE)) {
/*
* Scan to see that no delegation for this client and file
* doesn't already exist.
* There also shouldn't yet be an Open for this file and
* openowner.
*/
LIST_FOREACH(stp, &lfp->lf_deleg, ls_file) {
if (stp->ls_clp == clp)
break;
}
if (stp == LIST_END(&lfp->lf_deleg) && openstp == NULL) {
/*
* This is the Claim_Previous case with a delegation
* type != Delegate_None.
*/
/*
* First, add the delegation. (Although we must issue the
* delegation, we can also ask for an immediate return.)
*/
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
new_deleg->ls_stateid.seqid = delegstateidp->seqid = 1;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
new_deleg->ls_stateid.other[0] = delegstateidp->other[0] =
clp->lc_clientid.lval[0];
new_deleg->ls_stateid.other[1] = delegstateidp->other[1] =
clp->lc_clientid.lval[1];
new_deleg->ls_stateid.other[2] = delegstateidp->other[2] =
nfsrv_nextstateindex(clp);
if (new_stp->ls_flags & NFSLCK_DELEGWRITE) {
new_deleg->ls_flags = (NFSLCK_DELEGWRITE |
NFSLCK_READACCESS | NFSLCK_WRITEACCESS);
*rflagsp |= NFSV4OPEN_WRITEDELEGATE;
Avoid the overhead of acquiring a lock in nfsrv_checkgetattr() when there are no write delegations issued. manu@ reported on the freebsd-current@ mailing list that there was a significant performance hit in nfsrv_checkgetattr() caused by the acquisition/release of a state lock, even when there were no write delegations issued. This patch add a count of outstanding issued write delegations to the NFSv4 server. This count allows nfsrv_checkgetattr() to return without acquiring any lock when the count is 0, avoiding the performance hit for the case where no write delegations are issued. Reported by: manu Reviewed by: kib MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D13327
2017-12-04 21:50:27 +00:00
nfsrv_writedelegcnt++;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
} else {
new_deleg->ls_flags = (NFSLCK_DELEGREAD |
NFSLCK_READACCESS);
*rflagsp |= NFSV4OPEN_READDELEGATE;
}
new_deleg->ls_uid = new_stp->ls_uid;
new_deleg->ls_lfp = lfp;
new_deleg->ls_clp = clp;
new_deleg->ls_filerev = filerev;
new_deleg->ls_compref = nd->nd_compref;
nfsd: make the server repeat CB_RECALL every couple of seconds Commit 01ae8969a9ee stopped the NFSv4.1/4.2 server from implicitly binding the back channel to a new TCP connection so that it conforms to RFC5661, for NFSv4.1/4.2. An effect of this for the Linux NFS client is that it will do a BindConnectionToSession when it sees NFSV4SEQ_CBPATHDOWN set in a sequence reply. This will fix the back channel, but the first attempt at a callback like CB_RECALL will already have failed. Without this patch, a CB_RECALL will not be retried and that can result in a 5 minute delay until the delegation times out. This patch modifies the code so that it will retry the CB_RECALL every couple of seconds, often avoiding the 5 minute delay. This is not critical for correct behaviour, but avoids the 5 minute delay for the case where the Linux client re-binds the back channel via BindConnectionToSession. MFC after: 2 weeks
2021-04-05 01:15:54 +00:00
new_deleg->ls_lastrecall = 0;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
LIST_INSERT_HEAD(&lfp->lf_deleg, new_deleg, ls_file);
LIST_INSERT_HEAD(NFSSTATEHASH(clp,
new_deleg->ls_stateid), new_deleg, ls_hash);
LIST_INSERT_HEAD(&clp->lc_deleg, new_deleg, ls_list);
new_deleg = NULL;
if (delegate == 2 || nfsrv_issuedelegs == 0 ||
(clp->lc_flags & (LCL_CALLBACKSON | LCL_CBDOWN)) !=
LCL_CALLBACKSON ||
NFSRV_V4DELEGLIMIT(nfsrv_delegatecnt) ||
!NFSVNO_DELEGOK(vp))
*rflagsp |= NFSV4OPEN_RECALL;
nfsd: Wrap nfsstatsv1_p in the NFSD_VNET() macro Commit 7344856e3a6d added a lot of macros that will front end vnet macros so that nfsd(8) can run in vnet prison. The nfsstatsv1_p variable got missed. This patch wraps all uses of nfsstatsv1_p with the NFSD_VNET() macro. The NFSD_VNET() macro is still a null macro. MFC after: 3 months
2023-02-16 01:39:07 +00:00
NFSD_VNET(nfsstatsv1_p)->srvdelegates++;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_openpluslock++;
nfsrv_delegatecnt++;
/*
* Now, do the associated open.
*/
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
new_open->ls_stateid.seqid = 1;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
new_open->ls_stateid.other[0] = clp->lc_clientid.lval[0];
new_open->ls_stateid.other[1] = clp->lc_clientid.lval[1];
new_open->ls_stateid.other[2] = nfsrv_nextstateindex(clp);
new_open->ls_flags = (new_stp->ls_flags & NFSLCK_DENYBITS) |
NFSLCK_OPEN;
if (new_stp->ls_flags & NFSLCK_DELEGWRITE)
new_open->ls_flags |= (NFSLCK_READACCESS |
NFSLCK_WRITEACCESS);
else
new_open->ls_flags |= NFSLCK_READACCESS;
new_open->ls_uid = new_stp->ls_uid;
new_open->ls_lfp = lfp;
new_open->ls_clp = clp;
LIST_INIT(&new_open->ls_open);
LIST_INSERT_HEAD(&lfp->lf_open, new_open, ls_file);
LIST_INSERT_HEAD(NFSSTATEHASH(clp, new_open->ls_stateid),
new_open, ls_hash);
/*
* and handle the open owner
*/
if (ownerstp) {
new_open->ls_openowner = ownerstp;
LIST_INSERT_HEAD(&ownerstp->ls_open, new_open, ls_list);
} else {
new_open->ls_openowner = new_stp;
new_stp->ls_flags = 0;
nfsrvd_refcache(new_stp->ls_op);
new_stp->ls_noopens = 0;
LIST_INIT(&new_stp->ls_open);
LIST_INSERT_HEAD(&new_stp->ls_open, new_open, ls_list);
LIST_INSERT_HEAD(&clp->lc_open, new_stp, ls_list);
*new_stpp = NULL;
nfsd: Wrap nfsstatsv1_p in the NFSD_VNET() macro Commit 7344856e3a6d added a lot of macros that will front end vnet macros so that nfsd(8) can run in vnet prison. The nfsstatsv1_p variable got missed. This patch wraps all uses of nfsstatsv1_p with the NFSD_VNET() macro. The NFSD_VNET() macro is still a null macro. MFC after: 3 months
2023-02-16 01:39:07 +00:00
NFSD_VNET(nfsstatsv1_p)->srvopenowners++;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_openpluslock++;
}
openstp = new_open;
new_open = NULL;
nfsd: Wrap nfsstatsv1_p in the NFSD_VNET() macro Commit 7344856e3a6d added a lot of macros that will front end vnet macros so that nfsd(8) can run in vnet prison. The nfsstatsv1_p variable got missed. This patch wraps all uses of nfsstatsv1_p with the NFSD_VNET() macro. The NFSD_VNET() macro is still a null macro. MFC after: 3 months
2023-02-16 01:39:07 +00:00
NFSD_VNET(nfsstatsv1_p)->srvopens++;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_openpluslock++;
} else {
error = NFSERR_RECLAIMCONFLICT;
}
} else if (ownerstp) {
if (ownerstp->ls_flags & NFSLCK_NEEDSCONFIRM) {
/* Replace the open */
if (ownerstp->ls_op)
nfsrvd_derefcache(ownerstp->ls_op);
ownerstp->ls_op = new_stp->ls_op;
nfsrvd_refcache(ownerstp->ls_op);
ownerstp->ls_seq = new_stp->ls_seq;
*rflagsp |= NFSV4OPEN_RESULTCONFIRM;
stp = LIST_FIRST(&ownerstp->ls_open);
stp->ls_flags = (new_stp->ls_flags & NFSLCK_SHAREBITS) |
NFSLCK_OPEN;
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
stp->ls_stateid.seqid = 1;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
stp->ls_uid = new_stp->ls_uid;
if (lfp != stp->ls_lfp) {
LIST_REMOVE(stp, ls_file);
LIST_INSERT_HEAD(&lfp->lf_open, stp, ls_file);
stp->ls_lfp = lfp;
}
openstp = stp;
} else if (openstp) {
openstp->ls_flags |= (new_stp->ls_flags & NFSLCK_SHAREBITS);
openstp->ls_stateid.seqid++;
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
if ((nd->nd_flag & ND_NFSV41) != 0 &&
openstp->ls_stateid.seqid == 0)
openstp->ls_stateid.seqid = 1;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* This is where we can choose to issue a delegation.
*/
Modify the reasons for not issuing a delegation in the NFSv4.1 server. The ESXi NFSv4.1 client will generate warning messages when the reason for not issuing a delegation is two. Two refers to a resource limit and I do not see why it would be considered invalid. However it probably was not the best choice of reason for not issuing a delegation. This patch changes the reasons used to ones that the ESXi client doesn't complain about. This change does not affect the FreeBSD client and does not appear to affect behaviour of the Linux NFSv4.1 client. RFC5661 defines these "reasons" but does not give any guidance w.r.t. which ones are more appropriate to return to a client. Tested by: andreas.nagy@frequentis.com PR: 226650 MFC after: 2 weeks
2018-07-16 21:32:50 +00:00
if ((new_stp->ls_flags & NFSLCK_WANTNODELEG) != 0)
*rflagsp |= NFSV4OPEN_WDNOTWANTED;
else if (nfsrv_issuedelegs == 0)
*rflagsp |= NFSV4OPEN_WDSUPPFTYPE;
else if (NFSRV_V4DELEGLIMIT(nfsrv_delegatecnt))
*rflagsp |= NFSV4OPEN_WDRESOURCE;
else if (delegate == 0 || writedeleg == 0 ||
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
NFSVNO_EXRDONLY(exp) || (readonly != 0 &&
nfsrv_writedelegifpos == 0) ||
!NFSVNO_DELEGOK(vp) ||
(new_stp->ls_flags & NFSLCK_WANTRDELEG) != 0 ||
(clp->lc_flags & (LCL_CALLBACKSON | LCL_CBDOWN)) !=
LCL_CALLBACKSON)
*rflagsp |= NFSV4OPEN_WDCONTENTION;
else {
new_deleg->ls_stateid.seqid = delegstateidp->seqid = 1;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
new_deleg->ls_stateid.other[0] = delegstateidp->other[0]
= clp->lc_clientid.lval[0];
new_deleg->ls_stateid.other[1] = delegstateidp->other[1]
= clp->lc_clientid.lval[1];
new_deleg->ls_stateid.other[2] = delegstateidp->other[2]
= nfsrv_nextstateindex(clp);
new_deleg->ls_flags = (NFSLCK_DELEGWRITE |
NFSLCK_READACCESS | NFSLCK_WRITEACCESS);
*rflagsp |= NFSV4OPEN_WRITEDELEGATE;
new_deleg->ls_uid = new_stp->ls_uid;
new_deleg->ls_lfp = lfp;
new_deleg->ls_clp = clp;
new_deleg->ls_filerev = filerev;
new_deleg->ls_compref = nd->nd_compref;
nfsd: make the server repeat CB_RECALL every couple of seconds Commit 01ae8969a9ee stopped the NFSv4.1/4.2 server from implicitly binding the back channel to a new TCP connection so that it conforms to RFC5661, for NFSv4.1/4.2. An effect of this for the Linux NFS client is that it will do a BindConnectionToSession when it sees NFSV4SEQ_CBPATHDOWN set in a sequence reply. This will fix the back channel, but the first attempt at a callback like CB_RECALL will already have failed. Without this patch, a CB_RECALL will not be retried and that can result in a 5 minute delay until the delegation times out. This patch modifies the code so that it will retry the CB_RECALL every couple of seconds, often avoiding the 5 minute delay. This is not critical for correct behaviour, but avoids the 5 minute delay for the case where the Linux client re-binds the back channel via BindConnectionToSession. MFC after: 2 weeks
2021-04-05 01:15:54 +00:00
new_deleg->ls_lastrecall = 0;
Avoid the overhead of acquiring a lock in nfsrv_checkgetattr() when there are no write delegations issued. manu@ reported on the freebsd-current@ mailing list that there was a significant performance hit in nfsrv_checkgetattr() caused by the acquisition/release of a state lock, even when there were no write delegations issued. This patch add a count of outstanding issued write delegations to the NFSv4 server. This count allows nfsrv_checkgetattr() to return without acquiring any lock when the count is 0, avoiding the performance hit for the case where no write delegations are issued. Reported by: manu Reviewed by: kib MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D13327
2017-12-04 21:50:27 +00:00
nfsrv_writedelegcnt++;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
LIST_INSERT_HEAD(&lfp->lf_deleg, new_deleg, ls_file);
LIST_INSERT_HEAD(NFSSTATEHASH(clp,
new_deleg->ls_stateid), new_deleg, ls_hash);
LIST_INSERT_HEAD(&clp->lc_deleg, new_deleg, ls_list);
new_deleg = NULL;
nfsd: Wrap nfsstatsv1_p in the NFSD_VNET() macro Commit 7344856e3a6d added a lot of macros that will front end vnet macros so that nfsd(8) can run in vnet prison. The nfsstatsv1_p variable got missed. This patch wraps all uses of nfsstatsv1_p with the NFSD_VNET() macro. The NFSD_VNET() macro is still a null macro. MFC after: 3 months
2023-02-16 01:39:07 +00:00
NFSD_VNET(nfsstatsv1_p)->srvdelegates++;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_openpluslock++;
nfsrv_delegatecnt++;
}
} else {
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
new_open->ls_stateid.seqid = 1;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
new_open->ls_stateid.other[0] = clp->lc_clientid.lval[0];
new_open->ls_stateid.other[1] = clp->lc_clientid.lval[1];
new_open->ls_stateid.other[2] = nfsrv_nextstateindex(clp);
new_open->ls_flags = (new_stp->ls_flags & NFSLCK_SHAREBITS)|
NFSLCK_OPEN;
new_open->ls_uid = new_stp->ls_uid;
new_open->ls_openowner = ownerstp;
new_open->ls_lfp = lfp;
new_open->ls_clp = clp;
LIST_INIT(&new_open->ls_open);
LIST_INSERT_HEAD(&lfp->lf_open, new_open, ls_file);
LIST_INSERT_HEAD(&ownerstp->ls_open, new_open, ls_list);
LIST_INSERT_HEAD(NFSSTATEHASH(clp, new_open->ls_stateid),
new_open, ls_hash);
openstp = new_open;
new_open = NULL;
nfsd: Wrap nfsstatsv1_p in the NFSD_VNET() macro Commit 7344856e3a6d added a lot of macros that will front end vnet macros so that nfsd(8) can run in vnet prison. The nfsstatsv1_p variable got missed. This patch wraps all uses of nfsstatsv1_p with the NFSD_VNET() macro. The NFSD_VNET() macro is still a null macro. MFC after: 3 months
2023-02-16 01:39:07 +00:00
NFSD_VNET(nfsstatsv1_p)->srvopens++;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_openpluslock++;
/*
* This is where we can choose to issue a delegation.
*/
Modify the reasons for not issuing a delegation in the NFSv4.1 server. The ESXi NFSv4.1 client will generate warning messages when the reason for not issuing a delegation is two. Two refers to a resource limit and I do not see why it would be considered invalid. However it probably was not the best choice of reason for not issuing a delegation. This patch changes the reasons used to ones that the ESXi client doesn't complain about. This change does not affect the FreeBSD client and does not appear to affect behaviour of the Linux NFSv4.1 client. RFC5661 defines these "reasons" but does not give any guidance w.r.t. which ones are more appropriate to return to a client. Tested by: andreas.nagy@frequentis.com PR: 226650 MFC after: 2 weeks
2018-07-16 21:32:50 +00:00
if ((new_stp->ls_flags & NFSLCK_WANTNODELEG) != 0)
*rflagsp |= NFSV4OPEN_WDNOTWANTED;
else if (nfsrv_issuedelegs == 0)
*rflagsp |= NFSV4OPEN_WDSUPPFTYPE;
else if (NFSRV_V4DELEGLIMIT(nfsrv_delegatecnt))
*rflagsp |= NFSV4OPEN_WDRESOURCE;
else if (delegate == 0 || (writedeleg == 0 &&
readonly == 0) || !NFSVNO_DELEGOK(vp) ||
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
(clp->lc_flags & (LCL_CALLBACKSON | LCL_CBDOWN)) !=
LCL_CALLBACKSON)
*rflagsp |= NFSV4OPEN_WDCONTENTION;
else {
new_deleg->ls_stateid.seqid = delegstateidp->seqid = 1;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
new_deleg->ls_stateid.other[0] = delegstateidp->other[0]
= clp->lc_clientid.lval[0];
new_deleg->ls_stateid.other[1] = delegstateidp->other[1]
= clp->lc_clientid.lval[1];
new_deleg->ls_stateid.other[2] = delegstateidp->other[2]
= nfsrv_nextstateindex(clp);
if (writedeleg && !NFSVNO_EXRDONLY(exp) &&
Revert crap accidentally committed
2017-01-28 16:31:23 +00:00
(nfsrv_writedelegifpos || !readonly) &&
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
(new_stp->ls_flags & NFSLCK_WANTRDELEG) == 0) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
new_deleg->ls_flags = (NFSLCK_DELEGWRITE |
NFSLCK_READACCESS | NFSLCK_WRITEACCESS);
*rflagsp |= NFSV4OPEN_WRITEDELEGATE;
Avoid the overhead of acquiring a lock in nfsrv_checkgetattr() when there are no write delegations issued. manu@ reported on the freebsd-current@ mailing list that there was a significant performance hit in nfsrv_checkgetattr() caused by the acquisition/release of a state lock, even when there were no write delegations issued. This patch add a count of outstanding issued write delegations to the NFSv4 server. This count allows nfsrv_checkgetattr() to return without acquiring any lock when the count is 0, avoiding the performance hit for the case where no write delegations are issued. Reported by: manu Reviewed by: kib MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D13327
2017-12-04 21:50:27 +00:00
nfsrv_writedelegcnt++;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
} else {
new_deleg->ls_flags = (NFSLCK_DELEGREAD |
NFSLCK_READACCESS);
*rflagsp |= NFSV4OPEN_READDELEGATE;
}
new_deleg->ls_uid = new_stp->ls_uid;
new_deleg->ls_lfp = lfp;
new_deleg->ls_clp = clp;
new_deleg->ls_filerev = filerev;
new_deleg->ls_compref = nd->nd_compref;
nfsd: make the server repeat CB_RECALL every couple of seconds Commit 01ae8969a9ee stopped the NFSv4.1/4.2 server from implicitly binding the back channel to a new TCP connection so that it conforms to RFC5661, for NFSv4.1/4.2. An effect of this for the Linux NFS client is that it will do a BindConnectionToSession when it sees NFSV4SEQ_CBPATHDOWN set in a sequence reply. This will fix the back channel, but the first attempt at a callback like CB_RECALL will already have failed. Without this patch, a CB_RECALL will not be retried and that can result in a 5 minute delay until the delegation times out. This patch modifies the code so that it will retry the CB_RECALL every couple of seconds, often avoiding the 5 minute delay. This is not critical for correct behaviour, but avoids the 5 minute delay for the case where the Linux client re-binds the back channel via BindConnectionToSession. MFC after: 2 weeks
2021-04-05 01:15:54 +00:00
new_deleg->ls_lastrecall = 0;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
LIST_INSERT_HEAD(&lfp->lf_deleg, new_deleg, ls_file);
LIST_INSERT_HEAD(NFSSTATEHASH(clp,
new_deleg->ls_stateid), new_deleg, ls_hash);
LIST_INSERT_HEAD(&clp->lc_deleg, new_deleg, ls_list);
new_deleg = NULL;
nfsd: Wrap nfsstatsv1_p in the NFSD_VNET() macro Commit 7344856e3a6d added a lot of macros that will front end vnet macros so that nfsd(8) can run in vnet prison. The nfsstatsv1_p variable got missed. This patch wraps all uses of nfsstatsv1_p with the NFSD_VNET() macro. The NFSD_VNET() macro is still a null macro. MFC after: 3 months
2023-02-16 01:39:07 +00:00
NFSD_VNET(nfsstatsv1_p)->srvdelegates++;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_openpluslock++;
nfsrv_delegatecnt++;
}
}
} else {
/*
* New owner case. Start the open_owner sequence with a
* Needs confirmation (unless a reclaim) and hang the
* new open off it.
*/
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
new_open->ls_stateid.seqid = 1;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
new_open->ls_stateid.other[0] = clp->lc_clientid.lval[0];
new_open->ls_stateid.other[1] = clp->lc_clientid.lval[1];
new_open->ls_stateid.other[2] = nfsrv_nextstateindex(clp);
new_open->ls_flags = (new_stp->ls_flags & NFSLCK_SHAREBITS) |
NFSLCK_OPEN;
new_open->ls_uid = new_stp->ls_uid;
LIST_INIT(&new_open->ls_open);
new_open->ls_openowner = new_stp;
new_open->ls_lfp = lfp;
new_open->ls_clp = clp;
LIST_INSERT_HEAD(&lfp->lf_open, new_open, ls_file);
if (new_stp->ls_flags & NFSLCK_RECLAIM) {
new_stp->ls_flags = 0;
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
} else if ((nd->nd_flag & ND_NFSV41) != 0) {
/* NFSv4.1 never needs confirmation. */
new_stp->ls_flags = 0;
/*
* This is where we can choose to issue a delegation.
*/
if (delegate && nfsrv_issuedelegs &&
(writedeleg || readonly) &&
(clp->lc_flags & (LCL_CALLBACKSON | LCL_CBDOWN)) ==
LCL_CALLBACKSON &&
!NFSRV_V4DELEGLIMIT(nfsrv_delegatecnt) &&
NFSVNO_DELEGOK(vp) &&
((nd->nd_flag & ND_NFSV41) == 0 ||
(new_stp->ls_flags & NFSLCK_WANTNODELEG) == 0)) {
new_deleg->ls_stateid.seqid =
delegstateidp->seqid = 1;
new_deleg->ls_stateid.other[0] =
delegstateidp->other[0]
= clp->lc_clientid.lval[0];
new_deleg->ls_stateid.other[1] =
delegstateidp->other[1]
= clp->lc_clientid.lval[1];
new_deleg->ls_stateid.other[2] =
delegstateidp->other[2]
= nfsrv_nextstateindex(clp);
if (writedeleg && !NFSVNO_EXRDONLY(exp) &&
Revert crap accidentally committed
2017-01-28 16:31:23 +00:00
(nfsrv_writedelegifpos || !readonly) &&
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
((nd->nd_flag & ND_NFSV41) == 0 ||
(new_stp->ls_flags & NFSLCK_WANTRDELEG) ==
0)) {
new_deleg->ls_flags =
(NFSLCK_DELEGWRITE |
NFSLCK_READACCESS |
NFSLCK_WRITEACCESS);
*rflagsp |= NFSV4OPEN_WRITEDELEGATE;
Avoid the overhead of acquiring a lock in nfsrv_checkgetattr() when there are no write delegations issued. manu@ reported on the freebsd-current@ mailing list that there was a significant performance hit in nfsrv_checkgetattr() caused by the acquisition/release of a state lock, even when there were no write delegations issued. This patch add a count of outstanding issued write delegations to the NFSv4 server. This count allows nfsrv_checkgetattr() to return without acquiring any lock when the count is 0, avoiding the performance hit for the case where no write delegations are issued. Reported by: manu Reviewed by: kib MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D13327
2017-12-04 21:50:27 +00:00
nfsrv_writedelegcnt++;
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
} else {
new_deleg->ls_flags =
(NFSLCK_DELEGREAD |
NFSLCK_READACCESS);
*rflagsp |= NFSV4OPEN_READDELEGATE;
}
new_deleg->ls_uid = new_stp->ls_uid;
new_deleg->ls_lfp = lfp;
new_deleg->ls_clp = clp;
new_deleg->ls_filerev = filerev;
new_deleg->ls_compref = nd->nd_compref;
nfsd: make the server repeat CB_RECALL every couple of seconds Commit 01ae8969a9ee stopped the NFSv4.1/4.2 server from implicitly binding the back channel to a new TCP connection so that it conforms to RFC5661, for NFSv4.1/4.2. An effect of this for the Linux NFS client is that it will do a BindConnectionToSession when it sees NFSV4SEQ_CBPATHDOWN set in a sequence reply. This will fix the back channel, but the first attempt at a callback like CB_RECALL will already have failed. Without this patch, a CB_RECALL will not be retried and that can result in a 5 minute delay until the delegation times out. This patch modifies the code so that it will retry the CB_RECALL every couple of seconds, often avoiding the 5 minute delay. This is not critical for correct behaviour, but avoids the 5 minute delay for the case where the Linux client re-binds the back channel via BindConnectionToSession. MFC after: 2 weeks
2021-04-05 01:15:54 +00:00
new_deleg->ls_lastrecall = 0;
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
LIST_INSERT_HEAD(&lfp->lf_deleg, new_deleg,
ls_file);
LIST_INSERT_HEAD(NFSSTATEHASH(clp,
new_deleg->ls_stateid), new_deleg, ls_hash);
LIST_INSERT_HEAD(&clp->lc_deleg, new_deleg,
ls_list);
new_deleg = NULL;
nfsd: Wrap nfsstatsv1_p in the NFSD_VNET() macro Commit 7344856e3a6d added a lot of macros that will front end vnet macros so that nfsd(8) can run in vnet prison. The nfsstatsv1_p variable got missed. This patch wraps all uses of nfsstatsv1_p with the NFSD_VNET() macro. The NFSD_VNET() macro is still a null macro. MFC after: 3 months
2023-02-16 01:39:07 +00:00
NFSD_VNET(nfsstatsv1_p)->srvdelegates++;
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
nfsrv_openpluslock++;
nfsrv_delegatecnt++;
}
Fix the NFSv4.1 server for Open reclaim after a reboot. The NFSv4.1 server failed to update the nfs-stablerestart file for a client when the client was issued its first Open. As such, recovery of Opens after a server reboot failed with NFSERR_NOGRACE. This patch fixes this. It also changes the code so that it malloc()'s the 1024 byte array instead of allocating it on the kernel stack for both NFSv4.0 and NFSv4.1. Note that this bug only affected NFSv4.1 and only when clients attempted to reclaim Opens after a server reboot. MFC after: 2 weeks
2016-12-05 22:36:25 +00:00
/*
* Since NFSv4.1 never does an OpenConfirm, the first
* open state will be acquired here.
*/
if (!(clp->lc_flags & LCL_STAMPEDSTABLE)) {
clp->lc_flags |= LCL_STAMPEDSTABLE;
len = clp->lc_idlen;
NFSBCOPY(clp->lc_id, clidp, len);
gotstate = 1;
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
} else {
*rflagsp |= NFSV4OPEN_RESULTCONFIRM;
new_stp->ls_flags = NFSLCK_NEEDSCONFIRM;
}
nfsrvd_refcache(new_stp->ls_op);
new_stp->ls_noopens = 0;
LIST_INIT(&new_stp->ls_open);
LIST_INSERT_HEAD(&new_stp->ls_open, new_open, ls_list);
LIST_INSERT_HEAD(&clp->lc_open, new_stp, ls_list);
LIST_INSERT_HEAD(NFSSTATEHASH(clp, new_open->ls_stateid),
new_open, ls_hash);
openstp = new_open;
new_open = NULL;
*new_stpp = NULL;
nfsd: Wrap nfsstatsv1_p in the NFSD_VNET() macro Commit 7344856e3a6d added a lot of macros that will front end vnet macros so that nfsd(8) can run in vnet prison. The nfsstatsv1_p variable got missed. This patch wraps all uses of nfsstatsv1_p with the NFSD_VNET() macro. The NFSD_VNET() macro is still a null macro. MFC after: 3 months
2023-02-16 01:39:07 +00:00
NFSD_VNET(nfsstatsv1_p)->srvopens++;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_openpluslock++;
nfsd: Wrap nfsstatsv1_p in the NFSD_VNET() macro Commit 7344856e3a6d added a lot of macros that will front end vnet macros so that nfsd(8) can run in vnet prison. The nfsstatsv1_p variable got missed. This patch wraps all uses of nfsstatsv1_p with the NFSD_VNET() macro. The NFSD_VNET() macro is still a null macro. MFC after: 3 months
2023-02-16 01:39:07 +00:00
NFSD_VNET(nfsstatsv1_p)->srvopenowners++;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_openpluslock++;
}
if (!error) {
stateidp->seqid = openstp->ls_stateid.seqid;
stateidp->other[0] = openstp->ls_stateid.other[0];
stateidp->other[1] = openstp->ls_stateid.other[1];
stateidp->other[2] = openstp->ls_stateid.other[2];
}
NFSUNLOCKSTATE();
if (haslock) {
NFSLOCKV4ROOTMUTEX();
nfsv4_unlock(&nfsv4rootfs_lock, 1);
NFSUNLOCKV4ROOTMUTEX();
}
if (new_open)
style: Remove remaining deprecated MALLOC/FREE macros Mechanically replace uses of MALLOC/FREE with appropriate invocations of malloc(9) / free(9) (a series of sed expressions). Something like: * MALLOC(a, b, ... -> a = malloc(... * FREE( -> free( * free((caddr_t) -> free( No functional change. For now, punt on modifying contrib ipfilter code, leaving a definition of the macro in its KMALLOC(). Reported by: jhb Reviewed by: cy, imp, markj, rmacklem Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D14035
2018-01-25 22:25:13 +00:00
free(new_open, M_NFSDSTATE);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (new_deleg)
style: Remove remaining deprecated MALLOC/FREE macros Mechanically replace uses of MALLOC/FREE with appropriate invocations of malloc(9) / free(9) (a series of sed expressions). Something like: * MALLOC(a, b, ... -> a = malloc(... * FREE( -> free( * free((caddr_t) -> free( No functional change. For now, punt on modifying contrib ipfilter code, leaving a definition of the macro in its KMALLOC(). Reported by: jhb Reviewed by: cy, imp, markj, rmacklem Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D14035
2018-01-25 22:25:13 +00:00
free(new_deleg, M_NFSDSTATE);
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
Fix the NFSv4.1 server for Open reclaim after a reboot. The NFSv4.1 server failed to update the nfs-stablerestart file for a client when the client was issued its first Open. As such, recovery of Opens after a server reboot failed with NFSERR_NOGRACE. This patch fixes this. It also changes the code so that it malloc()'s the 1024 byte array instead of allocating it on the kernel stack for both NFSv4.0 and NFSv4.1. Note that this bug only affected NFSv4.1 and only when clients attempted to reclaim Opens after a server reboot. MFC after: 2 weeks
2016-12-05 22:36:25 +00:00
/*
* If the NFSv4.1 client just acquired its first open, write a timestamp
* to the stable storage file.
*/
if (gotstate != 0) {
nfsrv_writestable(clidp, len, NFSNST_NEWSTATE, p);
nfsrv_backupstable();
}
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
out:
Fix the NFSv4.1 server for Open reclaim after a reboot. The NFSv4.1 server failed to update the nfs-stablerestart file for a client when the client was issued its first Open. As such, recovery of Opens after a server reboot failed with NFSERR_NOGRACE. This patch fixes this. It also changes the code so that it malloc()'s the 1024 byte array instead of allocating it on the kernel stack for both NFSv4.0 and NFSv4.1. Note that this bug only affected NFSv4.1 and only when clients attempted to reclaim Opens after a server reboot. MFC after: 2 weeks
2016-12-05 22:36:25 +00:00
free(clidp, M_TEMP);
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
NFSEXITCODE2(error, nd);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
return (error);
}
/*
* Open update. Does the confirm, downgrade and close.
*/
nfs: Remove APPLESTATIC macro It is no longer useful. Reviewed by: rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24811
2020-05-12 13:23:25 +00:00
int
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_openupdate(vnode_t vp, struct nfsstate *new_stp, nfsquad_t clientid,
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
nfsv4stateid_t *stateidp, struct nfsrv_descript *nd, NFSPROC_T *p,
int *retwriteaccessp)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
{
Do pass removing some write-only variables from the kernel. This reduces noise when kernel is compiled by newer GCC versions, such as one used by external toolchain ports. Reviewed by: kib, andrew(sys/arm and sys/arm64), emaste(partial), erj(partial) Reviewed by: jhb (sys/dev/pci/* sys/kern/vfs_aio.c and sys/kern/kern_synch.c) Differential Revision: https://reviews.freebsd.org/D10385
2017-12-25 04:48:39 +00:00
struct nfsstate *stp;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
struct nfsclient *clp;
struct nfslockfile *lfp;
u_int32_t bits;
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
int error = 0, gotstate = 0, len = 0;
Fix the NFSv4.1 server for Open reclaim after a reboot. The NFSv4.1 server failed to update the nfs-stablerestart file for a client when the client was issued its first Open. As such, recovery of Opens after a server reboot failed with NFSERR_NOGRACE. This patch fixes this. It also changes the code so that it malloc()'s the 1024 byte array instead of allocating it on the kernel stack for both NFSv4.0 and NFSv4.1. Note that this bug only affected NFSv4.1 and only when clients attempted to reclaim Opens after a server reboot. MFC after: 2 weeks
2016-12-05 22:36:25 +00:00
u_char *clidp = NULL;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* Check for restart conditions (client and server).
*/
error = nfsrv_checkrestart(clientid, new_stp->ls_flags,
&new_stp->ls_stateid, 0);
if (error)
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
Fix the NFSv4.1 server for Open reclaim after a reboot. The NFSv4.1 server failed to update the nfs-stablerestart file for a client when the client was issued its first Open. As such, recovery of Opens after a server reboot failed with NFSERR_NOGRACE. This patch fixes this. It also changes the code so that it malloc()'s the 1024 byte array instead of allocating it on the kernel stack for both NFSv4.0 and NFSv4.1. Note that this bug only affected NFSv4.1 and only when clients attempted to reclaim Opens after a server reboot. MFC after: 2 weeks
2016-12-05 22:36:25 +00:00
clidp = malloc(NFSV4_OPAQUELIMIT, M_TEMP, M_WAITOK);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
NFSLOCKSTATE();
/*
* Get the open structure via clientid and stateid.
*/
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
error = nfsrv_getclient(clientid, CLOPS_RENEW, &clp, NULL,
(nfsquad_t)((u_quad_t)0), 0, nd, p);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (!error)
error = nfsrv_getstate(clp, &new_stp->ls_stateid,
new_stp->ls_flags, &stp);
/*
* Sanity check the open.
*/
if (!error && (!(stp->ls_flags & NFSLCK_OPEN) ||
(!(new_stp->ls_flags & NFSLCK_CONFIRM) &&
(stp->ls_openowner->ls_flags & NFSLCK_NEEDSCONFIRM)) ||
((new_stp->ls_flags & NFSLCK_CONFIRM) &&
(!(stp->ls_openowner->ls_flags & NFSLCK_NEEDSCONFIRM)))))
error = NFSERR_BADSTATEID;
if (!error)
error = nfsrv_checkseqid(nd, new_stp->ls_seq,
stp->ls_openowner, new_stp->ls_op);
if (!error && stp->ls_stateid.seqid != new_stp->ls_stateid.seqid &&
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
(((nd->nd_flag & ND_NFSV41) == 0 &&
!(new_stp->ls_flags & NFSLCK_CONFIRM)) ||
((nd->nd_flag & ND_NFSV41) != 0 &&
new_stp->ls_stateid.seqid != 0)))
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
error = NFSERR_OLDSTATEID;
nfsd: Clean up the code by not using the vnode_vtype() macro The vnode_vtype() macro was used to make the code compatible with Mac OSX, for the Mac OSX port. For FreeBSD, this macro just obscured the code, so avoid using it to clean up the code. This commit should not result in a semantics change.
2022-06-22 20:20:32 +00:00
if (!error && vp->v_type != VREG) {
if (vp->v_type == VDIR)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
error = NFSERR_ISDIR;
else
error = NFSERR_INVAL;
}
if (error) {
/*
* If a client tries to confirm an Open with a bad
* seqid# and there are no byte range locks or other Opens
* on the openowner, just throw it away, so the next use of the
* openowner will start a fresh seq#.
*/
if (error == NFSERR_BADSEQID &&
(new_stp->ls_flags & NFSLCK_CONFIRM) &&
nfsrv_nootherstate(stp))
nfsrv_freeopenowner(stp->ls_openowner, 0, p);
NFSUNLOCKSTATE();
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* Set the return stateid.
*/
stateidp->seqid = stp->ls_stateid.seqid + 1;
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
if ((nd->nd_flag & ND_NFSV41) != 0 && stateidp->seqid == 0)
stateidp->seqid = 1;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
stateidp->other[0] = stp->ls_stateid.other[0];
stateidp->other[1] = stp->ls_stateid.other[1];
stateidp->other[2] = stp->ls_stateid.other[2];
/*
* Now, handle the three cases.
*/
if (new_stp->ls_flags & NFSLCK_CONFIRM) {
/*
* If the open doesn't need confirmation, it seems to me that
* there is a client error, but I'll just log it and keep going?
*/
if (!(stp->ls_openowner->ls_flags & NFSLCK_NEEDSCONFIRM))
printf("Nfsv4d: stray open confirm\n");
stp->ls_openowner->ls_flags = 0;
stp->ls_stateid.seqid++;
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
if ((nd->nd_flag & ND_NFSV41) != 0 &&
stp->ls_stateid.seqid == 0)
stp->ls_stateid.seqid = 1;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (!(clp->lc_flags & LCL_STAMPEDSTABLE)) {
clp->lc_flags |= LCL_STAMPEDSTABLE;
len = clp->lc_idlen;
Fix the NFSv4.1 server for Open reclaim after a reboot. The NFSv4.1 server failed to update the nfs-stablerestart file for a client when the client was issued its first Open. As such, recovery of Opens after a server reboot failed with NFSERR_NOGRACE. This patch fixes this. It also changes the code so that it malloc()'s the 1024 byte array instead of allocating it on the kernel stack for both NFSv4.0 and NFSv4.1. Note that this bug only affected NFSv4.1 and only when clients attempted to reclaim Opens after a server reboot. MFC after: 2 weeks
2016-12-05 22:36:25 +00:00
NFSBCOPY(clp->lc_id, clidp, len);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
gotstate = 1;
}
NFSUNLOCKSTATE();
} else if (new_stp->ls_flags & NFSLCK_CLOSE) {
lfp = stp->ls_lfp;
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
if (retwriteaccessp != NULL) {
if ((stp->ls_flags & NFSLCK_WRITEACCESS) != 0)
*retwriteaccessp = 1;
else
*retwriteaccessp = 0;
}
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
if (nfsrv_dolocallocks != 0 && !LIST_EMPTY(&stp->ls_open)) {
/* Get the lf lock */
nfsrv_locklf(lfp);
NFSUNLOCKSTATE();
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
ASSERT_VOP_ELOCKED(vp, "nfsrv_openupdate");
vfs: drop the mostly unused flags argument from VOP_UNLOCK Filesystems which want to use it in limited capacity can employ the VOP_UNLOCK_FLAGS macro. Reviewed by: kib (previous version) Differential Revision: https://reviews.freebsd.org/D21427
2020-01-03 22:29:58 +00:00
NFSVOPUNLOCK(vp);
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
if (nfsrv_freeopen(stp, vp, 1, p) == 0) {
NFSLOCKSTATE();
nfsrv_unlocklf(lfp);
NFSUNLOCKSTATE();
}
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
NFSVOPLOCK(vp, LK_EXCLUSIVE | LK_RETRY);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
} else {
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
(void) nfsrv_freeopen(stp, NULL, 0, p);
NFSUNLOCKSTATE();
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
} else {
/*
* Update the share bits, making sure that the new set are a
* subset of the old ones.
*/
bits = (new_stp->ls_flags & NFSLCK_SHAREBITS);
if (~(stp->ls_flags) & bits) {
NFSUNLOCKSTATE();
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
error = NFSERR_INVAL;
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
stp->ls_flags = (bits | NFSLCK_OPEN);
stp->ls_stateid.seqid++;
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
if ((nd->nd_flag & ND_NFSV41) != 0 &&
stp->ls_stateid.seqid == 0)
stp->ls_stateid.seqid = 1;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
NFSUNLOCKSTATE();
}
/*
* If the client just confirmed its first open, write a timestamp
* to the stable storage file.
*/
Modify the experimental NFSv4 server so that it posts a SIGUSR2 signal to the master nfsd daemon whenever the stable restart file has been modified. This will allow the master nfsd daemon to maintain an up to date backup copy of the file. This is enabled via the nfssvc() syscall, so that older nfsd daemons will not be signaled. Reviewed by: jhb MFC after: 1 week
2011-01-14 23:30:35 +00:00
if (gotstate != 0) {
Fix the NFSv4.1 server for Open reclaim after a reboot. The NFSv4.1 server failed to update the nfs-stablerestart file for a client when the client was issued its first Open. As such, recovery of Opens after a server reboot failed with NFSERR_NOGRACE. This patch fixes this. It also changes the code so that it malloc()'s the 1024 byte array instead of allocating it on the kernel stack for both NFSv4.0 and NFSv4.1. Note that this bug only affected NFSv4.1 and only when clients attempted to reclaim Opens after a server reboot. MFC after: 2 weeks
2016-12-05 22:36:25 +00:00
nfsrv_writestable(clidp, len, NFSNST_NEWSTATE, p);
Modify the experimental NFSv4 server so that it posts a SIGUSR2 signal to the master nfsd daemon whenever the stable restart file has been modified. This will allow the master nfsd daemon to maintain an up to date backup copy of the file. This is enabled via the nfssvc() syscall, so that older nfsd daemons will not be signaled. Reviewed by: jhb MFC after: 1 week
2011-01-14 23:30:35 +00:00
nfsrv_backupstable();
}
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
out:
Fix the NFSv4.1 server for Open reclaim after a reboot. The NFSv4.1 server failed to update the nfs-stablerestart file for a client when the client was issued its first Open. As such, recovery of Opens after a server reboot failed with NFSERR_NOGRACE. This patch fixes this. It also changes the code so that it malloc()'s the 1024 byte array instead of allocating it on the kernel stack for both NFSv4.0 and NFSv4.1. Note that this bug only affected NFSv4.1 and only when clients attempted to reclaim Opens after a server reboot. MFC after: 2 weeks
2016-12-05 22:36:25 +00:00
free(clidp, M_TEMP);
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
NFSEXITCODE2(error, nd);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
return (error);
}
/*
* Delegation update. Does the purge and return.
*/
nfs: Remove APPLESTATIC macro It is no longer useful. Reviewed by: rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24811
2020-05-12 13:23:25 +00:00
int
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
nfsrv_delegupdate(struct nfsrv_descript *nd, nfsquad_t clientid,
nfsv4stateid_t *stateidp, vnode_t vp, int op, struct ucred *cred,
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
NFSPROC_T *p, int *retwriteaccessp)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
{
struct nfsstate *stp;
struct nfsclient *clp;
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
int error = 0;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
fhandle_t fh;
/*
* Do a sanity check against the file handle for DelegReturn.
*/
if (vp) {
error = nfsvno_getfh(vp, &fh, p);
if (error)
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* Check for restart conditions (client and server).
*/
if (op == NFSV4OP_DELEGRETURN)
error = nfsrv_checkrestart(clientid, NFSLCK_DELEGRETURN,
stateidp, 0);
else
error = nfsrv_checkrestart(clientid, NFSLCK_DELEGPURGE,
stateidp, 0);
NFSLOCKSTATE();
/*
* Get the open structure via clientid and stateid.
*/
if (!error)
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
error = nfsrv_getclient(clientid, CLOPS_RENEW, &clp, NULL,
(nfsquad_t)((u_quad_t)0), 0, nd, p);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (error) {
if (error == NFSERR_CBPATHDOWN)
error = 0;
if (error == NFSERR_STALECLIENTID && op == NFSV4OP_DELEGRETURN)
error = NFSERR_STALESTATEID;
}
if (!error && op == NFSV4OP_DELEGRETURN) {
error = nfsrv_getstate(clp, stateidp, NFSLCK_DELEGRETURN, &stp);
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
if (!error && stp->ls_stateid.seqid != stateidp->seqid &&
((nd->nd_flag & ND_NFSV41) == 0 || stateidp->seqid != 0))
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
error = NFSERR_OLDSTATEID;
}
/*
* NFSERR_EXPIRED means that the state has gone away,
* so Delegations have been purged. Just return ok.
*/
if (error == NFSERR_EXPIRED && op == NFSV4OP_DELEGPURGE) {
NFSUNLOCKSTATE();
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
error = 0;
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
if (error) {
NFSUNLOCKSTATE();
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
if (op == NFSV4OP_DELEGRETURN) {
if (NFSBCMP((caddr_t)&fh, (caddr_t)&stp->ls_lfp->lf_fh,
sizeof (fhandle_t))) {
NFSUNLOCKSTATE();
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
error = NFSERR_BADSTATEID;
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
if (retwriteaccessp != NULL) {
if ((stp->ls_flags & NFSLCK_DELEGWRITE) != 0)
*retwriteaccessp = 1;
else
*retwriteaccessp = 0;
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_freedeleg(stp);
} else {
nfsrv_freedeleglist(&clp->lc_olddeleg);
}
NFSUNLOCKSTATE();
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
error = 0;
out:
NFSEXITCODE(error);
return (error);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* Release lock owner.
*/
nfs: Remove APPLESTATIC macro It is no longer useful. Reviewed by: rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24811
2020-05-12 13:23:25 +00:00
int
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_releaselckown(struct nfsstate *new_stp, nfsquad_t clientid,
NFSPROC_T *p)
{
struct nfsstate *stp, *nstp, *openstp, *ownstp;
struct nfsclient *clp;
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
int error = 0;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* Check for restart conditions (client and server).
*/
error = nfsrv_checkrestart(clientid, new_stp->ls_flags,
&new_stp->ls_stateid, 0);
if (error)
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
NFSLOCKSTATE();
/*
* Get the lock owner by name.
*/
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
error = nfsrv_getclient(clientid, CLOPS_RENEW, &clp, NULL,
(nfsquad_t)((u_quad_t)0), 0, NULL, p);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (error) {
NFSUNLOCKSTATE();
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
LIST_FOREACH(ownstp, &clp->lc_open, ls_list) {
LIST_FOREACH(openstp, &ownstp->ls_open, ls_list) {
stp = LIST_FIRST(&openstp->ls_open);
while (stp != LIST_END(&openstp->ls_open)) {
nstp = LIST_NEXT(stp, ls_list);
/*
* If the owner matches, check for locks and
* then free or return an error.
*/
if (stp->ls_ownerlen == new_stp->ls_ownerlen &&
!NFSBCMP(stp->ls_owner, new_stp->ls_owner,
stp->ls_ownerlen)){
if (LIST_EMPTY(&stp->ls_lock)) {
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
nfsrv_freelockowner(stp, NULL, 0, p);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
} else {
NFSUNLOCKSTATE();
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
error = NFSERR_LOCKSHELD;
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
}
stp = nstp;
}
}
}
NFSUNLOCKSTATE();
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
out:
NFSEXITCODE(error);
return (error);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* Get the file handle for a lock structure.
*/
static int
Remove an unnecessary level of indirection for an argument. This simplifies the code and should avoid the clang sparc port from generating an abort() call. Requested by: rdivacky Submitted by: jhb MFC after: 2 weeks
2014-04-23 23:13:46 +00:00
nfsrv_getlockfh(vnode_t vp, u_short flags, struct nfslockfile *new_lfp,
fhandle_t *nfhp, NFSPROC_T *p)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
{
fhandle_t *fhp = NULL;
int error;
/*
* For lock, use the new nfslock structure, otherwise just
* a fhandle_t on the stack.
*/
if (flags & NFSLCK_OPEN) {
Remove an unnecessary level of indirection for an argument. This simplifies the code and should avoid the clang sparc port from generating an abort() call. Requested by: rdivacky Submitted by: jhb MFC after: 2 weeks
2014-04-23 23:13:46 +00:00
KASSERT(new_lfp != NULL, ("nfsrv_getlockfh: new_lfp NULL"));
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
fhp = &new_lfp->lf_fh;
} else if (nfhp) {
fhp = nfhp;
} else {
panic("nfsrv_getlockfh");
}
error = nfsvno_getfh(vp, fhp, p);
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
NFSEXITCODE(error);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
return (error);
}
/*
* Get an nfs lock structure. Allocate one, as required, and return a
* pointer to it.
* Returns an NFSERR_xxx upon failure or -1 to indicate no current lock.
*/
static int
nfsrv_getlockfile(u_short flags, struct nfslockfile **new_lfpp,
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
struct nfslockfile **lfpp, fhandle_t *nfhp, int lockit)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
{
struct nfslockfile *lfp;
fhandle_t *fhp = NULL, *tfhp;
struct nfslockhashhead *hp;
struct nfslockfile *new_lfp = NULL;
/*
* For lock, use the new nfslock structure, otherwise just
* a fhandle_t on the stack.
*/
if (flags & NFSLCK_OPEN) {
new_lfp = *new_lfpp;
fhp = &new_lfp->lf_fh;
} else if (nfhp) {
fhp = nfhp;
} else {
panic("nfsrv_getlockfile");
}
hp = NFSLOCKHASH(fhp);
LIST_FOREACH(lfp, hp, lf_hash) {
tfhp = &lfp->lf_fh;
if (NFSVNO_CMPFH(fhp, tfhp)) {
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
if (lockit)
nfsrv_locklf(lfp);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
*lfpp = lfp;
return (0);
}
}
if (!(flags & NFSLCK_OPEN))
return (-1);
/*
* No match, so chain the new one into the list.
*/
LIST_INIT(&new_lfp->lf_open);
LIST_INIT(&new_lfp->lf_lock);
LIST_INIT(&new_lfp->lf_deleg);
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
LIST_INIT(&new_lfp->lf_locallock);
LIST_INIT(&new_lfp->lf_rollback);
new_lfp->lf_locallock_lck.nfslock_usecnt = 0;
new_lfp->lf_locallock_lck.nfslock_lock = 0;
new_lfp->lf_usecount = 0;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
LIST_INSERT_HEAD(hp, new_lfp, lf_hash);
*lfpp = new_lfp;
*new_lfpp = NULL;
return (0);
}
/*
* This function adds a nfslock lock structure to the list for the associated
* nfsstate and nfslockfile structures. It will be inserted after the
* entry pointed at by insert_lop.
*/
static void
nfsrv_insertlock(struct nfslock *new_lop, struct nfslock *insert_lop,
struct nfsstate *stp, struct nfslockfile *lfp)
{
struct nfslock *lop, *nlop;
new_lop->lo_stp = stp;
new_lop->lo_lfp = lfp;
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
if (stp != NULL) {
/* Insert in increasing lo_first order */
lop = LIST_FIRST(&lfp->lf_lock);
if (lop == LIST_END(&lfp->lf_lock) ||
new_lop->lo_first <= lop->lo_first) {
LIST_INSERT_HEAD(&lfp->lf_lock, new_lop, lo_lckfile);
} else {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nlop = LIST_NEXT(lop, lo_lckfile);
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
while (nlop != LIST_END(&lfp->lf_lock) &&
nlop->lo_first < new_lop->lo_first) {
lop = nlop;
nlop = LIST_NEXT(lop, lo_lckfile);
}
LIST_INSERT_AFTER(lop, new_lop, lo_lckfile);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
} else {
new_lop->lo_lckfile.le_prev = NULL; /* list not used */
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
* Insert after insert_lop, which is overloaded as stp or lfp for
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
* an empty list.
*/
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
if (stp == NULL && (struct nfslockfile *)insert_lop == lfp)
LIST_INSERT_HEAD(&lfp->lf_locallock, new_lop, lo_lckowner);
else if ((struct nfsstate *)insert_lop == stp)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
LIST_INSERT_HEAD(&stp->ls_lock, new_lop, lo_lckowner);
else
LIST_INSERT_AFTER(insert_lop, new_lop, lo_lckowner);
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
if (stp != NULL) {
nfsd: Wrap nfsstatsv1_p in the NFSD_VNET() macro Commit 7344856e3a6d added a lot of macros that will front end vnet macros so that nfsd(8) can run in vnet prison. The nfsstatsv1_p variable got missed. This patch wraps all uses of nfsstatsv1_p with the NFSD_VNET() macro. The NFSD_VNET() macro is still a null macro. MFC after: 3 months
2023-02-16 01:39:07 +00:00
NFSD_VNET(nfsstatsv1_p)->srvlocks++;
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
nfsrv_openpluslock++;
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* This function updates the locking for a lock owner and given file. It
* maintains a list of lock ranges ordered on increasing file offset that
* are NFSLCK_READ or NFSLCK_WRITE and non-overlapping (aka POSIX style).
* It always adds new_lop to the list and sometimes uses the one pointed
* at by other_lopp.
*/
static void
nfsrv_updatelock(struct nfsstate *stp, struct nfslock **new_lopp,
struct nfslock **other_lopp, struct nfslockfile *lfp)
{
struct nfslock *new_lop = *new_lopp;
struct nfslock *lop, *tlop, *ilop;
struct nfslock *other_lop = *other_lopp;
int unlock = 0, myfile = 0;
u_int64_t tmp;
/*
* Work down the list until the lock is merged.
*/
if (new_lop->lo_flags & NFSLCK_UNLOCK)
unlock = 1;
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
if (stp != NULL) {
ilop = (struct nfslock *)stp;
lop = LIST_FIRST(&stp->ls_lock);
} else {
ilop = (struct nfslock *)lfp;
lop = LIST_FIRST(&lfp->lf_locallock);
}
while (lop != NULL) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* Only check locks for this file that aren't before the start of
* new lock's range.
*/
if (lop->lo_lfp == lfp) {
myfile = 1;
if (lop->lo_end >= new_lop->lo_first) {
if (new_lop->lo_end < lop->lo_first) {
/*
* If the new lock ends before the start of the
* current lock's range, no merge, just insert
* the new lock.
*/
break;
}
if (new_lop->lo_flags == lop->lo_flags ||
(new_lop->lo_first <= lop->lo_first &&
new_lop->lo_end >= lop->lo_end)) {
/*
* This lock can be absorbed by the new lock/unlock.
* This happens when it covers the entire range
* of the old lock or is contiguous
* with the old lock and is of the same type or an
* unlock.
*/
if (lop->lo_first < new_lop->lo_first)
new_lop->lo_first = lop->lo_first;
if (lop->lo_end > new_lop->lo_end)
new_lop->lo_end = lop->lo_end;
tlop = lop;
lop = LIST_NEXT(lop, lo_lckowner);
nfsrv_freenfslock(tlop);
continue;
}
/*
* All these cases are for contiguous locks that are not the
* same type, so they can't be merged.
*/
if (new_lop->lo_first <= lop->lo_first) {
/*
* This case is where the new lock overlaps with the
* first part of the old lock. Move the start of the
* old lock to just past the end of the new lock. The
* new lock will be inserted in front of the old, since
* ilop hasn't been updated. (We are done now.)
*/
lop->lo_first = new_lop->lo_end;
break;
}
if (new_lop->lo_end >= lop->lo_end) {
/*
* This case is where the new lock overlaps with the
* end of the old lock's range. Move the old lock's
* end to just before the new lock's first and insert
* the new lock after the old lock.
* Might not be done yet, since the new lock could
* overlap further locks with higher ranges.
*/
lop->lo_end = new_lop->lo_first;
ilop = lop;
lop = LIST_NEXT(lop, lo_lckowner);
continue;
}
/*
* The final case is where the new lock's range is in the
* middle of the current lock's and splits the current lock
* up. Use *other_lopp to handle the second part of the
* split old lock range. (We are done now.)
* For unlock, we use new_lop as other_lop and tmp, since
* other_lop and new_lop are the same for this case.
* We noted the unlock case above, so we don't need
* new_lop->lo_flags any longer.
*/
tmp = new_lop->lo_first;
if (other_lop == NULL) {
if (!unlock)
panic("nfsd srv update unlock");
other_lop = new_lop;
*new_lopp = NULL;
}
other_lop->lo_first = new_lop->lo_end;
other_lop->lo_end = lop->lo_end;
other_lop->lo_flags = lop->lo_flags;
other_lop->lo_stp = stp;
other_lop->lo_lfp = lfp;
lop->lo_end = tmp;
nfsrv_insertlock(other_lop, lop, stp, lfp);
*other_lopp = NULL;
ilop = lop;
break;
}
}
ilop = lop;
lop = LIST_NEXT(lop, lo_lckowner);
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
if (myfile && (lop == NULL || lop->lo_lfp != lfp))
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
break;
}
/*
* Insert the new lock in the list at the appropriate place.
*/
if (!unlock) {
nfsrv_insertlock(new_lop, ilop, stp, lfp);
*new_lopp = NULL;
}
}
/*
* This function handles sequencing of locks, etc.
* It returns an error that indicates what the caller should do.
*/
static int
nfsrv_checkseqid(struct nfsrv_descript *nd, u_int32_t seqid,
struct nfsstate *stp, struct nfsrvcache *op)
{
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
int error = 0;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
if ((nd->nd_flag & ND_NFSV41) != 0)
/* NFSv4.1 ignores the open_seqid and lock_seqid. */
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (op != nd->nd_rp)
panic("nfsrvstate checkseqid");
if (!(op->rc_flag & RC_INPROG))
panic("nfsrvstate not inprog");
if (stp->ls_op && stp->ls_op->rc_refcnt <= 0) {
printf("refcnt=%d\n", stp->ls_op->rc_refcnt);
panic("nfsrvstate op refcnt");
}
nfsd: fix NFSv4.0 seqid handling for ERELOOKUP Commit 774a36851e0e fixed the NFS server so that it could handle ERELOOKUP returns from VOP calls by redoing the operation/RPC. However, for NFSv4.0, redoing an Open would increment the open_owner's seqid multiple times, breaking the protocol. This patch sets a new flag called ND_ERELOOKUP on the RPC when a redo is in progress. Then the code that increments the seqid avoids the seqid increment/check when the flag is set, since it indicates this has already been done for the Open.
2021-01-01 22:21:51 +00:00
/* If ND_ERELOOKUP is set, the seqid has already been handled. */
if ((nd->nd_flag & ND_ERELOOKUP) != 0)
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if ((stp->ls_seq + 1) == seqid) {
if (stp->ls_op)
nfsrvd_derefcache(stp->ls_op);
stp->ls_op = op;
nfsrvd_refcache(op);
stp->ls_seq = seqid;
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
} else if (stp->ls_seq == seqid && stp->ls_op &&
op->rc_xid == stp->ls_op->rc_xid &&
op->rc_refcnt == 0 &&
op->rc_reqlen == stp->ls_op->rc_reqlen &&
op->rc_cksum == stp->ls_op->rc_cksum) {
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
if (stp->ls_op->rc_flag & RC_INPROG) {
error = NFSERR_DONTREPLY;
goto out;
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nd->nd_rp = stp->ls_op;
nd->nd_rp->rc_flag |= RC_INPROG;
nfsrvd_delcache(op);
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
error = NFSERR_REPLYFROMCACHE;
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
error = NFSERR_BADSEQID;
out:
NFSEXITCODE2(error, nd);
return (error);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* Get the client ip address for callbacks. If the strings can't be parsed,
* just set lc_program to 0 to indicate no callbacks are possible.
* (For cases where the address can't be parsed or is 0.0.0.0.0.0, set
* the address to the client's transport address. This won't be used
Update the nfsstats structure to include the changes needed by the patch in D1626 plus changes so that it includes counts for NFSv4.1 (and the draft of NFSv4.2). Also, make all the counts uint64_t and add a vers field at the beginning, so that future revisions can easily be implemented. There is code in place to handle the old vesion of the nfsstats structure for backwards binary compatibility. Subsequent commits will update nfsstat(8) to use the new fields. Submitted by: will (earlier version) Reviewed by: ken MFC after: 1 month Relnotes: yes Differential Revision: https://reviews.freebsd.org/D1626
2016-08-12 22:44:59 +00:00
* for callbacks, but can be printed out by nfsstats for info.)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
* Return error if the xdr can't be parsed, 0 otherwise.
*/
nfs: Remove APPLESTATIC macro It is no longer useful. Reviewed by: rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24811
2020-05-12 13:23:25 +00:00
int
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_getclientipaddr(struct nfsrv_descript *nd, struct nfsclient *clp)
{
u_int32_t *tl;
u_char *cp, *cp2;
Add support for INET6 addresses to the kernel code that dumps open/lock state. PR#223036 reported that INET6 callback addresses were not printed by nfsdumpstate(8). This kernel patch adds INET6 addresses to the dump structure, so that nfsdumpstate(8) can print them out, post-r346190. The patch also includes the addition of #ifdef INET, INET6 as requested by bz@. PR: 223036 Reviewed by: bz, rgrimes MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D19839
2019-04-13 22:00:09 +00:00
int i, j, maxalen = 0, minalen = 0;
sa_family_t af;
#ifdef INET
Silence some "might not be initialized" warnings for riscv64. None of these case were actually using the variable(s) uninitialized, but I figured that silencing the warnings via initializing them made sense. Some of these predated r355677.
2019-12-13 21:38:08 +00:00
struct sockaddr_in *rin = NULL, *sin;
Add support for INET6 addresses to the kernel code that dumps open/lock state. PR#223036 reported that INET6 callback addresses were not printed by nfsdumpstate(8). This kernel patch adds INET6 addresses to the dump structure, so that nfsdumpstate(8) can print them out, post-r346190. The patch also includes the addition of #ifdef INET, INET6 as requested by bz@. PR: 223036 Reviewed by: bz, rgrimes MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D19839
2019-04-13 22:00:09 +00:00
#endif
#ifdef INET6
Silence some "might not be initialized" warnings for riscv64. None of these case were actually using the variable(s) uninitialized, but I figured that silencing the warnings via initializing them made sense. Some of these predated r355677.
2019-12-13 21:38:08 +00:00
struct sockaddr_in6 *rin6 = NULL, *sin6;
Add support for INET6 addresses to the kernel code that dumps open/lock state. PR#223036 reported that INET6 callback addresses were not printed by nfsdumpstate(8). This kernel patch adds INET6 addresses to the dump structure, so that nfsdumpstate(8) can print them out, post-r346190. The patch also includes the addition of #ifdef INET, INET6 as requested by bz@. PR: 223036 Reviewed by: bz, rgrimes MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D19839
2019-04-13 22:00:09 +00:00
#endif
u_char *addr;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
int error = 0, cantparse = 0;
union {
Fix the client IP address reported by nfsdumpstate for 64bit arch and NFSv4.1. The client IP address was not being reported for some NFSv4 mounts by nfsdumpstate. Upon investigation, two problems were found for mounts using IPv4. One was that the code (originally written and tested on i386) assumed that a "u_long" was a "uint32_t" and would exactly store an IPv4 host address. Not correct for 64bit arches. Also, for NFSv4.1 mounts, the field was not being filled in. This was basically correct, because NFSv4.1 does not use a callback address. However, it meant that nfsdumpstate could not report the client IP addr. This patch should fix both of these issues. For IPv6, the address will still not be reported. The original NFSv4 RFC only specified IPv4 callback addresses. I think this has changed and, if so, a future commit to fix reporting of IPv6 addresses will be needed. Reported by: manu PR: 223036 MFC after: 2 weeks
2017-10-15 22:22:27 +00:00
in_addr_t ival;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
u_char cval[4];
} ip;
union {
Fix the client IP address reported by nfsdumpstate for 64bit arch and NFSv4.1. The client IP address was not being reported for some NFSv4 mounts by nfsdumpstate. Upon investigation, two problems were found for mounts using IPv4. One was that the code (originally written and tested on i386) assumed that a "u_long" was a "uint32_t" and would exactly store an IPv4 host address. Not correct for 64bit arches. Also, for NFSv4.1 mounts, the field was not being filled in. This was basically correct, because NFSv4.1 does not use a callback address. However, it meant that nfsdumpstate could not report the client IP addr. This patch should fix both of these issues. For IPv6, the address will still not be reported. The original NFSv4 RFC only specified IPv4 callback addresses. I think this has changed and, if so, a future commit to fix reporting of IPv6 addresses will be needed. Reported by: manu PR: 223036 MFC after: 2 weeks
2017-10-15 22:22:27 +00:00
in_port_t sval;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
u_char cval[2];
} port;
Add support for INET6 addresses to the kernel code that dumps open/lock state. PR#223036 reported that INET6 callback addresses were not printed by nfsdumpstate(8). This kernel patch adds INET6 addresses to the dump structure, so that nfsdumpstate(8) can print them out, post-r346190. The patch also includes the addition of #ifdef INET, INET6 as requested by bz@. PR: 223036 Reviewed by: bz, rgrimes MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D19839
2019-04-13 22:00:09 +00:00
/* 8 is the maximum length of the port# string. */
addr = malloc(INET6_ADDRSTRLEN + 8, M_TEMP, M_WAITOK);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
clp->lc_req.nr_client = NULL;
clp->lc_req.nr_lock = 0;
Add support for INET6 addresses to the kernel code that dumps open/lock state. PR#223036 reported that INET6 callback addresses were not printed by nfsdumpstate(8). This kernel patch adds INET6 addresses to the dump structure, so that nfsdumpstate(8) can print them out, post-r346190. The patch also includes the addition of #ifdef INET, INET6 as requested by bz@. PR: 223036 Reviewed by: bz, rgrimes MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D19839
2019-04-13 22:00:09 +00:00
af = AF_UNSPEC;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
i = fxdr_unsigned(int, *tl);
if (i >= 3 && i <= 4) {
Add support for INET6 addresses to the kernel code that dumps open/lock state. PR#223036 reported that INET6 callback addresses were not printed by nfsdumpstate(8). This kernel patch adds INET6 addresses to the dump structure, so that nfsdumpstate(8) can print them out, post-r346190. The patch also includes the addition of #ifdef INET, INET6 as requested by bz@. PR: 223036 Reviewed by: bz, rgrimes MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D19839
2019-04-13 22:00:09 +00:00
error = nfsrv_mtostr(nd, addr, i);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (error)
goto nfsmout;
Add support for INET6 addresses to the kernel code that dumps open/lock state. PR#223036 reported that INET6 callback addresses were not printed by nfsdumpstate(8). This kernel patch adds INET6 addresses to the dump structure, so that nfsdumpstate(8) can print them out, post-r346190. The patch also includes the addition of #ifdef INET, INET6 as requested by bz@. PR: 223036 Reviewed by: bz, rgrimes MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D19839
2019-04-13 22:00:09 +00:00
#ifdef INET
if (!strcmp(addr, "tcp")) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
clp->lc_flags |= LCL_TCPCALLBACK;
clp->lc_req.nr_sotype = SOCK_STREAM;
clp->lc_req.nr_soproto = IPPROTO_TCP;
Add support for INET6 addresses to the kernel code that dumps open/lock state. PR#223036 reported that INET6 callback addresses were not printed by nfsdumpstate(8). This kernel patch adds INET6 addresses to the dump structure, so that nfsdumpstate(8) can print them out, post-r346190. The patch also includes the addition of #ifdef INET, INET6 as requested by bz@. PR: 223036 Reviewed by: bz, rgrimes MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D19839
2019-04-13 22:00:09 +00:00
af = AF_INET;
} else if (!strcmp(addr, "udp")) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
clp->lc_req.nr_sotype = SOCK_DGRAM;
clp->lc_req.nr_soproto = IPPROTO_UDP;
Add support for INET6 addresses to the kernel code that dumps open/lock state. PR#223036 reported that INET6 callback addresses were not printed by nfsdumpstate(8). This kernel patch adds INET6 addresses to the dump structure, so that nfsdumpstate(8) can print them out, post-r346190. The patch also includes the addition of #ifdef INET, INET6 as requested by bz@. PR: 223036 Reviewed by: bz, rgrimes MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D19839
2019-04-13 22:00:09 +00:00
af = AF_INET;
}
#endif
#ifdef INET6
if (af == AF_UNSPEC) {
if (!strcmp(addr, "tcp6")) {
clp->lc_flags |= LCL_TCPCALLBACK;
clp->lc_req.nr_sotype = SOCK_STREAM;
clp->lc_req.nr_soproto = IPPROTO_TCP;
af = AF_INET6;
} else if (!strcmp(addr, "udp6")) {
clp->lc_req.nr_sotype = SOCK_DGRAM;
clp->lc_req.nr_soproto = IPPROTO_UDP;
af = AF_INET6;
}
}
#endif
if (af == AF_UNSPEC) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
cantparse = 1;
}
} else {
cantparse = 1;
if (i > 0) {
error = nfsm_advance(nd, NFSM_RNDUP(i), -1);
if (error)
goto nfsmout;
}
}
Add support for INET6 addresses to the kernel code that dumps open/lock state. PR#223036 reported that INET6 callback addresses were not printed by nfsdumpstate(8). This kernel patch adds INET6 addresses to the dump structure, so that nfsdumpstate(8) can print them out, post-r346190. The patch also includes the addition of #ifdef INET, INET6 as requested by bz@. PR: 223036 Reviewed by: bz, rgrimes MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D19839
2019-04-13 22:00:09 +00:00
/*
* The caller has allocated clp->lc_req.nr_nam to be large enough
* for either AF_INET or AF_INET6 and zeroed out the contents.
* maxalen is set to the maximum length of the host IP address string
* plus 8 for the maximum length of the port#.
* minalen is set to the minimum length of the host IP address string
* plus 4 for the minimum length of the port#.
* These lengths do not include NULL termination,
* so INET[6]_ADDRSTRLEN - 1 is used in the calculations.
*/
switch (af) {
#ifdef INET
case AF_INET:
rin = (struct sockaddr_in *)clp->lc_req.nr_nam;
rin->sin_family = AF_INET;
rin->sin_len = sizeof(struct sockaddr_in);
maxalen = INET_ADDRSTRLEN - 1 + 8;
minalen = 7 + 4;
break;
#endif
#ifdef INET6
case AF_INET6:
rin6 = (struct sockaddr_in6 *)clp->lc_req.nr_nam;
rin6->sin6_family = AF_INET6;
rin6->sin6_len = sizeof(struct sockaddr_in6);
maxalen = INET6_ADDRSTRLEN - 1 + 8;
minalen = 3 + 4;
break;
#endif
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
i = fxdr_unsigned(int, *tl);
if (i < 0) {
error = NFSERR_BADXDR;
goto nfsmout;
} else if (i == 0) {
cantparse = 1;
Add support for INET6 addresses to the kernel code that dumps open/lock state. PR#223036 reported that INET6 callback addresses were not printed by nfsdumpstate(8). This kernel patch adds INET6 addresses to the dump structure, so that nfsdumpstate(8) can print them out, post-r346190. The patch also includes the addition of #ifdef INET, INET6 as requested by bz@. PR: 223036 Reviewed by: bz, rgrimes MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D19839
2019-04-13 22:00:09 +00:00
} else if (!cantparse && i <= maxalen && i >= minalen) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
error = nfsrv_mtostr(nd, addr, i);
if (error)
goto nfsmout;
/*
* Parse out the address fields. We expect 6 decimal numbers
Add support for INET6 addresses to the kernel code that dumps open/lock state. PR#223036 reported that INET6 callback addresses were not printed by nfsdumpstate(8). This kernel patch adds INET6 addresses to the dump structure, so that nfsdumpstate(8) can print them out, post-r346190. The patch also includes the addition of #ifdef INET, INET6 as requested by bz@. PR: 223036 Reviewed by: bz, rgrimes MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D19839
2019-04-13 22:00:09 +00:00
* separated by '.'s for AF_INET and two decimal numbers
* preceeded by '.'s for AF_INET6.
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
*/
Add support for INET6 addresses to the kernel code that dumps open/lock state. PR#223036 reported that INET6 callback addresses were not printed by nfsdumpstate(8). This kernel patch adds INET6 addresses to the dump structure, so that nfsdumpstate(8) can print them out, post-r346190. The patch also includes the addition of #ifdef INET, INET6 as requested by bz@. PR: 223036 Reviewed by: bz, rgrimes MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D19839
2019-04-13 22:00:09 +00:00
cp = NULL;
switch (af) {
#ifdef INET6
/*
* For AF_INET6, first parse the host address.
*/
case AF_INET6:
cp = strchr(addr, '.');
if (cp != NULL) {
*cp++ = '\0';
if (inet_pton(af, addr, &rin6->sin6_addr) == 1)
i = 4;
else {
cp = NULL;
cantparse = 1;
}
}
break;
#endif
#ifdef INET
case AF_INET:
cp = addr;
i = 0;
break;
#endif
}
while (cp != NULL && *cp && i < 6) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
cp2 = cp;
while (*cp2 && *cp2 != '.')
cp2++;
if (*cp2)
*cp2++ = '\0';
else if (i != 5) {
cantparse = 1;
break;
}
j = nfsrv_getipnumber(cp);
if (j >= 0) {
if (i < 4)
ip.cval[3 - i] = j;
else
port.cval[5 - i] = j;
} else {
cantparse = 1;
break;
}
cp = cp2;
i++;
}
if (!cantparse) {
Add support for INET6 addresses to the kernel code that dumps open/lock state. PR#223036 reported that INET6 callback addresses were not printed by nfsdumpstate(8). This kernel patch adds INET6 addresses to the dump structure, so that nfsdumpstate(8) can print them out, post-r346190. The patch also includes the addition of #ifdef INET, INET6 as requested by bz@. PR: 223036 Reviewed by: bz, rgrimes MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D19839
2019-04-13 22:00:09 +00:00
/*
* The host address INADDR_ANY is (mis)used to indicate
* "there is no valid callback address".
*/
switch (af) {
#ifdef INET6
case AF_INET6:
if (!IN6_ARE_ADDR_EQUAL(&rin6->sin6_addr,
&in6addr_any))
rin6->sin6_port = htons(port.sval);
else
cantparse = 1;
break;
#endif
#ifdef INET
case AF_INET:
if (ip.ival != INADDR_ANY) {
rin->sin_addr.s_addr = htonl(ip.ival);
rin->sin_port = htons(port.sval);
} else {
cantparse = 1;
}
break;
#endif
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
}
} else {
cantparse = 1;
if (i > 0) {
error = nfsm_advance(nd, NFSM_RNDUP(i), -1);
if (error)
goto nfsmout;
}
}
if (cantparse) {
Add support for INET6 addresses to the kernel code that dumps open/lock state. PR#223036 reported that INET6 callback addresses were not printed by nfsdumpstate(8). This kernel patch adds INET6 addresses to the dump structure, so that nfsdumpstate(8) can print them out, post-r346190. The patch also includes the addition of #ifdef INET, INET6 as requested by bz@. PR: 223036 Reviewed by: bz, rgrimes MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D19839
2019-04-13 22:00:09 +00:00
switch (nd->nd_nam->sa_family) {
#ifdef INET
case AF_INET:
sin = (struct sockaddr_in *)nd->nd_nam;
rin = (struct sockaddr_in *)clp->lc_req.nr_nam;
rin->sin_family = AF_INET;
rin->sin_len = sizeof(struct sockaddr_in);
rin->sin_addr.s_addr = sin->sin_addr.s_addr;
rin->sin_port = 0x0;
break;
#endif
#ifdef INET6
case AF_INET6:
sin6 = (struct sockaddr_in6 *)nd->nd_nam;
rin6 = (struct sockaddr_in6 *)clp->lc_req.nr_nam;
rin6->sin6_family = AF_INET6;
rin6->sin6_len = sizeof(struct sockaddr_in6);
rin6->sin6_addr = sin6->sin6_addr;
rin6->sin6_port = 0x0;
break;
#endif
Fix the client IP address reported by nfsdumpstate for 64bit arch and NFSv4.1. The client IP address was not being reported for some NFSv4 mounts by nfsdumpstate. Upon investigation, two problems were found for mounts using IPv4. One was that the code (originally written and tested on i386) assumed that a "u_long" was a "uint32_t" and would exactly store an IPv4 host address. Not correct for 64bit arches. Also, for NFSv4.1 mounts, the field was not being filled in. This was basically correct, because NFSv4.1 does not use a callback address. However, it meant that nfsdumpstate could not report the client IP addr. This patch should fix both of these issues. For IPv6, the address will still not be reported. The original NFSv4 RFC only specified IPv4 callback addresses. I think this has changed and, if so, a future commit to fix reporting of IPv6 addresses will be needed. Reported by: manu PR: 223036 MFC after: 2 weeks
2017-10-15 22:22:27 +00:00
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
clp->lc_program = 0;
}
nfsmout:
Add support for INET6 addresses to the kernel code that dumps open/lock state. PR#223036 reported that INET6 callback addresses were not printed by nfsdumpstate(8). This kernel patch adds INET6 addresses to the dump structure, so that nfsdumpstate(8) can print them out, post-r346190. The patch also includes the addition of #ifdef INET, INET6 as requested by bz@. PR: 223036 Reviewed by: bz, rgrimes MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D19839
2019-04-13 22:00:09 +00:00
free(addr, M_TEMP);
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
NFSEXITCODE2(error, nd);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
return (error);
}
/*
* Turn a string of up to three decimal digits into a number. Return -1 upon
* error.
*/
static int
nfsrv_getipnumber(u_char *cp)
{
int i = 0, j = 0;
while (*cp) {
if (j > 2 || *cp < '0' || *cp > '9')
return (-1);
i *= 10;
i += (*cp - '0');
cp++;
j++;
}
if (i < 256)
return (i);
return (-1);
}
/*
* This function checks for restart conditions.
*/
static int
nfsrv_checkrestart(nfsquad_t clientid, u_int32_t flags,
nfsv4stateid_t *stateidp, int specialid)
{
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
int ret = 0;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* First check for a server restart. Open, LockT, ReleaseLockOwner
* and DelegPurge have a clientid, the rest a stateid.
*/
if (flags &
(NFSLCK_OPEN | NFSLCK_TEST | NFSLCK_RELEASE | NFSLCK_DELEGPURGE)) {
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
if (clientid.lval[0] != NFSD_VNET(nfsrvboottime)) {
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
ret = NFSERR_STALECLIENTID;
goto out;
}
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
} else if (stateidp->other[0] != NFSD_VNET(nfsrvboottime) &&
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
specialid == 0) {
ret = NFSERR_STALESTATEID;
goto out;
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* Read, Write, Setattr and LockT can return NFSERR_GRACE and do
* not use a lock/open owner seqid#, so the check can be done now.
* (The others will be checked, as required, later.)
*/
if (!(flags & (NFSLCK_CHECK | NFSLCK_TEST)))
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
NFSLOCKSTATE();
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
ret = nfsrv_checkgrace(NULL, NULL, flags);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
NFSUNLOCKSTATE();
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
out:
NFSEXITCODE(ret);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
return (ret);
}
/*
* Check for grace.
*/
static int
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
nfsrv_checkgrace(struct nfsrv_descript *nd, struct nfsclient *clp,
u_int32_t flags)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
{
End grace for the NFSv4 server if all mounts do ReclaimComplete. The NFSv4 protocol requires that the server only allow reclaim of state and not issue any new open/lock state for a grace period after booting. The NFSv4.0 protocol required this grace period to be greater than the lease duration (over 2minutes). For NFSv4.1, the client tells the server that it has done reclaiming state by doing a ReclaimComplete operation. If all NFSv4 clients are NFSv4.1, the grace period can end once all the clients have done ReclaimComplete, shortening the time period considerably. This patch does this. If there are any NFSv4.0 mounts, the grace period will still be over 2minutes. This change is only an optimization and does not affect correct operation. Tested by: andreas.nagy@frequentis.com MFC after: 2 months
2018-05-15 20:28:50 +00:00
int error = 0, notreclaimed;
struct nfsrv_stable *sp;
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
if ((NFSD_VNET(nfsrv_stablefirst).nsf_flags & (NFSNSF_UPDATEDONE |
End grace for the NFSv4 server if all mounts do ReclaimComplete. The NFSv4 protocol requires that the server only allow reclaim of state and not issue any new open/lock state for a grace period after booting. The NFSv4.0 protocol required this grace period to be greater than the lease duration (over 2minutes). For NFSv4.1, the client tells the server that it has done reclaiming state by doing a ReclaimComplete operation. If all NFSv4 clients are NFSv4.1, the grace period can end once all the clients have done ReclaimComplete, shortening the time period considerably. This patch does this. If there are any NFSv4.0 mounts, the grace period will still be over 2minutes. This change is only an optimization and does not affect correct operation. Tested by: andreas.nagy@frequentis.com MFC after: 2 months
2018-05-15 20:28:50 +00:00
NFSNSF_GRACEOVER)) == 0) {
/*
* First, check to see if all of the clients have done a
* ReclaimComplete. If so, grace can end now.
*/
notreclaimed = 0;
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
LIST_FOREACH(sp, &NFSD_VNET(nfsrv_stablefirst).nsf_head,
nst_list) {
End grace for the NFSv4 server if all mounts do ReclaimComplete. The NFSv4 protocol requires that the server only allow reclaim of state and not issue any new open/lock state for a grace period after booting. The NFSv4.0 protocol required this grace period to be greater than the lease duration (over 2minutes). For NFSv4.1, the client tells the server that it has done reclaiming state by doing a ReclaimComplete operation. If all NFSv4 clients are NFSv4.1, the grace period can end once all the clients have done ReclaimComplete, shortening the time period considerably. This patch does this. If there are any NFSv4.0 mounts, the grace period will still be over 2minutes. This change is only an optimization and does not affect correct operation. Tested by: andreas.nagy@frequentis.com MFC after: 2 months
2018-05-15 20:28:50 +00:00
if ((sp->nst_flag & NFSNST_RECLAIMED) == 0) {
notreclaimed = 1;
break;
}
}
if (notreclaimed == 0)
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
NFSD_VNET(nfsrv_stablefirst).nsf_flags |=
(NFSNSF_GRACEOVER | NFSNSF_NEEDLOCK);
End grace for the NFSv4 server if all mounts do ReclaimComplete. The NFSv4 protocol requires that the server only allow reclaim of state and not issue any new open/lock state for a grace period after booting. The NFSv4.0 protocol required this grace period to be greater than the lease duration (over 2minutes). For NFSv4.1, the client tells the server that it has done reclaiming state by doing a ReclaimComplete operation. If all NFSv4 clients are NFSv4.1, the grace period can end once all the clients have done ReclaimComplete, shortening the time period considerably. This patch does this. If there are any NFSv4.0 mounts, the grace period will still be over 2minutes. This change is only an optimization and does not affect correct operation. Tested by: andreas.nagy@frequentis.com MFC after: 2 months
2018-05-15 20:28:50 +00:00
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
if ((NFSD_VNET(nfsrv_stablefirst).nsf_flags & NFSNSF_GRACEOVER) != 0) {
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
if (flags & NFSLCK_RECLAIM) {
error = NFSERR_NOGRACE;
goto out;
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
} else {
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
if (!(flags & NFSLCK_RECLAIM)) {
error = NFSERR_GRACE;
goto out;
}
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
if (nd != NULL && clp != NULL &&
(nd->nd_flag & ND_NFSV41) != 0 &&
(clp->lc_flags & LCL_RECLAIMCOMPLETE) != 0) {
error = NFSERR_NOGRACE;
goto out;
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* If grace is almost over and we are still getting Reclaims,
* extend grace a bit.
*/
if ((NFSD_MONOSEC + NFSRV_LEASEDELTA) >
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
NFSD_VNET(nfsrv_stablefirst).nsf_eograce)
NFSD_VNET(nfsrv_stablefirst).nsf_eograce =
NFSD_MONOSEC + NFSRV_LEASEDELTA;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
out:
NFSEXITCODE(error);
return (error);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* Do a server callback.
Assorted fixes to handling of LayoutRecall callbacks, mostly error handling. After a re-read of the appropriate section of RFC5661, I decided that a few things should be changed related to LayoutRecall callback handling. Here are the things fixed by this patch. - For two of the three cases that LayoutRecall is done, I now think setting the clora_changed argument false is correct. - All errors other than NFSERR_DELAY returned by LayoutRecall appear permanent, so don't retry for any of them. (NFSERR_DELAY is retried by newnfs_request(), so it is not affected by this patch.) - Instead of waiting "forever" (actually until the process is SIGTERM'd) for Layouts to be returned during a mirror copy, fail and return ENXIO after about 1minute. Waiting for a <ctrl>C made sense when pnfsdscopymr() was done by itself, but did not make sense when done via find(1). This patch only affects the pNFS server.
2018-08-08 20:21:45 +00:00
* The "trunc" argument is slightly overloaded and refers to different
* boolean arguments for CBRECALL and CBLAYOUTRECALL.
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
*/
static int
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
nfsrv_docallback(struct nfsclient *clp, int procnum, nfsv4stateid_t *stateidp,
int trunc, fhandle_t *fhp, struct nfsvattr *nap, nfsattrbit_t *attrbitp,
int laytype, NFSPROC_T *p)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
{
Replace all instances of the typedef mbuf_t with "struct mbuf *". The typedef mbuf_t was used for the Mac OS/X port of the code long ago. Since this port is no longer used and the use of mbuf_t obscures what the code does (and is not consistent with style(9)), it is no longer needed. This patch replaces all instances of mbuf_t with "struct mbuf *", so that it is no longer used. This patch should not result in any semantic change.
2020-04-17 21:17:51 +00:00
struct mbuf *m;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
u_int32_t *tl;
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
struct nfsrv_descript *nd;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
struct ucred *cred;
nfsd: fix session slot handling for failed callbacks When the NFSv4.1/4.2 server does a callback to a client on the back channel, it will use a session slot in the back channel session. If the back channel has failed, the callback will fail and, without this patch, the session slot will not be released. As more callbacks are attempted, all session slots can become busy and then the nfsd thread gets stuck waiting for a back channel session slot. This patch frees the session slot upon callback failure to avoid this problem. Without this patch, the problem can be avoided by leaving delegations disabled in the NFS server. MFC after: 2 weeks
2021-04-23 22:24:47 +00:00
int error = 0, slotpos;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
u_int32_t callback;
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
struct nfsdsession *sep = NULL;
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
uint64_t tval;
Add an argument to newnfs_connect() that indicates use TLS for the connection. For NFSv4.0, the server creates a server->client TCP connection for callbacks. If the client mount on the server is using TLS, enable TLS for this callback TCP connection. TLS connections from clients will not be supported until the kernel RPC changes are committed. Since this changes the internal ABI between the NFS kernel modules that will require a version bump, delete newnfs_trimtrailing(), which is no longer used. Since LCL_TLSCB is not yet set, these changes should not have any semantic affect at this time.
2020-08-11 00:26:45 +00:00
bool dotls;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
nd = malloc(sizeof(*nd), M_TEMP, M_WAITOK | M_ZERO);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
cred = newnfs_getcred();
NFSLOCKSTATE(); /* mostly for lc_cbref++ */
if (clp->lc_flags & LCL_NEEDSCONFIRM) {
NFSUNLOCKSTATE();
panic("docallb");
}
clp->lc_cbref++;
Fixed the Null callback RPCs so that they work with the new krpc. This required two changes: setting the program and version numbers before connect and fixing the handling of the Null Rpc case in newnfs_request(). Approved by: kib (mentor)
2009-05-16 03:12:55 +00:00
/*
* Fill the callback program# and version into the request
* structure for newnfs_connect() to use.
*/
clp->lc_req.nr_prog = clp->lc_program;
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
#ifdef notnow
if ((clp->lc_flags & LCL_NFSV41) != 0)
clp->lc_req.nr_vers = NFSV41_CBVERS;
else
#endif
clp->lc_req.nr_vers = NFSV4_CBVERS;
Fixed the Null callback RPCs so that they work with the new krpc. This required two changes: setting the program and version numbers before connect and fixing the handling of the Null Rpc case in newnfs_request(). Approved by: kib (mentor)
2009-05-16 03:12:55 +00:00
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* First, fill in some of the fields of nd and cr.
*/
nd->nd_flag = ND_NFSV4;
if (clp->lc_flags & LCL_GSS)
nd->nd_flag |= ND_KERBV;
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
if ((clp->lc_flags & LCL_NFSV41) != 0)
nd->nd_flag |= ND_NFSV41;
Add support for NFSv4.2 to the NFS client and server. This patch adds support for NFSv4.2 (RFC-7862) and Extended Attributes (RFC-8276) to the NFS client and server. NFSv4.2 is comprised of several optional features that can be supported in addition to NFSv4.1. This patch adds the following optional features: - posix_fadvise(POSIX_FADV_WILLNEED/POSIX_FADV_DONTNEED) - posix_fallocate() - intra server file range copying via the copy_file_range(2) syscall --> Avoiding data tranfer over the wire to/from the NFS client. - lseek(SEEK_DATA/SEEK_HOLE) - Extended attribute syscalls for "user" namespace attributes as defined by RFC-8276. Although this patch is fairly large, it should not affect support for the other versions of NFS. However it does add two new sysctls that allow a sysadmin to limit which minor versions of NFSv4 a server supports, allowing a sysadmin to disable NFSv4.2. Unfortunately, when the NFS stats structure was last revised, it was assumed that there would be no additional operations added beyond what was specified in RFC-7862. However RFC-8276 did add additional operations, forcing the NFS stats structure to revised again. It now has extra unused entries in all arrays, so that future extensions to NFSv4.2 can be accomodated without revising this structure again. A future commit will update nfsstat(1) to report counts for the new NFSv4.2 specific operations/procedures. This patch affects the internal interface between the nfscommon, nfscl and nfsd modules and, as such, they all must be upgraded simultaneously. I will do a version bump (although arguably not needed), due to this. This code has survived a "make universe" but has not been built with a recent GCC. If you encounter build problems, please email me. Relnotes: yes
2019-12-12 23:22:55 +00:00
if ((clp->lc_flags & LCL_NFSV42) != 0)
nd->nd_flag |= ND_NFSV42;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nd->nd_repstat = 0;
cred->cr_uid = clp->lc_uid;
cred->cr_gid = clp->lc_gid;
callback = clp->lc_callback;
NFSUNLOCKSTATE();
cred->cr_ngroups = 1;
/*
* Get the first mbuf for the request.
*/
Mechanically substitute flags from historic mbuf allocator with malloc(9) flags within sys. Exceptions: - sys/contrib not touched - sys/mbuf.h edited manually
2012-12-05 08:04:20 +00:00
MGET(m, M_WAITOK, MT_DATA);
Replace mbuf macros with the code they would generate in the NFS code. When the code was ported to Mac OS/X, mbuf handling functions were converted to using the Mac OS/X accessor functions. For FreeBSD, they are a simple set of macros in sys/fs/nfs/nfskpiport.h. Since porting to Mac OS/X is no longer a consideration, replacement of these macros with the code generated by them makes the code more readable. When support for external page mbufs is added as needed by the KERN_TLS, the patch becomes simpler if done without the macros. This patch should not result in any semantic change.
2020-04-11 20:57:15 +00:00
m->m_len = 0;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nd->nd_mreq = nd->nd_mb = m;
Replace mbuf macros with the code they would generate in the NFS code. When the code was ported to Mac OS/X, mbuf handling functions were converted to using the Mac OS/X accessor functions. For FreeBSD, they are a simple set of macros in sys/fs/nfs/nfskpiport.h. Since porting to Mac OS/X is no longer a consideration, replacement of these macros with the code generated by them makes the code more readable. When support for external page mbufs is added as needed by the KERN_TLS, the patch becomes simpler if done without the macros. This patch should not result in any semantic change.
2020-04-11 20:57:15 +00:00
nd->nd_bpos = mtod(m, caddr_t);
fs: clean up empty lines in .c and .h files
2020-09-01 21:18:40 +00:00
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* and build the callback request.
*/
if (procnum == NFSV4OP_CBGETATTR) {
nd->nd_procnum = NFSV4PROC_CBCOMPOUND;
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
error = nfsrv_cbcallargs(nd, clp, callback, NFSV4OP_CBGETATTR,
nfsd: fix session slot handling for failed callbacks When the NFSv4.1/4.2 server does a callback to a client on the back channel, it will use a session slot in the back channel session. If the back channel has failed, the callback will fail and, without this patch, the session slot will not be released. As more callbacks are attempted, all session slots can become busy and then the nfsd thread gets stuck waiting for a back channel session slot. This patch frees the session slot upon callback failure to avoid this problem. Without this patch, the problem can be avoided by leaving delegations disabled in the NFS server. MFC after: 2 weeks
2021-04-23 22:24:47 +00:00
"CB Getattr", &sep, &slotpos);
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
if (error != 0) {
Replace mbuf macros with the code they would generate in the NFS code. When the code was ported to Mac OS/X, mbuf handling functions were converted to using the Mac OS/X accessor functions. For FreeBSD, they are a simple set of macros in sys/fs/nfs/nfskpiport.h. Since porting to Mac OS/X is no longer a consideration, replacement of these macros with the code generated by them makes the code more readable. When support for external page mbufs is added as needed by the KERN_TLS, the patch becomes simpler if done without the macros. This patch should not result in any semantic change.
2020-04-11 20:57:15 +00:00
m_freem(nd->nd_mreq);
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
goto errout;
}
nfscl: Make coverity happy Coverity does not like code that checks a function's return value sometimes. Add "(void)" in front of the function when the return value does not matter to try and make it happy. A recent commit deleted "(void)"s in front of nfsm_fhtom(). This commit puts them back in. Reported by: emaste MFC after: 3 months
2023-03-29 00:08:45 +00:00
(void)nfsm_fhtom(NULL, nd, (u_int8_t *)fhp, NFSX_MYFH, 0);
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
(void)nfsrv_putattrbit(nd, attrbitp);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
} else if (procnum == NFSV4OP_CBRECALL) {
nd->nd_procnum = NFSV4PROC_CBCOMPOUND;
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
error = nfsrv_cbcallargs(nd, clp, callback, NFSV4OP_CBRECALL,
nfsd: fix session slot handling for failed callbacks When the NFSv4.1/4.2 server does a callback to a client on the back channel, it will use a session slot in the back channel session. If the back channel has failed, the callback will fail and, without this patch, the session slot will not be released. As more callbacks are attempted, all session slots can become busy and then the nfsd thread gets stuck waiting for a back channel session slot. This patch frees the session slot upon callback failure to avoid this problem. Without this patch, the problem can be avoided by leaving delegations disabled in the NFS server. MFC after: 2 weeks
2021-04-23 22:24:47 +00:00
"CB Recall", &sep, &slotpos);
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
if (error != 0) {
Replace mbuf macros with the code they would generate in the NFS code. When the code was ported to Mac OS/X, mbuf handling functions were converted to using the Mac OS/X accessor functions. For FreeBSD, they are a simple set of macros in sys/fs/nfs/nfskpiport.h. Since porting to Mac OS/X is no longer a consideration, replacement of these macros with the code generated by them makes the code more readable. When support for external page mbufs is added as needed by the KERN_TLS, the patch becomes simpler if done without the macros. This patch should not result in any semantic change.
2020-04-11 20:57:15 +00:00
m_freem(nd->nd_mreq);
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
goto errout;
}
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
*tl++ = txdr_unsigned(stateidp->seqid);
NFSBCOPY((caddr_t)stateidp->other, (caddr_t)tl,
NFSX_STATEIDOTHER);
tl += (NFSX_STATEIDOTHER / NFSX_UNSIGNED);
if (trunc)
*tl = newnfs_true;
else
*tl = newnfs_false;
nfscl: Make coverity happy Coverity does not like code that checks a function's return value sometimes. Add "(void)" in front of the function when the return value does not matter to try and make it happy. A recent commit deleted "(void)"s in front of nfsm_fhtom(). This commit puts them back in. Reported by: emaste MFC after: 3 months
2023-03-29 00:08:45 +00:00
(void)nfsm_fhtom(NULL, nd, (u_int8_t *)fhp, NFSX_MYFH, 0);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
} else if (procnum == NFSV4OP_CBLAYOUTRECALL) {
NFSD_DEBUG(4, "docallback layout recall\n");
nd->nd_procnum = NFSV4PROC_CBCOMPOUND;
error = nfsrv_cbcallargs(nd, clp, callback,
nfsd: fix session slot handling for failed callbacks When the NFSv4.1/4.2 server does a callback to a client on the back channel, it will use a session slot in the back channel session. If the back channel has failed, the callback will fail and, without this patch, the session slot will not be released. As more callbacks are attempted, all session slots can become busy and then the nfsd thread gets stuck waiting for a back channel session slot. This patch frees the session slot upon callback failure to avoid this problem. Without this patch, the problem can be avoided by leaving delegations disabled in the NFS server. MFC after: 2 weeks
2021-04-23 22:24:47 +00:00
NFSV4OP_CBLAYOUTRECALL, "CB Reclayout", &sep, &slotpos);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
NFSD_DEBUG(4, "aft cbcallargs=%d\n", error);
if (error != 0) {
Replace mbuf macros with the code they would generate in the NFS code. When the code was ported to Mac OS/X, mbuf handling functions were converted to using the Mac OS/X accessor functions. For FreeBSD, they are a simple set of macros in sys/fs/nfs/nfskpiport.h. Since porting to Mac OS/X is no longer a consideration, replacement of these macros with the code generated by them makes the code more readable. When support for external page mbufs is added as needed by the KERN_TLS, the patch becomes simpler if done without the macros. This patch should not result in any semantic change.
2020-04-11 20:57:15 +00:00
m_freem(nd->nd_mreq);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
goto errout;
}
NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
*tl++ = txdr_unsigned(laytype);
*tl++ = txdr_unsigned(NFSLAYOUTIOMODE_ANY);
Assorted fixes to handling of LayoutRecall callbacks, mostly error handling. After a re-read of the appropriate section of RFC5661, I decided that a few things should be changed related to LayoutRecall callback handling. Here are the things fixed by this patch. - For two of the three cases that LayoutRecall is done, I now think setting the clora_changed argument false is correct. - All errors other than NFSERR_DELAY returned by LayoutRecall appear permanent, so don't retry for any of them. (NFSERR_DELAY is retried by newnfs_request(), so it is not affected by this patch.) - Instead of waiting "forever" (actually until the process is SIGTERM'd) for Layouts to be returned during a mirror copy, fail and return ENXIO after about 1minute. Waiting for a <ctrl>C made sense when pnfsdscopymr() was done by itself, but did not make sense when done via find(1). This patch only affects the pNFS server.
2018-08-08 20:21:45 +00:00
if (trunc)
*tl++ = newnfs_true;
else
*tl++ = newnfs_false;
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
*tl = txdr_unsigned(NFSV4LAYOUTRET_FILE);
nfscl: Make coverity happy Coverity does not like code that checks a function's return value sometimes. Add "(void)" in front of the function when the return value does not matter to try and make it happy. A recent commit deleted "(void)"s in front of nfsm_fhtom(). This commit puts them back in. Reported by: emaste MFC after: 3 months
2023-03-29 00:08:45 +00:00
(void)nfsm_fhtom(NULL, nd, (uint8_t *)fhp, NFSX_MYFH, 0);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_HYPER + NFSX_STATEID);
tval = 0;
txdr_hyper(tval, tl); tl += 2;
tval = UINT64_MAX;
txdr_hyper(tval, tl); tl += 2;
*tl++ = txdr_unsigned(stateidp->seqid);
NFSBCOPY(stateidp->other, tl, NFSX_STATEIDOTHER);
tl += (NFSX_STATEIDOTHER / NFSX_UNSIGNED);
NFSD_DEBUG(4, "aft args\n");
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
} else if (procnum == NFSV4PROC_CBNULL) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nd->nd_procnum = NFSV4PROC_CBNULL;
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
if ((clp->lc_flags & LCL_NFSV41) != 0) {
error = nfsv4_getcbsession(clp, &sep);
if (error != 0) {
Replace mbuf macros with the code they would generate in the NFS code. When the code was ported to Mac OS/X, mbuf handling functions were converted to using the Mac OS/X accessor functions. For FreeBSD, they are a simple set of macros in sys/fs/nfs/nfskpiport.h. Since porting to Mac OS/X is no longer a consideration, replacement of these macros with the code generated by them makes the code more readable. When support for external page mbufs is added as needed by the KERN_TLS, the patch becomes simpler if done without the macros. This patch should not result in any semantic change.
2020-04-11 20:57:15 +00:00
m_freem(nd->nd_mreq);
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
goto errout;
}
}
} else {
error = NFSERR_SERVERFAULT;
Replace mbuf macros with the code they would generate in the NFS code. When the code was ported to Mac OS/X, mbuf handling functions were converted to using the Mac OS/X accessor functions. For FreeBSD, they are a simple set of macros in sys/fs/nfs/nfskpiport.h. Since porting to Mac OS/X is no longer a consideration, replacement of these macros with the code generated by them makes the code more readable. When support for external page mbufs is added as needed by the KERN_TLS, the patch becomes simpler if done without the macros. This patch should not result in any semantic change.
2020-04-11 20:57:15 +00:00
m_freem(nd->nd_mreq);
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
goto errout;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* Call newnfs_connect(), as required, and then newnfs_request().
*/
Add an argument to newnfs_connect() that indicates use TLS for the connection. For NFSv4.0, the server creates a server->client TCP connection for callbacks. If the client mount on the server is using TLS, enable TLS for this callback TCP connection. TLS connections from clients will not be supported until the kernel RPC changes are committed. Since this changes the internal ABI between the NFS kernel modules that will require a version bump, delete newnfs_trimtrailing(), which is no longer used. Since LCL_TLSCB is not yet set, these changes should not have any semantic affect at this time.
2020-08-11 00:26:45 +00:00
dotls = false;
if ((clp->lc_flags & LCL_TLSCB) != 0)
dotls = true;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
(void) newnfs_sndlock(&clp->lc_req.nr_lock);
if (clp->lc_req.nr_client == NULL) {
Add a missing nfsrv_freesession() call for an unlikely failure case. Since NFSv4.1 clients normally create a single session which supports both fore and back channels, it is unlikely that a callback will fail due to a lack of a back channel. However, if this failure occurred, the session wasn't being dereferenced and would never be free'd. Found by inspection during pNFS server development. Tested by: andreas.nagy@frequentis.com MFC after: 2 months
2018-05-17 21:17:20 +00:00
if ((clp->lc_flags & LCL_NFSV41) != 0) {
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
error = ECONNREFUSED;
nfsd: fix session slot handling for failed callbacks When the NFSv4.1/4.2 server does a callback to a client on the back channel, it will use a session slot in the back channel session. If the back channel has failed, the callback will fail and, without this patch, the session slot will not be released. As more callbacks are attempted, all session slots can become busy and then the nfsd thread gets stuck waiting for a back channel session slot. This patch frees the session slot upon callback failure to avoid this problem. Without this patch, the problem can be avoided by leaving delegations disabled in the NFS server. MFC after: 2 weeks
2021-04-23 22:24:47 +00:00
if (procnum != NFSV4PROC_CBNULL)
nfsd: fix the slot sequence# when a callback fails Commit 4281bfec3628 patched the server so that the callback session slot would be free'd for reuse when a callback attempt fails. However, this can often result in the sequence# for the session slot to be advanced such that the client end will reply NFSERR_SEQMISORDERED. To avoid the NFSERR_SEQMISORDERED client reply, this patch negates the sequence# advance for the case where the callback has failed. The common case is a failed back channel, where the callback cannot be sent to the client, and not advancing the sequence# is correct for this case. For the uncommon case where the client's reply to the callback is lost, not advancing the sequence# will indicate to the client that the next callback is a retry and not a new callback. But, since the FreeBSD server always sets "csa_cachethis" false in the callback sequence operation, a retry and a new callback should be handled the same way by the client, so this should not matter. Until you have this patch in your NFSv4.1/4.2 server, you should consider avoiding the use of delegations. Even with this patch, interoperation with the Linux NFSv4.1/4.2 client in kernel versions prior to 5.3 can result in frequent 15second delays if delegations are enabled. This occurs because, for kernels prior to 5.3, the Linux client does a TCP reconnect every time it sees multiple concurrent callbacks and then it takes 15seconds to recover the back channel after doing so. MFC after: 2 weeks
2021-04-26 23:24:10 +00:00
nfsv4_freeslot(&sep->sess_cbsess, slotpos,
true);
nfsd: Add support for the SP4_MACH_CRED case in ExchangeID Commit f4179ad46fa4 added support for operation bitmaps for NFSv4.1/4.2. This commit uses those to implement the SP4_MACH_CRED case for the NFSv4.1/4.2 ExchangeID operation since the Linux NFSv4.1/4.2 client is now using this for Kerberized mounts. The Linux Kerberized NFSv4.1/4.2 mounts currently work without support for this because Linux will fall back to SP4_NONE, but there is no guarantee this fallback will work forever. This commit only affects Kerberized NFSv4.1/4.2 mounts from Linux at this time. MFC after: 3 months
2023-04-07 19:49:23 +00:00
nfsrv_freesession(NULL, sep, NULL);
Add a missing nfsrv_freesession() call for an unlikely failure case. Since NFSv4.1 clients normally create a single session which supports both fore and back channels, it is unlikely that a callback will fail due to a lack of a back channel. However, if this failure occurred, the session wasn't being dereferenced and would never be free'd. Found by inspection during pNFS server development. Tested by: andreas.nagy@frequentis.com MFC after: 2 months
2018-05-17 21:17:20 +00:00
} else if (nd->nd_procnum == NFSV4PROC_CBNULL)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
error = newnfs_connect(NULL, &clp->lc_req, cred,
nfscl: Add a Linux compatible "nconnect" mount option Linux has had an "nconnect" NFS mount option for some time. It specifies that N (up to 16) TCP connections are to created for a mount, instead of just one TCP connection. A discussion on freebsd-net@ indicated that this could improve client<-->server network bandwidth, if either the client or server have one of the following: - multiple network ports aggregated to-gether with lagg/lacp. - a fast NIC that is using multiple queues It does result in using more IP port#s and might increase server peak load for a client. One difference from the Linux implementation is that this implementation uses the first TCP connection for all RPCs composed of small messages and uses the additional TCP connections for RPCs that normally have large messages (Read/Readdir/Write). The Linux implementation spreads all RPCs across all TCP connections in a round robin fashion, whereas this implementation spreads Read/Readdir/Write across the additional TCP connections in a round robin fashion. Reviewed by: markj MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D30970
2021-07-09 00:39:04 +00:00
NULL, 1, dotls, &clp->lc_req.nr_client);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
else
error = newnfs_connect(NULL, &clp->lc_req, cred,
nfscl: Add a Linux compatible "nconnect" mount option Linux has had an "nconnect" NFS mount option for some time. It specifies that N (up to 16) TCP connections are to created for a mount, instead of just one TCP connection. A discussion on freebsd-net@ indicated that this could improve client<-->server network bandwidth, if either the client or server have one of the following: - multiple network ports aggregated to-gether with lagg/lacp. - a fast NIC that is using multiple queues It does result in using more IP port#s and might increase server peak load for a client. One difference from the Linux implementation is that this implementation uses the first TCP connection for all RPCs composed of small messages and uses the additional TCP connections for RPCs that normally have large messages (Read/Readdir/Write). The Linux implementation spreads all RPCs across all TCP connections in a round robin fashion, whereas this implementation spreads Read/Readdir/Write across the additional TCP connections in a round robin fashion. Reviewed by: markj MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D30970
2021-07-09 00:39:04 +00:00
NULL, 3, dotls, &clp->lc_req.nr_client);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
newnfs_sndunlock(&clp->lc_req.nr_lock);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
NFSD_DEBUG(4, "aft sndunlock=%d\n", error);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (!error) {
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
if ((nd->nd_flag & ND_NFSV41) != 0) {
KASSERT(sep != NULL, ("sep NULL"));
When the nfsd threads are terminated, the NFSv4 server state (opens, locks, etc) is retained, which I believe is correct behaviour. However, for NFSv4.1, the server also retained a reference to the xprt (RPC transport socket structure) for the backchannel. This caused svcpool_destroy() to not call SVC_DESTROY() for the xprt and allowed a socket upcall to occur after the mutexes in the svcpool were destroyed, causing a crash. This patch fixes the code so that the backchannel xprt structure is dereferenced just before svcpool_destroy() is called, so the code does do an SVC_DESTROY() on the xprt, which shuts down the socket upcall. Tested by: g_amanakis@yahoo.com PR: 204340 MFC after: 2 weeks
2015-11-21 23:55:46 +00:00
if (sep->sess_cbsess.nfsess_xprt != NULL)
error = newnfs_request(nd, NULL, clp,
&clp->lc_req, NULL, NULL, cred,
clp->lc_program, clp->lc_req.nr_vers, NULL,
1, NULL, &sep->sess_cbsess);
else {
/*
* This should probably never occur, but if a
* client somehow does an RPC without a
* SequenceID Op that causes a callback just
* after the nfsd threads have been terminated
* and restared we could conceivably get here
* without a backchannel xprt.
*/
printf("nfsrv_docallback: no xprt\n");
error = ECONNREFUSED;
}
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
NFSD_DEBUG(4, "aft newnfs_request=%d\n", error);
nfsd: fix the slot sequence# when a callback fails Commit 4281bfec3628 patched the server so that the callback session slot would be free'd for reuse when a callback attempt fails. However, this can often result in the sequence# for the session slot to be advanced such that the client end will reply NFSERR_SEQMISORDERED. To avoid the NFSERR_SEQMISORDERED client reply, this patch negates the sequence# advance for the case where the callback has failed. The common case is a failed back channel, where the callback cannot be sent to the client, and not advancing the sequence# is correct for this case. For the uncommon case where the client's reply to the callback is lost, not advancing the sequence# will indicate to the client that the next callback is a retry and not a new callback. But, since the FreeBSD server always sets "csa_cachethis" false in the callback sequence operation, a retry and a new callback should be handled the same way by the client, so this should not matter. Until you have this patch in your NFSv4.1/4.2 server, you should consider avoiding the use of delegations. Even with this patch, interoperation with the Linux NFSv4.1/4.2 client in kernel versions prior to 5.3 can result in frequent 15second delays if delegations are enabled. This occurs because, for kernels prior to 5.3, the Linux client does a TCP reconnect every time it sees multiple concurrent callbacks and then it takes 15seconds to recover the back channel after doing so. MFC after: 2 weeks
2021-04-26 23:24:10 +00:00
if (error != 0 && procnum != NFSV4PROC_CBNULL) {
/*
* It is likely that the callback was never
* processed by the client and, as such,
* the sequence# for the session slot needs
* to be backed up by one to avoid a
* NFSERR_SEQMISORDERED error reply.
* For the unlikely case where the callback
* was processed by the client, this will
* make the next callback on the slot
* appear to be a retry.
* Since callbacks never specify that the
* reply be cached, this "apparent retry"
* should not be a problem.
*/
nfsv4_freeslot(&sep->sess_cbsess, slotpos,
true);
}
nfsd: Add support for the SP4_MACH_CRED case in ExchangeID Commit f4179ad46fa4 added support for operation bitmaps for NFSv4.1/4.2. This commit uses those to implement the SP4_MACH_CRED case for the NFSv4.1/4.2 ExchangeID operation since the Linux NFSv4.1/4.2 client is now using this for Kerberized mounts. The Linux Kerberized NFSv4.1/4.2 mounts currently work without support for this because Linux will fall back to SP4_NONE, but there is no guarantee this fallback will work forever. This commit only affects Kerberized NFSv4.1/4.2 mounts from Linux at this time. MFC after: 3 months
2023-04-07 19:49:23 +00:00
nfsrv_freesession(NULL, sep, NULL);
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
} else
error = newnfs_request(nd, NULL, clp, &clp->lc_req,
NULL, NULL, cred, clp->lc_program,
clp->lc_req.nr_vers, NULL, 1, NULL, NULL);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
errout:
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
NFSFREECRED(cred);
/*
* If error is set here, the Callback path isn't working
* properly, so twiddle the appropriate LCL_ flags.
* (nd_repstat != 0 indicates the Callback path is working,
* but the callback failed on the client.)
*/
if (error) {
/*
* Mark the callback pathway down, which disabled issuing
* of delegations and gets Renew to return NFSERR_CBPATHDOWN.
*/
NFSLOCKSTATE();
clp->lc_flags |= LCL_CBDOWN;
NFSUNLOCKSTATE();
} else {
/*
* Callback worked. If the callback path was down, disable
* callbacks, so no more delegations will be issued. (This
* is done on the assumption that the callback pathway is
* flakey.)
*/
NFSLOCKSTATE();
if (clp->lc_flags & LCL_CBDOWN)
clp->lc_flags &= ~(LCL_CBDOWN | LCL_CALLBACKSON);
NFSUNLOCKSTATE();
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
if (nd->nd_repstat) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
error = nd->nd_repstat;
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
NFSD_DEBUG(1, "nfsrv_docallback op=%d err=%d\n",
procnum, error);
} else if (error == 0 && procnum == NFSV4OP_CBGETATTR)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
NULL, NULL, NULL, NULL, NULL, 0, NULL, NULL, NULL,
p, NULL);
Replace mbuf macros with the code they would generate in the NFS code. When the code was ported to Mac OS/X, mbuf handling functions were converted to using the Mac OS/X accessor functions. For FreeBSD, they are a simple set of macros in sys/fs/nfs/nfskpiport.h. Since porting to Mac OS/X is no longer a consideration, replacement of these macros with the code generated by them makes the code more readable. When support for external page mbufs is added as needed by the KERN_TLS, the patch becomes simpler if done without the macros. This patch should not result in any semantic change.
2020-04-11 20:57:15 +00:00
m_freem(nd->nd_mrep);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
NFSLOCKSTATE();
clp->lc_cbref--;
if ((clp->lc_flags & LCL_WAKEUPWANTED) && clp->lc_cbref == 0) {
clp->lc_flags &= ~LCL_WAKEUPWANTED;
Fix two cases in the new NFS server where a tsleep() is used, when the code should actually protect the tested variable with a mutex. Since the tsleep()s had a 10sec timeout, the race would have only delayed the allocation of a new clientid for a client. The sleeps will also rarely occur, since having a callback in progress when a client acquires a new clientid, is unlikely. in practice, since having a callback in progress when a fresh clientid is being acquired by a client is unlikely. MFC after: 1 month
2012-05-12 22:20:55 +00:00
wakeup(clp);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
Fix two cases in the new NFS server where a tsleep() is used, when the code should actually protect the tested variable with a mutex. Since the tsleep()s had a 10sec timeout, the race would have only delayed the allocation of a new clientid for a client. The sleeps will also rarely occur, since having a callback in progress when a client acquires a new clientid, is unlikely. in practice, since having a callback in progress when a fresh clientid is being acquired by a client is unlikely. MFC after: 1 month
2012-05-12 22:20:55 +00:00
NFSUNLOCKSTATE();
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
free(nd, M_TEMP);
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
NFSEXITCODE(error);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
return (error);
}
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
/*
* Set up the compound RPC for the callback.
*/
static int
nfsrv_cbcallargs(struct nfsrv_descript *nd, struct nfsclient *clp,
nfsd: fix session slot handling for failed callbacks When the NFSv4.1/4.2 server does a callback to a client on the back channel, it will use a session slot in the back channel session. If the back channel has failed, the callback will fail and, without this patch, the session slot will not be released. As more callbacks are attempted, all session slots can become busy and then the nfsd thread gets stuck waiting for a back channel session slot. This patch frees the session slot upon callback failure to avoid this problem. Without this patch, the problem can be avoided by leaving delegations disabled in the NFS server. MFC after: 2 weeks
2021-04-23 22:24:47 +00:00
uint32_t callback, int op, const char *optag, struct nfsdsession **sepp,
int *slotposp)
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
{
uint32_t *tl;
int error, len;
len = strlen(optag);
(void)nfsm_strtom(nd, optag, len);
NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
if ((nd->nd_flag & ND_NFSV41) != 0) {
Add support for NFSv4.2 to the NFS client and server. This patch adds support for NFSv4.2 (RFC-7862) and Extended Attributes (RFC-8276) to the NFS client and server. NFSv4.2 is comprised of several optional features that can be supported in addition to NFSv4.1. This patch adds the following optional features: - posix_fadvise(POSIX_FADV_WILLNEED/POSIX_FADV_DONTNEED) - posix_fallocate() - intra server file range copying via the copy_file_range(2) syscall --> Avoiding data tranfer over the wire to/from the NFS client. - lseek(SEEK_DATA/SEEK_HOLE) - Extended attribute syscalls for "user" namespace attributes as defined by RFC-8276. Although this patch is fairly large, it should not affect support for the other versions of NFS. However it does add two new sysctls that allow a sysadmin to limit which minor versions of NFSv4 a server supports, allowing a sysadmin to disable NFSv4.2. Unfortunately, when the NFS stats structure was last revised, it was assumed that there would be no additional operations added beyond what was specified in RFC-7862. However RFC-8276 did add additional operations, forcing the NFS stats structure to revised again. It now has extra unused entries in all arrays, so that future extensions to NFSv4.2 can be accomodated without revising this structure again. A future commit will update nfsstat(1) to report counts for the new NFSv4.2 specific operations/procedures. This patch affects the internal interface between the nfscommon, nfscl and nfsd modules and, as such, they all must be upgraded simultaneously. I will do a version bump (although arguably not needed), due to this. This code has survived a "make universe" but has not been built with a recent GCC. If you encounter build problems, please email me. Relnotes: yes
2019-12-12 23:22:55 +00:00
if ((nd->nd_flag & ND_NFSV42) != 0)
*tl++ = txdr_unsigned(NFSV42_MINORVERSION);
else
*tl++ = txdr_unsigned(NFSV41_MINORVERSION);
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
*tl++ = txdr_unsigned(callback);
*tl++ = txdr_unsigned(2);
*tl = txdr_unsigned(NFSV4OP_CBSEQUENCE);
nfsd: fix session slot handling for failed callbacks When the NFSv4.1/4.2 server does a callback to a client on the back channel, it will use a session slot in the back channel session. If the back channel has failed, the callback will fail and, without this patch, the session slot will not be released. As more callbacks are attempted, all session slots can become busy and then the nfsd thread gets stuck waiting for a back channel session slot. This patch frees the session slot upon callback failure to avoid this problem. Without this patch, the problem can be avoided by leaving delegations disabled in the NFS server. MFC after: 2 weeks
2021-04-23 22:24:47 +00:00
error = nfsv4_setcbsequence(nd, clp, 1, sepp, slotposp);
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
if (error != 0)
return (error);
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(op);
} else {
*tl++ = txdr_unsigned(NFSV4_MINORVERSION);
*tl++ = txdr_unsigned(callback);
*tl++ = txdr_unsigned(1);
*tl = txdr_unsigned(op);
}
return (0);
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* Return the next index# for a clientid. Mostly just increment and return
* the next one, but... if the 32bit unsigned does actually wrap around,
Change the comment at the beginning of the function to reflect the change from panic() to printf() done by r192588.
2009-05-22 16:46:01 +00:00
* it should be rebooted.
* At an average rate of one new client per second, it will wrap around in
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
* approximately 136 years. (I think the server will have been shut
* down or rebooted before then.)
*/
static u_int32_t
nfsrv_nextclientindex(void)
{
static u_int32_t client_index = 0;
client_index++;
if (client_index != 0)
return (client_index);
Change the printf of r192595 to identify the function, as requested by Sam. Approved by: kib (mentor)
2009-05-22 19:05:48 +00:00
printf("%s: out of clientids\n", __func__);
Change the reboot panic that would have occurred if clientid numbers wrapped around to a printf() warning of a possible DOS attack, in the experimental nfsv4 server. Approved by: kib (mentor)
2009-05-22 16:41:33 +00:00
return (client_index);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* Return the next index# for a stateid. Mostly just increment and return
* the next one, but... if the 32bit unsigned does actually wrap around
* (will a BSD server stay up that long?), find
* new start and end values.
*/
static u_int32_t
nfsrv_nextstateindex(struct nfsclient *clp)
{
struct nfsstate *stp;
int i;
u_int32_t canuse, min_index, max_index;
if (!(clp->lc_flags & LCL_INDEXNOTOK)) {
clp->lc_stateindex++;
if (clp->lc_stateindex != clp->lc_statemaxindex)
return (clp->lc_stateindex);
}
/*
* Yuck, we've hit the end.
* Look for a new min and max.
*/
min_index = 0;
max_index = 0xffffffff;
Make the size of the hash tables used by the NFSv4 server tunable. No appreciable change in performance was observed after increasing the sizes of these tables and then testing with a single client. However, there was an email that indicated high CPU overheads for a heavily loaded NFSv4 and it is hoped that increasing the sizes of the hash tables via these tunables might help. The tables remain the same size by default. Differential Revision: https://reviews.freebsd.org/D2596 MFC after: 2 weeks
2015-05-27 22:00:05 +00:00
for (i = 0; i < nfsrv_statehashsize; i++) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
LIST_FOREACH(stp, &clp->lc_stateid[i], ls_hash) {
if (stp->ls_stateid.other[2] > 0x80000000) {
if (stp->ls_stateid.other[2] < max_index)
max_index = stp->ls_stateid.other[2];
} else {
if (stp->ls_stateid.other[2] > min_index)
min_index = stp->ls_stateid.other[2];
}
}
}
/*
* Yikes, highly unlikely, but I'll handle it anyhow.
*/
if (min_index == 0x80000000 && max_index == 0x80000001) {
canuse = 0;
/*
* Loop around until we find an unused entry. Return that
* and set LCL_INDEXNOTOK, so the search will continue next time.
* (This is one of those rare cases where a goto is the
* cleanest way to code the loop.)
*/
tryagain:
Make the size of the hash tables used by the NFSv4 server tunable. No appreciable change in performance was observed after increasing the sizes of these tables and then testing with a single client. However, there was an email that indicated high CPU overheads for a heavily loaded NFSv4 and it is hoped that increasing the sizes of the hash tables via these tunables might help. The tables remain the same size by default. Differential Revision: https://reviews.freebsd.org/D2596 MFC after: 2 weeks
2015-05-27 22:00:05 +00:00
for (i = 0; i < nfsrv_statehashsize; i++) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
LIST_FOREACH(stp, &clp->lc_stateid[i], ls_hash) {
if (stp->ls_stateid.other[2] == canuse) {
canuse++;
goto tryagain;
}
}
}
clp->lc_flags |= LCL_INDEXNOTOK;
return (canuse);
}
/*
* Ok to start again from min + 1.
*/
clp->lc_stateindex = min_index + 1;
clp->lc_statemaxindex = max_index;
clp->lc_flags &= ~LCL_INDEXNOTOK;
return (clp->lc_stateindex);
}
/*
* The following functions handle the stable storage file that deals with
* the edge conditions described in RFC3530 Sec. 8.6.3.
* The file is as follows:
* - a single record at the beginning that has the lease time of the
* previous server instance (before the last reboot) and the nfsrvboottime
* values for the previous server boots.
* These previous boot times are used to ensure that the current
* nfsrvboottime does not, somehow, get set to a previous one.
* (This is important so that Stale ClientIDs and StateIDs can
* be recognized.)
NFS: spelling fixes on comments. No funcional change.
2016-04-29 16:07:25 +00:00
* The number of previous nfsvrboottime values precedes the list.
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
* - followed by some number of appended records with:
* - client id string
* - flag that indicates it is a record revoking state via lease
* expiration or similar
* OR has successfully acquired state.
* These structures vary in length, with the client string at the end, up
* to NFSV4_OPAQUELIMIT in size.
*
* At the end of the grace period, the file is truncated, the first
* record is rewritten with updated information and any acquired state
* records for successful reclaims of state are written.
*
* Subsequent records are appended when the first state is issued to
* a client and when state is revoked for a client.
*
* When reading the file in, state issued records that come later in
* the file override older ones, since the append log is in cronological order.
* If, for some reason, the file can't be read, the grace period is
* immediately terminated and all reclaims get NFSERR_NOGRACE.
*/
/*
* Read in the stable storage file. Called by nfssvc() before the nfsd
* processes start servicing requests.
*/
nfs: Remove APPLESTATIC macro It is no longer useful. Reviewed by: rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24811
2020-05-12 13:23:25 +00:00
void
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_setupstable(NFSPROC_T *p)
{
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
struct nfsrv_stablefirst *sf = &NFSD_VNET(nfsrv_stablefirst);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
struct nfsrv_stable *sp, *nsp;
struct nfst_rec *tsp;
int error, i, tryagain;
off_t off = 0;
Fix found places where uio_resid is truncated to int. Add the sysctl debug.iosize_max_clamp, enabled by default. Setting the sysctl to zero allows to perform the SSIZE_MAX-sized i/o requests from the usermode. Discussed with: bde, das (previous versions) MFC after: 1 month
2012-02-21 01:05:12 +00:00
ssize_t aresid, len;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* If NFSNSF_UPDATEDONE is set, this is a restart of the nfsds without
* a reboot, so state has not been lost.
*/
if (sf->nsf_flags & NFSNSF_UPDATEDONE)
return;
/*
* Set Grace over just until the file reads successfully.
*/
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
NFSD_VNET(nfsrvboottime) = time_second;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
LIST_INIT(&sf->nsf_head);
sf->nsf_flags = (NFSNSF_GRACEOVER | NFSNSF_NEEDLOCK);
sf->nsf_eograce = NFSD_MONOSEC + NFSRV_LEASEDELTA;
if (sf->nsf_fp == NULL)
return;
error = NFSD_RDWR(UIO_READ, NFSFPVNODE(sf->nsf_fp),
(caddr_t)&sf->nsf_rec, sizeof (struct nfsf_rec), off, UIO_SYSSPACE,
0, NFSFPCRED(sf->nsf_fp), &aresid, p);
if (error || aresid || sf->nsf_numboots == 0 ||
sf->nsf_numboots > NFSNSF_MAXNUMBOOTS)
return;
/*
* Now, read in the boottimes.
*/
sf->nsf_bootvals = (time_t *)malloc((sf->nsf_numboots + 1) *
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
sizeof(time_t), M_TEMP, M_WAITOK);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
off = sizeof (struct nfsf_rec);
error = NFSD_RDWR(UIO_READ, NFSFPVNODE(sf->nsf_fp),
(caddr_t)sf->nsf_bootvals, sf->nsf_numboots * sizeof (time_t), off,
UIO_SYSSPACE, 0, NFSFPCRED(sf->nsf_fp), &aresid, p);
if (error || aresid) {
style: Remove remaining deprecated MALLOC/FREE macros Mechanically replace uses of MALLOC/FREE with appropriate invocations of malloc(9) / free(9) (a series of sed expressions). Something like: * MALLOC(a, b, ... -> a = malloc(... * FREE( -> free( * free((caddr_t) -> free( No functional change. For now, punt on modifying contrib ipfilter code, leaving a definition of the macro in its KMALLOC(). Reported by: jhb Reviewed by: cy, imp, markj, rmacklem Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D14035
2018-01-25 22:25:13 +00:00
free(sf->nsf_bootvals, M_TEMP);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
sf->nsf_bootvals = NULL;
return;
}
/*
* Make sure this nfsrvboottime is different from all recorded
* previous ones.
*/
do {
tryagain = 0;
for (i = 0; i < sf->nsf_numboots; i++) {
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
if (NFSD_VNET(nfsrvboottime) == sf->nsf_bootvals[i]) {
NFSD_VNET(nfsrvboottime)++;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
tryagain = 1;
break;
}
}
} while (tryagain);
sf->nsf_flags |= NFSNSF_OK;
off += (sf->nsf_numboots * sizeof (time_t));
/*
* Read through the file, building a list of records for grace
* checking.
* Each record is between sizeof (struct nfst_rec) and
* sizeof (struct nfst_rec) + NFSV4_OPAQUELIMIT - 1
* and is actually sizeof (struct nfst_rec) + nst_len - 1.
*/
tsp = (struct nfst_rec *)malloc(sizeof (struct nfst_rec) +
NFSV4_OPAQUELIMIT - 1, M_TEMP, M_WAITOK);
do {
error = NFSD_RDWR(UIO_READ, NFSFPVNODE(sf->nsf_fp),
(caddr_t)tsp, sizeof (struct nfst_rec) + NFSV4_OPAQUELIMIT - 1,
off, UIO_SYSSPACE, 0, NFSFPCRED(sf->nsf_fp), &aresid, p);
len = (sizeof (struct nfst_rec) + NFSV4_OPAQUELIMIT - 1) - aresid;
if (error || (len > 0 && (len < sizeof (struct nfst_rec) ||
len < (sizeof (struct nfst_rec) + tsp->len - 1)))) {
/*
* Yuck, the file has been corrupted, so just return
* after clearing out any restart state, so the grace period
* is over.
*/
LIST_FOREACH_SAFE(sp, &sf->nsf_head, nst_list, nsp) {
LIST_REMOVE(sp, nst_list);
style: Remove remaining deprecated MALLOC/FREE macros Mechanically replace uses of MALLOC/FREE with appropriate invocations of malloc(9) / free(9) (a series of sed expressions). Something like: * MALLOC(a, b, ... -> a = malloc(... * FREE( -> free( * free((caddr_t) -> free( No functional change. For now, punt on modifying contrib ipfilter code, leaving a definition of the macro in its KMALLOC(). Reported by: jhb Reviewed by: cy, imp, markj, rmacklem Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D14035
2018-01-25 22:25:13 +00:00
free(sp, M_TEMP);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
style: Remove remaining deprecated MALLOC/FREE macros Mechanically replace uses of MALLOC/FREE with appropriate invocations of malloc(9) / free(9) (a series of sed expressions). Something like: * MALLOC(a, b, ... -> a = malloc(... * FREE( -> free( * free((caddr_t) -> free( No functional change. For now, punt on modifying contrib ipfilter code, leaving a definition of the macro in its KMALLOC(). Reported by: jhb Reviewed by: cy, imp, markj, rmacklem Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D14035
2018-01-25 22:25:13 +00:00
free(tsp, M_TEMP);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
sf->nsf_flags &= ~NFSNSF_OK;
style: Remove remaining deprecated MALLOC/FREE macros Mechanically replace uses of MALLOC/FREE with appropriate invocations of malloc(9) / free(9) (a series of sed expressions). Something like: * MALLOC(a, b, ... -> a = malloc(... * FREE( -> free( * free((caddr_t) -> free( No functional change. For now, punt on modifying contrib ipfilter code, leaving a definition of the macro in its KMALLOC(). Reported by: jhb Reviewed by: cy, imp, markj, rmacklem Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D14035
2018-01-25 22:25:13 +00:00
free(sf->nsf_bootvals, M_TEMP);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
sf->nsf_bootvals = NULL;
return;
}
if (len > 0) {
off += sizeof (struct nfst_rec) + tsp->len - 1;
/*
* Search the list for a matching client.
*/
LIST_FOREACH(sp, &sf->nsf_head, nst_list) {
if (tsp->len == sp->nst_len &&
!NFSBCMP(tsp->client, sp->nst_client, tsp->len))
break;
}
if (sp == LIST_END(&sf->nsf_head)) {
sp = (struct nfsrv_stable *)malloc(tsp->len +
sizeof (struct nfsrv_stable) - 1, M_TEMP,
M_WAITOK);
NFSBCOPY((caddr_t)tsp, (caddr_t)&sp->nst_rec,
sizeof (struct nfst_rec) + tsp->len - 1);
LIST_INSERT_HEAD(&sf->nsf_head, sp, nst_list);
} else {
if (tsp->flag == NFSNST_REVOKE)
sp->nst_flag |= NFSNST_REVOKE;
else
/*
* A subsequent timestamp indicates the client
* did a setclientid/confirm and any previous
* revoke is no longer relevant.
*/
sp->nst_flag &= ~NFSNST_REVOKE;
}
}
} while (len > 0);
style: Remove remaining deprecated MALLOC/FREE macros Mechanically replace uses of MALLOC/FREE with appropriate invocations of malloc(9) / free(9) (a series of sed expressions). Something like: * MALLOC(a, b, ... -> a = malloc(... * FREE( -> free( * free((caddr_t) -> free( No functional change. For now, punt on modifying contrib ipfilter code, leaving a definition of the macro in its KMALLOC(). Reported by: jhb Reviewed by: cy, imp, markj, rmacklem Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D14035
2018-01-25 22:25:13 +00:00
free(tsp, M_TEMP);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
sf->nsf_flags = NFSNSF_OK;
sf->nsf_eograce = NFSD_MONOSEC + sf->nsf_lease +
NFSRV_LEASEDELTA;
}
/*
* Update the stable storage file, now that the grace period is over.
*/
nfs: Remove APPLESTATIC macro It is no longer useful. Reviewed by: rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24811
2020-05-12 13:23:25 +00:00
void
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_updatestable(NFSPROC_T *p)
{
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
struct nfsrv_stablefirst *sf = &NFSD_VNET(nfsrv_stablefirst);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
struct nfsrv_stable *sp, *nsp;
int i;
struct nfsvattr nva;
vnode_t vp;
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
#if defined(__FreeBSD_version) && (__FreeBSD_version >= 500000)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
mount_t mp = NULL;
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
#endif
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
int error;
if (sf->nsf_fp == NULL || (sf->nsf_flags & NFSNSF_UPDATEDONE))
return;
sf->nsf_flags |= NFSNSF_UPDATEDONE;
/*
* Ok, we need to rewrite the stable storage file.
* - truncate to 0 length
* - write the new first structure
* - loop through the data structures, writing out any that
* have timestamps older than the old boot
*/
if (sf->nsf_bootvals) {
sf->nsf_numboots++;
for (i = sf->nsf_numboots - 2; i >= 0; i--)
sf->nsf_bootvals[i + 1] = sf->nsf_bootvals[i];
} else {
sf->nsf_numboots = 1;
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
sf->nsf_bootvals = (time_t *)malloc(sizeof(time_t),
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
M_TEMP, M_WAITOK);
}
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
sf->nsf_bootvals[0] = NFSD_VNET(nfsrvboottime);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
sf->nsf_lease = nfsrv_lease;
NFSVNO_ATTRINIT(&nva);
NFSVNO_SETATTRVAL(&nva, size, 0);
vp = NFSFPVNODE(sf->nsf_fp);
Delete the NFS_STARTWRITE() and NFS_ENDWRITE() macros that obscured vn_start_write() and vn_finished_write() for the old OpenBSD port, since most uses have been replaced by the correct calls. MFC after: 12 days
2011-01-06 20:31:33 +00:00
vn_start_write(vp, &mp, V_WAIT);
Simple find/replace of vn_lock -> NFSVOPLOCK. This is done so that NFSVOPLOCK can be modified later to add enhanced logging and assertions. Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:05:31 +00:00
if (NFSVOPLOCK(vp, LK_EXCLUSIVE) == 0) {
Add checks for VI_DOOMED and vn_lock() failures to the experimental NFS server, to handle the case where an exported file system is forced dismounted while an RPC is in progress. Further commits will fix the cases where a mount point is used when the associated vnode isn't locked. Reviewed by: kib MFC after: 2 weeks
2011-01-02 19:58:39 +00:00
error = nfsvno_setattr(vp, &nva, NFSFPCRED(sf->nsf_fp), p,
NULL);
vfs: drop the mostly unused flags argument from VOP_UNLOCK Filesystems which want to use it in limited capacity can employ the VOP_UNLOCK_FLAGS macro. Reviewed by: kib (previous version) Differential Revision: https://reviews.freebsd.org/D21427
2020-01-03 22:29:58 +00:00
NFSVOPUNLOCK(vp);
Add checks for VI_DOOMED and vn_lock() failures to the experimental NFS server, to handle the case where an exported file system is forced dismounted while an RPC is in progress. Further commits will fix the cases where a mount point is used when the associated vnode isn't locked. Reviewed by: kib MFC after: 2 weeks
2011-01-02 19:58:39 +00:00
} else
error = EPERM;
Delete the NFS_STARTWRITE() and NFS_ENDWRITE() macros that obscured vn_start_write() and vn_finished_write() for the old OpenBSD port, since most uses have been replaced by the correct calls. MFC after: 12 days
2011-01-06 20:31:33 +00:00
vn_finished_write(mp);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (!error)
error = NFSD_RDWR(UIO_WRITE, vp,
(caddr_t)&sf->nsf_rec, sizeof (struct nfsf_rec), (off_t)0,
UIO_SYSSPACE, IO_SYNC, NFSFPCRED(sf->nsf_fp), NULL, p);
if (!error)
error = NFSD_RDWR(UIO_WRITE, vp,
(caddr_t)sf->nsf_bootvals,
sf->nsf_numboots * sizeof (time_t),
(off_t)(sizeof (struct nfsf_rec)),
UIO_SYSSPACE, IO_SYNC, NFSFPCRED(sf->nsf_fp), NULL, p);
style: Remove remaining deprecated MALLOC/FREE macros Mechanically replace uses of MALLOC/FREE with appropriate invocations of malloc(9) / free(9) (a series of sed expressions). Something like: * MALLOC(a, b, ... -> a = malloc(... * FREE( -> free( * free((caddr_t) -> free( No functional change. For now, punt on modifying contrib ipfilter code, leaving a definition of the macro in its KMALLOC(). Reported by: jhb Reviewed by: cy, imp, markj, rmacklem Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D14035
2018-01-25 22:25:13 +00:00
free(sf->nsf_bootvals, M_TEMP);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
sf->nsf_bootvals = NULL;
if (error) {
sf->nsf_flags &= ~NFSNSF_OK;
printf("EEK! Can't write NfsV4 stable storage file\n");
return;
}
sf->nsf_flags |= NFSNSF_OK;
/*
* Loop through the list and write out timestamp records for
* any clients that successfully reclaimed state.
*/
LIST_FOREACH_SAFE(sp, &sf->nsf_head, nst_list, nsp) {
if (sp->nst_flag & NFSNST_GOTSTATE) {
nfsrv_writestable(sp->nst_client, sp->nst_len,
NFSNST_NEWSTATE, p);
sp->nst_clp->lc_flags |= LCL_STAMPEDSTABLE;
}
LIST_REMOVE(sp, nst_list);
style: Remove remaining deprecated MALLOC/FREE macros Mechanically replace uses of MALLOC/FREE with appropriate invocations of malloc(9) / free(9) (a series of sed expressions). Something like: * MALLOC(a, b, ... -> a = malloc(... * FREE( -> free( * free((caddr_t) -> free( No functional change. For now, punt on modifying contrib ipfilter code, leaving a definition of the macro in its KMALLOC(). Reported by: jhb Reviewed by: cy, imp, markj, rmacklem Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D14035
2018-01-25 22:25:13 +00:00
free(sp, M_TEMP);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
Modify the experimental NFSv4 server so that it posts a SIGUSR2 signal to the master nfsd daemon whenever the stable restart file has been modified. This will allow the master nfsd daemon to maintain an up to date backup copy of the file. This is enabled via the nfssvc() syscall, so that older nfsd daemons will not be signaled. Reviewed by: jhb MFC after: 1 week
2011-01-14 23:30:35 +00:00
nfsrv_backupstable();
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* Append a record to the stable storage file.
*/
nfs: Remove APPLESTATIC macro It is no longer useful. Reviewed by: rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24811
2020-05-12 13:23:25 +00:00
void
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_writestable(u_char *client, int len, int flag, NFSPROC_T *p)
{
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
struct nfsrv_stablefirst *sf = &NFSD_VNET(nfsrv_stablefirst);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
struct nfst_rec *sp;
int error;
if (!(sf->nsf_flags & NFSNSF_OK) || sf->nsf_fp == NULL)
return;
sp = (struct nfst_rec *)malloc(sizeof (struct nfst_rec) +
len - 1, M_TEMP, M_WAITOK);
sp->len = len;
NFSBCOPY(client, sp->client, len);
sp->flag = flag;
error = NFSD_RDWR(UIO_WRITE, NFSFPVNODE(sf->nsf_fp),
(caddr_t)sp, sizeof (struct nfst_rec) + len - 1, (off_t)0,
UIO_SYSSPACE, (IO_SYNC | IO_APPEND), NFSFPCRED(sf->nsf_fp), NULL, p);
style: Remove remaining deprecated MALLOC/FREE macros Mechanically replace uses of MALLOC/FREE with appropriate invocations of malloc(9) / free(9) (a series of sed expressions). Something like: * MALLOC(a, b, ... -> a = malloc(... * FREE( -> free( * free((caddr_t) -> free( No functional change. For now, punt on modifying contrib ipfilter code, leaving a definition of the macro in its KMALLOC(). Reported by: jhb Reviewed by: cy, imp, markj, rmacklem Sponsored by: Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D14035
2018-01-25 22:25:13 +00:00
free(sp, M_TEMP);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (error) {
sf->nsf_flags &= ~NFSNSF_OK;
printf("EEK! Can't write NfsV4 stable storage file\n");
}
}
/*
* This function is called during the grace period to mark a client
* that successfully reclaimed state.
*/
static void
nfsrv_markstable(struct nfsclient *clp)
{
struct nfsrv_stable *sp;
/*
* First find the client structure.
*/
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
LIST_FOREACH(sp, &NFSD_VNET(nfsrv_stablefirst).nsf_head, nst_list) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (sp->nst_len == clp->lc_idlen &&
!NFSBCMP(sp->nst_client, clp->lc_id, sp->nst_len))
break;
}
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
if (sp == LIST_END(&NFSD_VNET(nfsrv_stablefirst).nsf_head))
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
return;
/*
* Now, just mark it and set the nfsclient back pointer.
*/
sp->nst_flag |= NFSNST_GOTSTATE;
sp->nst_clp = clp;
}
End grace for the NFSv4 server if all mounts do ReclaimComplete. The NFSv4 protocol requires that the server only allow reclaim of state and not issue any new open/lock state for a grace period after booting. The NFSv4.0 protocol required this grace period to be greater than the lease duration (over 2minutes). For NFSv4.1, the client tells the server that it has done reclaiming state by doing a ReclaimComplete operation. If all NFSv4 clients are NFSv4.1, the grace period can end once all the clients have done ReclaimComplete, shortening the time period considerably. This patch does this. If there are any NFSv4.0 mounts, the grace period will still be over 2minutes. This change is only an optimization and does not affect correct operation. Tested by: andreas.nagy@frequentis.com MFC after: 2 months
2018-05-15 20:28:50 +00:00
/*
* This function is called when a NFSv4.1 client does a ReclaimComplete.
* Very similar to nfsrv_markstable(), except for the flag being set.
*/
static void
nfsrv_markreclaim(struct nfsclient *clp)
{
struct nfsrv_stable *sp;
/*
* First find the client structure.
*/
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
LIST_FOREACH(sp, &NFSD_VNET(nfsrv_stablefirst).nsf_head, nst_list) {
End grace for the NFSv4 server if all mounts do ReclaimComplete. The NFSv4 protocol requires that the server only allow reclaim of state and not issue any new open/lock state for a grace period after booting. The NFSv4.0 protocol required this grace period to be greater than the lease duration (over 2minutes). For NFSv4.1, the client tells the server that it has done reclaiming state by doing a ReclaimComplete operation. If all NFSv4 clients are NFSv4.1, the grace period can end once all the clients have done ReclaimComplete, shortening the time period considerably. This patch does this. If there are any NFSv4.0 mounts, the grace period will still be over 2minutes. This change is only an optimization and does not affect correct operation. Tested by: andreas.nagy@frequentis.com MFC after: 2 months
2018-05-15 20:28:50 +00:00
if (sp->nst_len == clp->lc_idlen &&
!NFSBCMP(sp->nst_client, clp->lc_id, sp->nst_len))
break;
}
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
if (sp == LIST_END(&NFSD_VNET(nfsrv_stablefirst).nsf_head))
End grace for the NFSv4 server if all mounts do ReclaimComplete. The NFSv4 protocol requires that the server only allow reclaim of state and not issue any new open/lock state for a grace period after booting. The NFSv4.0 protocol required this grace period to be greater than the lease duration (over 2minutes). For NFSv4.1, the client tells the server that it has done reclaiming state by doing a ReclaimComplete operation. If all NFSv4 clients are NFSv4.1, the grace period can end once all the clients have done ReclaimComplete, shortening the time period considerably. This patch does this. If there are any NFSv4.0 mounts, the grace period will still be over 2minutes. This change is only an optimization and does not affect correct operation. Tested by: andreas.nagy@frequentis.com MFC after: 2 months
2018-05-15 20:28:50 +00:00
return;
/*
* Now, just set the flag.
*/
sp->nst_flag |= NFSNST_RECLAIMED;
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* This function is called for a reclaim, to see if it gets grace.
* It returns 0 if a reclaim is allowed, 1 otherwise.
*/
static int
nfsrv_checkstable(struct nfsclient *clp)
{
struct nfsrv_stable *sp;
/*
* First, find the entry for the client.
*/
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
LIST_FOREACH(sp, &NFSD_VNET(nfsrv_stablefirst).nsf_head, nst_list) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (sp->nst_len == clp->lc_idlen &&
!NFSBCMP(sp->nst_client, clp->lc_id, sp->nst_len))
break;
}
/*
* If not in the list, state was revoked or no state was issued
* since the previous reboot, a reclaim is denied.
*/
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
if (sp == LIST_END(&NFSD_VNET(nfsrv_stablefirst).nsf_head) ||
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
(sp->nst_flag & NFSNST_REVOKE) ||
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
!(NFSD_VNET(nfsrv_stablefirst).nsf_flags & NFSNSF_OK))
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
return (1);
return (0);
}
/*
* Test for and try to clear out a conflicting client. This is called by
* nfsrv_lockctrl() and nfsrv_openctrl() when conflicts with other clients
* a found.
* The trick here is that it can't revoke a conflicting client with an
* expired lease unless it holds the v4root lock, so...
* If no v4root lock, get the lock and return 1 to indicate "try again".
* Return 0 to indicate the conflict can't be revoked and 1 to indicate
* the revocation worked and the conflicting client is "bye, bye", so it
* can be tried again.
vfs: introduce v_irflag and make v_type smaller The current vnode layout is not smp-friendly by having frequently read data avoidably sharing cachelines with very frequently modified fields. In particular v_iflag inspected for VI_DOOMED can be found in the same line with v_usecount. Instead make it available in the same cacheline as the v_op, v_data and v_type which all get read all the time. v_type is avoidably 4 bytes while the necessary data will easily fit in 1. Shrinking it frees up 3 bytes, 2 of which get used here to introduce a new flag field with a new value: VIRF_DOOMED. Reviewed by: kib, jeff Differential Revision: https://reviews.freebsd.org/D22715
2019-12-08 21:30:04 +00:00
* Return 2 to indicate that the vnode is VIRF_DOOMED after NFSVOPLOCK().
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
* Unlocks State before a non-zero value is returned.
*/
static int
Add checks for VI_DOOMED and vn_lock() failures to the experimental NFS server, to handle the case where an exported file system is forced dismounted while an RPC is in progress. Further commits will fix the cases where a mount point is used when the associated vnode isn't locked. Reviewed by: kib MFC after: 2 weeks
2011-01-02 19:58:39 +00:00
nfsrv_clientconflict(struct nfsclient *clp, int *haslockp, vnode_t vp,
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
NFSPROC_T *p)
{
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
int gotlock, lktype = 0;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* If lease hasn't expired, we can't fix it.
*/
if (clp->lc_expiry >= NFSD_MONOSEC ||
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
!(NFSD_VNET(nfsrv_stablefirst).nsf_flags & NFSNSF_UPDATEDONE))
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
return (0);
if (*haslockp == 0) {
NFSUNLOCKSTATE();
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
if (vp != NULL) {
lktype = NFSVOPISLOCKED(vp);
vfs: drop the mostly unused flags argument from VOP_UNLOCK Filesystems which want to use it in limited capacity can employ the VOP_UNLOCK_FLAGS macro. Reviewed by: kib (previous version) Differential Revision: https://reviews.freebsd.org/D21427
2020-01-03 22:29:58 +00:00
NFSVOPUNLOCK(vp);
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
NFSLOCKV4ROOTMUTEX();
nfsv4_relref(&nfsv4rootfs_lock);
do {
gotlock = nfsv4_lock(&nfsv4rootfs_lock, 1, NULL,
Fix the new NFS client so that it handles NFSv4 state correctly during a forced dismount. This required that the exclusive and shared (refcnt) sleep lock functions check for MNTK_UMOUNTF before sleeping, so that they won't block while nfscl_umount() is getting rid of the state. As such, a "struct mount *" argument was added to the locking functions. I believe the only remaining case where a forced dismount can get hung in the kernel is when a thread is already attempting to do a TCP connect to a dead server when the krpc client structure called nr_client is NULL. This will only happen just after a "mount -u" with options that force a new TCP connection is done, so it shouldn't be a problem in practice. MFC after: 2 weeks
2011-05-27 22:05:10 +00:00
NFSV4ROOTLOCKMUTEXPTR, NULL);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
} while (!gotlock);
NFSUNLOCKV4ROOTMUTEX();
*haslockp = 1;
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
if (vp != NULL) {
NFSVOPLOCK(vp, lktype | LK_RETRY);
vfs: introduce v_irflag and make v_type smaller The current vnode layout is not smp-friendly by having frequently read data avoidably sharing cachelines with very frequently modified fields. In particular v_iflag inspected for VI_DOOMED can be found in the same line with v_usecount. Instead make it available in the same cacheline as the v_op, v_data and v_type which all get read all the time. v_type is avoidably 4 bytes while the necessary data will easily fit in 1. Shrinking it frees up 3 bytes, 2 of which get used here to introduce a new flag field with a new value: VIRF_DOOMED. Reviewed by: kib, jeff Differential Revision: https://reviews.freebsd.org/D22715
2019-12-08 21:30:04 +00:00
if (VN_IS_DOOMED(vp))
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
return (2);
}
return (1);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
NFSUNLOCKSTATE();
/*
* Ok, we can expire the conflicting client.
*/
nfsrv_writestable(clp->lc_id, clp->lc_idlen, NFSNST_REVOKE, p);
Modify the experimental NFSv4 server so that it posts a SIGUSR2 signal to the master nfsd daemon whenever the stable restart file has been modified. This will allow the master nfsd daemon to maintain an up to date backup copy of the file. This is enabled via the nfssvc() syscall, so that older nfsd daemons will not be signaled. Reviewed by: jhb MFC after: 1 week
2011-01-14 23:30:35 +00:00
nfsrv_backupstable();
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_cleanclient(clp, p);
nfsrv_freedeleglist(&clp->lc_deleg);
nfsrv_freedeleglist(&clp->lc_olddeleg);
LIST_REMOVE(clp, lc_hash);
nfsrv_zapclient(clp, p);
return (1);
}
/*
* Resolve a delegation conflict.
* Returns 0 to indicate the conflict was resolved without sleeping.
* Return -1 to indicate that the caller should check for conflicts again.
* Return > 0 for an error that should be returned, normally NFSERR_DELAY.
*
* Also, manipulate the nfsv4root_lock, as required. It isn't changed
* for a return of 0, since there was no sleep and it could be required
* later. It is released for a return of NFSERR_DELAY, since the caller
* will return that error. It is released when a sleep was done waiting
* for the delegation to be returned or expire (so that other nfsds can
* handle ops). Then, it must be acquired for the write to stable storage.
* (This function is somewhat similar to nfsrv_clientconflict(), but
* the semantics differ in a couple of subtle ways. The return of 0
* indicates the conflict was resolved without sleeping here, not
* that the conflict can't be resolved and the handling of nfsv4root_lock
* differs, as noted above.)
* Unlocks State before returning a non-zero value.
*/
static int
nfsrv_delegconflict(struct nfsstate *stp, int *haslockp, NFSPROC_T *p,
Fix two vnode locking problems in nfsd_recalldelegation() in the experimental NFSv4 server. The first was a bogus use of VOP_ISLOCKED() in a KASSERT() and the second was the need to lock the vnode for the nfsrv_checkremove() call. Also, delete a "__unused" that was bogus, since the argument is used. Reviewed by: zack.kirsch at isilon.com MFC after: 2 weeks
2010-12-17 22:18:09 +00:00
vnode_t vp)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
{
struct nfsclient *clp = stp->ls_clp;
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
int gotlock, error, lktype = 0, retrycnt, zapped_clp;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsv4stateid_t tstateid;
fhandle_t tfh;
/*
* If the conflict is with an old delegation...
*/
if (stp->ls_flags & NFSLCK_OLDDELEG) {
/*
* You can delete it, if it has expired.
*/
if (clp->lc_delegtime < NFSD_MONOSEC) {
nfsrv_freedeleg(stp);
NFSUNLOCKSTATE();
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
error = -1;
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
NFSUNLOCKSTATE();
/*
* During this delay, the old delegation could expire or it
* could be recovered by the client via an Open with
* CLAIM_DELEGATE_PREV.
* Release the nfsv4root_lock, if held.
*/
if (*haslockp) {
*haslockp = 0;
NFSLOCKV4ROOTMUTEX();
nfsv4_unlock(&nfsv4rootfs_lock, 1);
NFSUNLOCKV4ROOTMUTEX();
}
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
error = NFSERR_DELAY;
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* It's a current delegation, so:
* - check to see if the delegation has expired
* - if so, get the v4root lock and then expire it
*/
NFSv4 server: Re-establish the delegation recall timeout Commit 7a606f280a3e allowed the server to do retries of CB_RECALL callbacks every couple of seconds. This was needed to allow the Linux client to re-establish the back channel. However this patch broke the delegation timeout check, such that it would just keep retrying CB_RECALLS. If the client has crashed or been network patitioned from the server, this continues until the client TCP reconnects to the server and re-establishes the back channel. This patch modifies the code such that it still times out the delegation recall after some minutes, so that the server will allow the conflicting client request once the delegation times out. This patch only affects the NFSv4 server when delegations are enabled and a NFSv4 client that holds a delegation has crashed or been network partitioned from the server for at least several minutes when a delegation needs to be recalled. MFC after: 2 weeks
2021-05-16 23:40:01 +00:00
if ((stp->ls_flags & NFSLCK_DELEGRECALL) == 0 || (stp->ls_lastrecall <
NFSD_MONOSEC && clp->lc_expiry >= NFSD_MONOSEC &&
stp->ls_delegtime >= NFSD_MONOSEC)) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* - do a recall callback, since not yet done
* For now, never allow truncate to be set. To use
* truncate safely, it must be guaranteed that the
* Remove, Rename or Setattr with size of 0 will
* succeed and that would require major changes to
* the VFS/Vnode OPs.
* Set the expiry time large enough so that it won't expire
* until after the callback, then set it correctly, once
* the callback is done. (The delegation will now time
* out whether or not the Recall worked ok. The timeout
* will be extended when ops are done on the delegation
* stateid, up to the timelimit.)
*/
nfsd: make the server repeat CB_RECALL every couple of seconds Commit 01ae8969a9ee stopped the NFSv4.1/4.2 server from implicitly binding the back channel to a new TCP connection so that it conforms to RFC5661, for NFSv4.1/4.2. An effect of this for the Linux NFS client is that it will do a BindConnectionToSession when it sees NFSV4SEQ_CBPATHDOWN set in a sequence reply. This will fix the back channel, but the first attempt at a callback like CB_RECALL will already have failed. Without this patch, a CB_RECALL will not be retried and that can result in a 5 minute delay until the delegation times out. This patch modifies the code so that it will retry the CB_RECALL every couple of seconds, often avoiding the 5 minute delay. This is not critical for correct behaviour, but avoids the 5 minute delay for the case where the Linux client re-binds the back channel via BindConnectionToSession. MFC after: 2 weeks
2021-04-05 01:15:54 +00:00
if ((stp->ls_flags & NFSLCK_DELEGRECALL) == 0) {
stp->ls_delegtime = NFSD_MONOSEC + (2 * nfsrv_lease) +
NFSRV_LEASEDELTA;
stp->ls_delegtimelimit = NFSD_MONOSEC + (6 *
nfsrv_lease) + NFSRV_LEASEDELTA;
stp->ls_flags |= NFSLCK_DELEGRECALL;
}
stp->ls_lastrecall = time_uptime + 1;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* Loop NFSRV_CBRETRYCNT times while the CBRecall replies
* NFSERR_BADSTATEID or NFSERR_BADHANDLE. This is done
* in order to try and avoid a race that could happen
* when a CBRecall request passed the Open reply with
* the delegation in it when transitting the network.
* Since nfsrv_docallback will sleep, don't use stp after
* the call.
*/
NFSBCOPY((caddr_t)&stp->ls_stateid, (caddr_t)&tstateid,
sizeof (tstateid));
NFSBCOPY((caddr_t)&stp->ls_lfp->lf_fh, (caddr_t)&tfh,
sizeof (tfh));
NFSUNLOCKSTATE();
if (*haslockp) {
*haslockp = 0;
NFSLOCKV4ROOTMUTEX();
nfsv4_unlock(&nfsv4rootfs_lock, 1);
NFSUNLOCKV4ROOTMUTEX();
}
retrycnt = 0;
do {
error = nfsrv_docallback(clp, NFSV4OP_CBRECALL,
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
&tstateid, 0, &tfh, NULL, NULL, 0, p);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
retrycnt++;
} while ((error == NFSERR_BADSTATEID ||
error == NFSERR_BADHANDLE) && retrycnt < NFSV4_CBRETRYCNT);
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
error = NFSERR_DELAY;
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
if (clp->lc_expiry >= NFSD_MONOSEC &&
stp->ls_delegtime >= NFSD_MONOSEC) {
NFSUNLOCKSTATE();
/*
* A recall has been done, but it has not yet expired.
* So, RETURN_DELAY.
*/
if (*haslockp) {
*haslockp = 0;
NFSLOCKV4ROOTMUTEX();
nfsv4_unlock(&nfsv4rootfs_lock, 1);
NFSUNLOCKV4ROOTMUTEX();
}
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
error = NFSERR_DELAY;
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* If we don't yet have the lock, just get it and then return,
* since we need that before deleting expired state, such as
* this delegation.
* When getting the lock, unlock the vnode, so other nfsds that
* are in progress, won't get stuck waiting for the vnode lock.
*/
if (*haslockp == 0) {
NFSUNLOCKSTATE();
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
if (vp != NULL) {
lktype = NFSVOPISLOCKED(vp);
vfs: drop the mostly unused flags argument from VOP_UNLOCK Filesystems which want to use it in limited capacity can employ the VOP_UNLOCK_FLAGS macro. Reviewed by: kib (previous version) Differential Revision: https://reviews.freebsd.org/D21427
2020-01-03 22:29:58 +00:00
NFSVOPUNLOCK(vp);
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
NFSLOCKV4ROOTMUTEX();
nfsv4_relref(&nfsv4rootfs_lock);
do {
gotlock = nfsv4_lock(&nfsv4rootfs_lock, 1, NULL,
Fix the new NFS client so that it handles NFSv4 state correctly during a forced dismount. This required that the exclusive and shared (refcnt) sleep lock functions check for MNTK_UMOUNTF before sleeping, so that they won't block while nfscl_umount() is getting rid of the state. As such, a "struct mount *" argument was added to the locking functions. I believe the only remaining case where a forced dismount can get hung in the kernel is when a thread is already attempting to do a TCP connect to a dead server when the krpc client structure called nr_client is NULL. This will only happen just after a "mount -u" with options that force a new TCP connection is done, so it shouldn't be a problem in practice. MFC after: 2 weeks
2011-05-27 22:05:10 +00:00
NFSV4ROOTLOCKMUTEXPTR, NULL);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
} while (!gotlock);
NFSUNLOCKV4ROOTMUTEX();
*haslockp = 1;
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
if (vp != NULL) {
NFSVOPLOCK(vp, lktype | LK_RETRY);
vfs: introduce v_irflag and make v_type smaller The current vnode layout is not smp-friendly by having frequently read data avoidably sharing cachelines with very frequently modified fields. In particular v_iflag inspected for VI_DOOMED can be found in the same line with v_usecount. Instead make it available in the same cacheline as the v_op, v_data and v_type which all get read all the time. v_type is avoidably 4 bytes while the necessary data will easily fit in 1. Shrinking it frees up 3 bytes, 2 of which get used here to introduce a new flag field with a new value: VIRF_DOOMED. Reviewed by: kib, jeff Differential Revision: https://reviews.freebsd.org/D22715
2019-12-08 21:30:04 +00:00
if (VN_IS_DOOMED(vp)) {
A deadlock in the NFSv4 server with vfs.nfsd.enable_locallocks=1 was reported via email. This was caused by a LOR between the sleep lock used to serialize the local locking (nfsrv_locklf()) and locking the vnode. I believe this patch fixes the problem by delaying relocking of the vnode until the sleep lock is unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side effect of unlocking the vnode, unlocking the vnode was moved to before the functions that call nfsvno_advlock(). It shouldn't affect the execution of the default case where vfs.nfsd.enable_locallocks=0. Reported by: loic.blot@unix-experience.fr Discussed with: kib MFC after: 1 week
2014-12-25 01:55:17 +00:00
*haslockp = 0;
NFSLOCKV4ROOTMUTEX();
nfsv4_unlock(&nfsv4rootfs_lock, 1);
NFSUNLOCKV4ROOTMUTEX();
error = NFSERR_PERM;
goto out;
}
Add checks for VI_DOOMED and vn_lock() failures to the experimental NFS server, to handle the case where an exported file system is forced dismounted while an RPC is in progress. Further commits will fix the cases where a mount point is used when the associated vnode isn't locked. Reviewed by: kib MFC after: 2 weeks
2011-01-02 19:58:39 +00:00
}
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
error = -1;
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
NFSUNLOCKSTATE();
/*
* Ok, we can delete the expired delegation.
* First, write the Revoke record to stable storage and then
* clear out the conflict.
* Since all other nfsd threads are now blocked, we can safely
* sleep without the state changing.
*/
nfsrv_writestable(clp->lc_id, clp->lc_idlen, NFSNST_REVOKE, p);
Modify the experimental NFSv4 server so that it posts a SIGUSR2 signal to the master nfsd daemon whenever the stable restart file has been modified. This will allow the master nfsd daemon to maintain an up to date backup copy of the file. This is enabled via the nfssvc() syscall, so that older nfsd daemons will not be signaled. Reviewed by: jhb MFC after: 1 week
2011-01-14 23:30:35 +00:00
nfsrv_backupstable();
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (clp->lc_expiry < NFSD_MONOSEC) {
nfsrv_cleanclient(clp, p);
nfsrv_freedeleglist(&clp->lc_deleg);
nfsrv_freedeleglist(&clp->lc_olddeleg);
LIST_REMOVE(clp, lc_hash);
zapped_clp = 1;
} else {
nfsrv_freedeleg(stp);
zapped_clp = 0;
}
if (zapped_clp)
nfsrv_zapclient(clp, p);
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
error = -1;
out:
NFSEXITCODE(error);
return (error);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* Check for a remove allowed, if remove is set to 1 and get rid of
* delegations.
*/
nfs: Remove APPLESTATIC macro It is no longer useful. Reviewed by: rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24811
2020-05-12 13:23:25 +00:00
int
Add support for NFSv4.2 to the NFS client and server. This patch adds support for NFSv4.2 (RFC-7862) and Extended Attributes (RFC-8276) to the NFS client and server. NFSv4.2 is comprised of several optional features that can be supported in addition to NFSv4.1. This patch adds the following optional features: - posix_fadvise(POSIX_FADV_WILLNEED/POSIX_FADV_DONTNEED) - posix_fallocate() - intra server file range copying via the copy_file_range(2) syscall --> Avoiding data tranfer over the wire to/from the NFS client. - lseek(SEEK_DATA/SEEK_HOLE) - Extended attribute syscalls for "user" namespace attributes as defined by RFC-8276. Although this patch is fairly large, it should not affect support for the other versions of NFS. However it does add two new sysctls that allow a sysadmin to limit which minor versions of NFSv4 a server supports, allowing a sysadmin to disable NFSv4.2. Unfortunately, when the NFS stats structure was last revised, it was assumed that there would be no additional operations added beyond what was specified in RFC-7862. However RFC-8276 did add additional operations, forcing the NFS stats structure to revised again. It now has extra unused entries in all arrays, so that future extensions to NFSv4.2 can be accomodated without revising this structure again. A future commit will update nfsstat(1) to report counts for the new NFSv4.2 specific operations/procedures. This patch affects the internal interface between the nfscommon, nfscl and nfsd modules and, as such, they all must be upgraded simultaneously. I will do a version bump (although arguably not needed), due to this. This code has survived a "make universe" but has not been built with a recent GCC. If you encounter build problems, please email me. Relnotes: yes
2019-12-12 23:22:55 +00:00
nfsrv_checkremove(vnode_t vp, int remove, struct nfsrv_descript *nd,
nfsquad_t clientid, NFSPROC_T *p)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
{
Add support for NFSv4.2 to the NFS client and server. This patch adds support for NFSv4.2 (RFC-7862) and Extended Attributes (RFC-8276) to the NFS client and server. NFSv4.2 is comprised of several optional features that can be supported in addition to NFSv4.1. This patch adds the following optional features: - posix_fadvise(POSIX_FADV_WILLNEED/POSIX_FADV_DONTNEED) - posix_fallocate() - intra server file range copying via the copy_file_range(2) syscall --> Avoiding data tranfer over the wire to/from the NFS client. - lseek(SEEK_DATA/SEEK_HOLE) - Extended attribute syscalls for "user" namespace attributes as defined by RFC-8276. Although this patch is fairly large, it should not affect support for the other versions of NFS. However it does add two new sysctls that allow a sysadmin to limit which minor versions of NFSv4 a server supports, allowing a sysadmin to disable NFSv4.2. Unfortunately, when the NFS stats structure was last revised, it was assumed that there would be no additional operations added beyond what was specified in RFC-7862. However RFC-8276 did add additional operations, forcing the NFS stats structure to revised again. It now has extra unused entries in all arrays, so that future extensions to NFSv4.2 can be accomodated without revising this structure again. A future commit will update nfsstat(1) to report counts for the new NFSv4.2 specific operations/procedures. This patch affects the internal interface between the nfscommon, nfscl and nfsd modules and, as such, they all must be upgraded simultaneously. I will do a version bump (although arguably not needed), due to this. This code has survived a "make universe" but has not been built with a recent GCC. If you encounter build problems, please email me. Relnotes: yes
2019-12-12 23:22:55 +00:00
struct nfsclient *clp;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
struct nfsstate *stp;
struct nfslockfile *lfp;
int error, haslock = 0;
fhandle_t nfh;
Add support for NFSv4.2 to the NFS client and server. This patch adds support for NFSv4.2 (RFC-7862) and Extended Attributes (RFC-8276) to the NFS client and server. NFSv4.2 is comprised of several optional features that can be supported in addition to NFSv4.1. This patch adds the following optional features: - posix_fadvise(POSIX_FADV_WILLNEED/POSIX_FADV_DONTNEED) - posix_fallocate() - intra server file range copying via the copy_file_range(2) syscall --> Avoiding data tranfer over the wire to/from the NFS client. - lseek(SEEK_DATA/SEEK_HOLE) - Extended attribute syscalls for "user" namespace attributes as defined by RFC-8276. Although this patch is fairly large, it should not affect support for the other versions of NFS. However it does add two new sysctls that allow a sysadmin to limit which minor versions of NFSv4 a server supports, allowing a sysadmin to disable NFSv4.2. Unfortunately, when the NFS stats structure was last revised, it was assumed that there would be no additional operations added beyond what was specified in RFC-7862. However RFC-8276 did add additional operations, forcing the NFS stats structure to revised again. It now has extra unused entries in all arrays, so that future extensions to NFSv4.2 can be accomodated without revising this structure again. A future commit will update nfsstat(1) to report counts for the new NFSv4.2 specific operations/procedures. This patch affects the internal interface between the nfscommon, nfscl and nfsd modules and, as such, they all must be upgraded simultaneously. I will do a version bump (although arguably not needed), due to this. This code has survived a "make universe" but has not been built with a recent GCC. If you encounter build problems, please email me. Relnotes: yes
2019-12-12 23:22:55 +00:00
clp = NULL;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* First, get the lock file structure.
* (A return of -1 means no associated state, so remove ok.)
*/
error = nfsrv_getlockfh(vp, NFSLCK_CHECK, NULL, &nfh, p);
tryagain:
NFSLOCKSTATE();
Add support for NFSv4.2 to the NFS client and server. This patch adds support for NFSv4.2 (RFC-7862) and Extended Attributes (RFC-8276) to the NFS client and server. NFSv4.2 is comprised of several optional features that can be supported in addition to NFSv4.1. This patch adds the following optional features: - posix_fadvise(POSIX_FADV_WILLNEED/POSIX_FADV_DONTNEED) - posix_fallocate() - intra server file range copying via the copy_file_range(2) syscall --> Avoiding data tranfer over the wire to/from the NFS client. - lseek(SEEK_DATA/SEEK_HOLE) - Extended attribute syscalls for "user" namespace attributes as defined by RFC-8276. Although this patch is fairly large, it should not affect support for the other versions of NFS. However it does add two new sysctls that allow a sysadmin to limit which minor versions of NFSv4 a server supports, allowing a sysadmin to disable NFSv4.2. Unfortunately, when the NFS stats structure was last revised, it was assumed that there would be no additional operations added beyond what was specified in RFC-7862. However RFC-8276 did add additional operations, forcing the NFS stats structure to revised again. It now has extra unused entries in all arrays, so that future extensions to NFSv4.2 can be accomodated without revising this structure again. A future commit will update nfsstat(1) to report counts for the new NFSv4.2 specific operations/procedures. This patch affects the internal interface between the nfscommon, nfscl and nfsd modules and, as such, they all must be upgraded simultaneously. I will do a version bump (although arguably not needed), due to this. This code has survived a "make universe" but has not been built with a recent GCC. If you encounter build problems, please email me. Relnotes: yes
2019-12-12 23:22:55 +00:00
if (error == 0 && clientid.qval != 0)
error = nfsrv_getclient(clientid, CLOPS_RENEW, &clp, NULL,
(nfsquad_t)((u_quad_t)0), 0, nd, p);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (!error)
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
error = nfsrv_getlockfile(NFSLCK_CHECK, NULL, &lfp, &nfh, 0);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (error) {
NFSUNLOCKSTATE();
if (haslock) {
NFSLOCKV4ROOTMUTEX();
nfsv4_unlock(&nfsv4rootfs_lock, 1);
NFSUNLOCKV4ROOTMUTEX();
}
if (error == -1)
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
error = 0;
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* Now, we must Recall any delegations.
*/
Add support for NFSv4.2 to the NFS client and server. This patch adds support for NFSv4.2 (RFC-7862) and Extended Attributes (RFC-8276) to the NFS client and server. NFSv4.2 is comprised of several optional features that can be supported in addition to NFSv4.1. This patch adds the following optional features: - posix_fadvise(POSIX_FADV_WILLNEED/POSIX_FADV_DONTNEED) - posix_fallocate() - intra server file range copying via the copy_file_range(2) syscall --> Avoiding data tranfer over the wire to/from the NFS client. - lseek(SEEK_DATA/SEEK_HOLE) - Extended attribute syscalls for "user" namespace attributes as defined by RFC-8276. Although this patch is fairly large, it should not affect support for the other versions of NFS. However it does add two new sysctls that allow a sysadmin to limit which minor versions of NFSv4 a server supports, allowing a sysadmin to disable NFSv4.2. Unfortunately, when the NFS stats structure was last revised, it was assumed that there would be no additional operations added beyond what was specified in RFC-7862. However RFC-8276 did add additional operations, forcing the NFS stats structure to revised again. It now has extra unused entries in all arrays, so that future extensions to NFSv4.2 can be accomodated without revising this structure again. A future commit will update nfsstat(1) to report counts for the new NFSv4.2 specific operations/procedures. This patch affects the internal interface between the nfscommon, nfscl and nfsd modules and, as such, they all must be upgraded simultaneously. I will do a version bump (although arguably not needed), due to this. This code has survived a "make universe" but has not been built with a recent GCC. If you encounter build problems, please email me. Relnotes: yes
2019-12-12 23:22:55 +00:00
error = nfsrv_cleandeleg(vp, lfp, clp, &haslock, p);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (error) {
/*
* nfsrv_cleandeleg() unlocks state for non-zero
* return.
*/
if (error == -1)
goto tryagain;
if (haslock) {
NFSLOCKV4ROOTMUTEX();
nfsv4_unlock(&nfsv4rootfs_lock, 1);
NFSUNLOCKV4ROOTMUTEX();
}
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* Now, look for a conflicting open share.
*/
if (remove) {
Fix failures and warnings reported by newpynfs20090424 test tool. This fix addresses only issues with the pynfs reports, none of these issues are know to create problems for extant real clients. Submitted by: Bart Hsiao <bart.hsiao@gmail.com> Reworked by: myself Reviewed by: rmacklem Approved by: rmacklem Sponsored by: QNAP Systems Inc.
2014-10-03 02:24:41 +00:00
/*
* If the entry in the directory was the last reference to the
* corresponding filesystem object, the object can be destroyed
* */
if(lfp->lf_usecount>1)
LIST_FOREACH(stp, &lfp->lf_open, ls_file) {
if (stp->ls_flags & NFSLCK_WRITEDENY) {
error = NFSERR_FILEOPEN;
break;
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
}
NFSUNLOCKSTATE();
if (haslock) {
NFSLOCKV4ROOTMUTEX();
nfsv4_unlock(&nfsv4rootfs_lock, 1);
NFSUNLOCKV4ROOTMUTEX();
}
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
out:
NFSEXITCODE(error);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
return (error);
}
/*
* Clear out all delegations for the file referred to by lfp.
* May return NFSERR_DELAY, if there will be a delay waiting for
* delegations to expire.
* Returns -1 to indicate it slept while recalling a delegation.
* This function has the side effect of deleting the nfslockfile structure,
* if it no longer has associated state and didn't have to sleep.
* Unlocks State before a non-zero value is returned.
*/
static int
nfsrv_cleandeleg(vnode_t vp, struct nfslockfile *lfp,
struct nfsclient *clp, int *haslockp, NFSPROC_T *p)
{
struct nfsstate *stp, *nstp;
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
int ret = 0;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
stp = LIST_FIRST(&lfp->lf_deleg);
while (stp != LIST_END(&lfp->lf_deleg)) {
nstp = LIST_NEXT(stp, ls_file);
if (stp->ls_clp != clp) {
ret = nfsrv_delegconflict(stp, haslockp, p, vp);
if (ret) {
/*
* nfsrv_delegconflict() unlocks state
* when it returns non-zero.
*/
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
}
stp = nstp;
}
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
out:
NFSEXITCODE(ret);
return (ret);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* There are certain operations that, when being done outside of NFSv4,
* require that any NFSv4 delegation for the file be recalled.
* This function is to be called for those cases:
* VOP_RENAME() - When a delegation is being recalled for any reason,
* the client may have to do Opens against the server, using the file's
* final component name. If the file has been renamed on the server,
* that component name will be incorrect and the Open will fail.
* VOP_REMOVE() - Theoretically, a client could Open a file after it has
* been removed on the server, if there is a delegation issued to
* that client for the file. I say "theoretically" since clients
* normally do an Access Op before the Open and that Access Op will
* fail with ESTALE. Note that NFSv2 and 3 don't even do Opens, so
* they will detect the file's removal in the same manner. (There is
* one case where RFC3530 allows a client to do an Open without first
* doing an Access Op, which is passage of a check against the ACE
* returned with a Write delegation, but current practice is to ignore
* the ACE and always do an Access Op.)
* Since the functions can only be called with an unlocked vnode, this
* can't be done at this time.
* VOP_ADVLOCK() - When a client holds a delegation, it can issue byte range
* locks locally in the client, which are not visible to the server. To
* deal with this, issuing of delegations for a vnode must be disabled
* and all delegations for the vnode recalled. This is done via the
* second function, using the VV_DISABLEDELEG vflag on the vnode.
*/
nfs: Remove APPLESTATIC macro It is no longer useful. Reviewed by: rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24811
2020-05-12 13:23:25 +00:00
void
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsd_recalldelegation(vnode_t vp, NFSPROC_T *p)
{
Further cleanups to use of timestamps in NFS: - Use NFSD_MONOSEC (which maps to time_uptime) instead of the seconds portion of wall-time stamps to manage timeouts on events. - Remove unused nd_starttime from the per-request structure in the new NFS server. - Use nanotime() for the modification time on a delegation to get as precise a time as possible. - Use time_second instead of extracting the second from a call to getmicrotime(). Submitted by: bde (3) Reviewed by: bde, rmacklem MFC after: 2 weeks
2013-01-25 15:25:24 +00:00
time_t starttime;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
int error;
/*
* First, check to see if the server is currently running and it has
* been called for a regular file when issuing delegations.
*/
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
if (NFSD_VNET(nfsrv_numnfsd) == 0 || vp->v_type != VREG ||
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_issuedelegs == 0)
return;
Simple find/replace of VOP_ISLOCKED -> NFSVOPISLOCKED. This is done so that NFSVOPISLOCKED can be modified later to add enhanced logging and assertions. Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:05:41 +00:00
KASSERT((NFSVOPISLOCKED(vp) != LK_EXCLUSIVE), ("vp %p is locked", vp));
Add acquisition of a reference count on nfsv4root_lock to the nfsd_recalldelegation() function, since this function is called by nfsd threads when they are handling NFSv2 or NFSv3 RPCs, where no reference count would have been acquired. MFC after: 2 weeks
2010-08-28 23:50:09 +00:00
/*
* First, get a reference on the nfsv4rootfs_lock so that an
* exclusive lock cannot be acquired by another thread.
*/
NFSLOCKV4ROOTMUTEX();
Fix the new NFS client so that it handles NFSv4 state correctly during a forced dismount. This required that the exclusive and shared (refcnt) sleep lock functions check for MNTK_UMOUNTF before sleeping, so that they won't block while nfscl_umount() is getting rid of the state. As such, a "struct mount *" argument was added to the locking functions. I believe the only remaining case where a forced dismount can get hung in the kernel is when a thread is already attempting to do a TCP connect to a dead server when the krpc client structure called nr_client is NULL. This will only happen just after a "mount -u" with options that force a new TCP connection is done, so it shouldn't be a problem in practice. MFC after: 2 weeks
2011-05-27 22:05:10 +00:00
nfsv4_getref(&nfsv4rootfs_lock, NULL, NFSV4ROOTLOCKMUTEXPTR, NULL);
Add acquisition of a reference count on nfsv4root_lock to the nfsd_recalldelegation() function, since this function is called by nfsd threads when they are handling NFSv2 or NFSv3 RPCs, where no reference count would have been acquired. MFC after: 2 weeks
2010-08-28 23:50:09 +00:00
NFSUNLOCKV4ROOTMUTEX();
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* Now, call nfsrv_checkremove() in a loop while it returns
* NFSERR_DELAY. Return upon any other error or when timed out.
*/
Further cleanups to use of timestamps in NFS: - Use NFSD_MONOSEC (which maps to time_uptime) instead of the seconds portion of wall-time stamps to manage timeouts on events. - Remove unused nd_starttime from the per-request structure in the new NFS server. - Use nanotime() for the modification time on a delegation to get as precise a time as possible. - Use time_second instead of extracting the second from a call to getmicrotime(). Submitted by: bde (3) Reviewed by: bde, rmacklem MFC after: 2 weeks
2013-01-25 15:25:24 +00:00
starttime = NFSD_MONOSEC;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
do {
Simple find/replace of vn_lock -> NFSVOPLOCK. This is done so that NFSVOPLOCK can be modified later to add enhanced logging and assertions. Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:05:31 +00:00
if (NFSVOPLOCK(vp, LK_EXCLUSIVE) == 0) {
Add support for NFSv4.2 to the NFS client and server. This patch adds support for NFSv4.2 (RFC-7862) and Extended Attributes (RFC-8276) to the NFS client and server. NFSv4.2 is comprised of several optional features that can be supported in addition to NFSv4.1. This patch adds the following optional features: - posix_fadvise(POSIX_FADV_WILLNEED/POSIX_FADV_DONTNEED) - posix_fallocate() - intra server file range copying via the copy_file_range(2) syscall --> Avoiding data tranfer over the wire to/from the NFS client. - lseek(SEEK_DATA/SEEK_HOLE) - Extended attribute syscalls for "user" namespace attributes as defined by RFC-8276. Although this patch is fairly large, it should not affect support for the other versions of NFS. However it does add two new sysctls that allow a sysadmin to limit which minor versions of NFSv4 a server supports, allowing a sysadmin to disable NFSv4.2. Unfortunately, when the NFS stats structure was last revised, it was assumed that there would be no additional operations added beyond what was specified in RFC-7862. However RFC-8276 did add additional operations, forcing the NFS stats structure to revised again. It now has extra unused entries in all arrays, so that future extensions to NFSv4.2 can be accomodated without revising this structure again. A future commit will update nfsstat(1) to report counts for the new NFSv4.2 specific operations/procedures. This patch affects the internal interface between the nfscommon, nfscl and nfsd modules and, as such, they all must be upgraded simultaneously. I will do a version bump (although arguably not needed), due to this. This code has survived a "make universe" but has not been built with a recent GCC. If you encounter build problems, please email me. Relnotes: yes
2019-12-12 23:22:55 +00:00
error = nfsrv_checkremove(vp, 0, NULL,
(nfsquad_t)((u_quad_t)0), p);
vfs: drop the mostly unused flags argument from VOP_UNLOCK Filesystems which want to use it in limited capacity can employ the VOP_UNLOCK_FLAGS macro. Reviewed by: kib (previous version) Differential Revision: https://reviews.freebsd.org/D21427
2020-01-03 22:29:58 +00:00
NFSVOPUNLOCK(vp);
Add checks for VI_DOOMED and vn_lock() failures to the experimental NFS server, to handle the case where an exported file system is forced dismounted while an RPC is in progress. Further commits will fix the cases where a mount point is used when the associated vnode isn't locked. Reviewed by: kib MFC after: 2 weeks
2011-01-02 19:58:39 +00:00
} else
Fix two vnode locking problems in nfsd_recalldelegation() in the experimental NFSv4 server. The first was a bogus use of VOP_ISLOCKED() in a KASSERT() and the second was the need to lock the vnode for the nfsrv_checkremove() call. Also, delete a "__unused" that was bogus, since the argument is used. Reviewed by: zack.kirsch at isilon.com MFC after: 2 weeks
2010-12-17 22:18:09 +00:00
error = EPERM;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (error == NFSERR_DELAY) {
Further cleanups to use of timestamps in NFS: - Use NFSD_MONOSEC (which maps to time_uptime) instead of the seconds portion of wall-time stamps to manage timeouts on events. - Remove unused nd_starttime from the per-request structure in the new NFS server. - Use nanotime() for the modification time on a delegation to get as precise a time as possible. - Use time_second instead of extracting the second from a call to getmicrotime(). Submitted by: bde (3) Reviewed by: bde, rmacklem MFC after: 2 weeks
2013-01-25 15:25:24 +00:00
if (NFSD_MONOSEC - starttime > NFS_REMOVETIMEO)
Add acquisition of a reference count on nfsv4root_lock to the nfsd_recalldelegation() function, since this function is called by nfsd threads when they are handling NFSv2 or NFSv3 RPCs, where no reference count would have been acquired. MFC after: 2 weeks
2010-08-28 23:50:09 +00:00
break;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/* Sleep for a short period of time */
An NFSv4 server will reply NFSERR_GRACE for non-recovery RPCs during the grace period after startup. This grace period must be at least the lease duration, which is typically 1-2 minutes. It seems prudent for the experimental NFS client to wait a few seconds before retrying such an RPC, so that the server isn't flooded with non-recovery RPCs during recovery. This patch adds an argument to nfs_catnap() to implement a 5 second delay for this case. MFC after: 1 week
2010-04-24 22:52:14 +00:00
(void) nfs_catnap(PZERO, 0, "nfsremove");
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
} while (error == NFSERR_DELAY);
Add acquisition of a reference count on nfsv4root_lock to the nfsd_recalldelegation() function, since this function is called by nfsd threads when they are handling NFSv2 or NFSv3 RPCs, where no reference count would have been acquired. MFC after: 2 weeks
2010-08-28 23:50:09 +00:00
NFSLOCKV4ROOTMUTEX();
nfsv4_relref(&nfsv4rootfs_lock);
NFSUNLOCKV4ROOTMUTEX();
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
nfs: Remove APPLESTATIC macro It is no longer useful. Reviewed by: rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24811
2020-05-12 13:23:25 +00:00
void
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsd_disabledelegation(vnode_t vp, NFSPROC_T *p)
{
#ifdef VV_DISABLEDELEG
/*
* First, flag issuance of delegations disabled.
*/
atomic_set_long(&vp->v_vflag, VV_DISABLEDELEG);
#endif
/*
* Then call nfsd_recalldelegation() to get rid of all extant
* delegations.
*/
nfsd_recalldelegation(vp, p);
}
/*
* Check for conflicting locks, etc. and then get rid of delegations.
* (At one point I thought that I should get rid of delegations for any
* Setattr, since it could potentially disallow the I/O op (read or write)
* allowed by the delegation. However, Setattr Ops that aren't changing
* the size get a stateid of all 0s, so you can't tell if it is a delegation
* for the same client or a different one, so I decided to only get rid
* of delegations for other clients when the size is being changed.)
* In general, a Setattr can disable NFS I/O Ops that are outstanding, such
* as Write backs, even if there is no delegation, so it really isn't any
* different?)
*/
nfs: Remove APPLESTATIC macro It is no longer useful. Reviewed by: rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24811
2020-05-12 13:23:25 +00:00
int
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_checksetattr(vnode_t vp, struct nfsrv_descript *nd,
nfsv4stateid_t *stateidp, struct nfsvattr *nvap, nfsattrbit_t *attrbitp,
struct nfsexstuff *exp, NFSPROC_T *p)
{
struct nfsstate st, *stp = &st;
struct nfslock lo, *lop = &lo;
int error = 0;
nfsquad_t clientid;
if (NFSISSET_ATTRBIT(attrbitp, NFSATTRBIT_SIZE)) {
stp->ls_flags = (NFSLCK_CHECK | NFSLCK_WRITEACCESS);
lop->lo_first = nvap->na_size;
} else {
stp->ls_flags = 0;
lop->lo_first = 0;
}
if (NFSISSET_ATTRBIT(attrbitp, NFSATTRBIT_OWNER) ||
NFSISSET_ATTRBIT(attrbitp, NFSATTRBIT_OWNERGROUP) ||
NFSISSET_ATTRBIT(attrbitp, NFSATTRBIT_MODE) ||
NFSISSET_ATTRBIT(attrbitp, NFSATTRBIT_ACL))
stp->ls_flags |= NFSLCK_SETATTR;
if (stp->ls_flags == 0)
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
lop->lo_end = NFS64BITSSET;
lop->lo_flags = NFSLCK_WRITE;
stp->ls_ownerlen = 0;
stp->ls_op = NULL;
stp->ls_uid = nd->nd_cred->cr_uid;
stp->ls_stateid.seqid = stateidp->seqid;
clientid.lval[0] = stp->ls_stateid.other[0] = stateidp->other[0];
clientid.lval[1] = stp->ls_stateid.other[1] = stateidp->other[1];
stp->ls_stateid.other[2] = stateidp->other[2];
error = nfsrv_lockctrl(vp, &stp, &lop, NULL, clientid,
stateidp, exp, nd, p);
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
out:
NFSEXITCODE2(error, nd);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
return (error);
}
/*
* Check for a write delegation and do a CBGETATTR if there is one, updating
* the attributes, as required.
* Should I return an error if I can't get the attributes? (For now, I'll
* just return ok.
*/
nfs: Remove APPLESTATIC macro It is no longer useful. Reviewed by: rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24811
2020-05-12 13:23:25 +00:00
int
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_checkgetattr(struct nfsrv_descript *nd, vnode_t vp,
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
struct nfsvattr *nvap, nfsattrbit_t *attrbitp, NFSPROC_T *p)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
{
struct nfsstate *stp;
struct nfslockfile *lfp;
struct nfsclient *clp;
struct nfsvattr nva;
fhandle_t nfh;
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
int error = 0;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsattrbit_t cbbits;
u_quad_t delegfilerev;
NFSCBGETATTR_ATTRBIT(attrbitp, &cbbits);
if (!NFSNONZERO_ATTRBIT(&cbbits))
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
goto out;
Avoid the overhead of acquiring a lock in nfsrv_checkgetattr() when there are no write delegations issued. manu@ reported on the freebsd-current@ mailing list that there was a significant performance hit in nfsrv_checkgetattr() caused by the acquisition/release of a state lock, even when there were no write delegations issued. This patch add a count of outstanding issued write delegations to the NFSv4 server. This count allows nfsrv_checkgetattr() to return without acquiring any lock when the count is 0, avoiding the performance hit for the case where no write delegations are issued. Reported by: manu Reviewed by: kib MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D13327
2017-12-04 21:50:27 +00:00
if (nfsrv_writedelegcnt == 0)
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* Get the lock file structure.
* (A return of -1 means no associated state, so return ok.)
*/
error = nfsrv_getlockfh(vp, NFSLCK_CHECK, NULL, &nfh, p);
NFSLOCKSTATE();
if (!error)
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
error = nfsrv_getlockfile(NFSLCK_CHECK, NULL, &lfp, &nfh, 0);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (error) {
NFSUNLOCKSTATE();
if (error == -1)
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
error = 0;
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* Now, look for a write delegation.
*/
LIST_FOREACH(stp, &lfp->lf_deleg, ls_file) {
if (stp->ls_flags & NFSLCK_DELEGWRITE)
break;
}
if (stp == LIST_END(&lfp->lf_deleg)) {
NFSUNLOCKSTATE();
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
clp = stp->ls_clp;
Fix a case where the NFSv4.0 server might crash if delegations are enabled. asomers@ reported a crash on an NFSv4.0 server with a backtrace of: kdb_backtrace vpanic panic nfsrv_docallback nfsrv_checkgetattr nfsrvd_getattr nfsrvd_dorpc nfssvc_program svc_run_internal svc_thread_start fork_exit fork_trampoline where the panic message was "docallb", which indicates that a callback was attempted when the ClientID is unconfirmed. This would not normally occur, but it is possible to have an unconfirmed ClientID structure with delegation structure(s) chained off it if the client were to issue a SetClientID with the same "id" but different "verifier" after acquiring delegations on the previously confirmed ClientID. The bug appears to be that nfsrv_checkgetattr() failed to check for this uncommon case of an unconfirmed ClientID with a delegation structure that no longer refers to a delegation the client knows about. This patch adds a check for this case, handling it as if no delegation exists, which is the case when the above occurs. Although difficult to reproduce, this change should avoid the panic(). PR: 249127 Reported by: asomers Reviewed by: asomers MFC after: 1 week Differential Revision: https://reviews.freebbsd.org/D26342
2020-09-14 00:44:50 +00:00
/* If the clientid is not confirmed, ignore the delegation. */
if (clp->lc_flags & LCL_NEEDSCONFIRM) {
NFSUNLOCKSTATE();
goto out;
}
delegfilerev = stp->ls_filerev;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* If the Write delegation was issued as a part of this Compound RPC
* or if we have an Implied Clientid (used in a previous Op in this
* compound) and it is the client the delegation was issued to,
* just return ok.
* I also assume that it is from the same client iff the network
* host IP address is the same as the callback address. (Not
* exactly correct by the RFC, but avoids a lot of Getattr
* callbacks.)
*/
if (nd->nd_compref == stp->ls_compref ||
((nd->nd_flag & ND_IMPLIEDCLID) &&
clp->lc_clientid.qval == nd->nd_clientid.qval) ||
nfsaddr2_match(clp->lc_req.nr_nam, nd->nd_nam)) {
NFSUNLOCKSTATE();
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
goto out;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* We are now done with the delegation state structure,
* so the statelock can be released and we can now tsleep().
*/
/*
* Now, we must do the CB Getattr callback, to see if Change or Size
* has changed.
*/
if (clp->lc_expiry >= NFSD_MONOSEC) {
NFSUNLOCKSTATE();
NFSVNO_ATTRINIT(&nva);
nva.na_filerev = NFS64BITSSET;
error = nfsrv_docallback(clp, NFSV4OP_CBGETATTR, NULL,
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
0, &nfh, &nva, &cbbits, 0, p);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (!error) {
if ((nva.na_filerev != NFS64BITSSET &&
nva.na_filerev > delegfilerev) ||
(NFSVNO_ISSETSIZE(&nva) &&
nva.na_size != nvap->na_size)) {
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
error = nfsvno_updfilerev(vp, nvap, nd, p);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (NFSVNO_ISSETSIZE(&nva))
nvap->na_size = nva.na_size;
}
The NFSv4 server would call VOP_SETATTR() with a shared locked vnode when a Getattr for a file is done by a client other than the one that holds the file's delegation. This would only happen when delegations are enabled and the problem is fixed by this patch. MFC after: 1 week
2013-12-25 01:03:14 +00:00
} else
error = 0; /* Ignore callback errors for now. */
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
} else {
NFSUNLOCKSTATE();
}
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
out:
NFSEXITCODE2(error, nd);
return (error);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
/*
* This function looks for openowners that haven't had any opens for
* a while and throws them away. Called by an nfsd when NFSNSF_NOOPENS
* is set.
*/
nfs: Remove APPLESTATIC macro It is no longer useful. Reviewed by: rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24811
2020-05-12 13:23:25 +00:00
void
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_throwawayopens(NFSPROC_T *p)
{
struct nfsclient *clp, *nclp;
struct nfsstate *stp, *nstp;
int i;
NFSLOCKSTATE();
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
NFSD_VNET(nfsrv_stablefirst).nsf_flags &= ~NFSNSF_NOOPENS;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* For each client...
*/
Make the size of the hash tables used by the NFSv4 server tunable. No appreciable change in performance was observed after increasing the sizes of these tables and then testing with a single client. However, there was an email that indicated high CPU overheads for a heavily loaded NFSv4 and it is hoped that increasing the sizes of the hash tables via these tunables might help. The tables remain the same size by default. Differential Revision: https://reviews.freebsd.org/D2596 MFC after: 2 weeks
2015-05-27 22:00:05 +00:00
for (i = 0; i < nfsrv_clienthashsize; i++) {
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
LIST_FOREACH_SAFE(clp, &NFSD_VNET(nfsclienthash)[i], lc_hash,
nclp) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
LIST_FOREACH_SAFE(stp, &clp->lc_open, ls_list, nstp) {
if (LIST_EMPTY(&stp->ls_open) &&
(stp->ls_noopens > NFSNOOPEN ||
(nfsrv_openpluslock * 2) >
Make the size of the hash tables used by the NFSv4 server tunable. No appreciable change in performance was observed after increasing the sizes of these tables and then testing with a single client. However, there was an email that indicated high CPU overheads for a heavily loaded NFSv4 and it is hoped that increasing the sizes of the hash tables via these tunables might help. The tables remain the same size by default. Differential Revision: https://reviews.freebsd.org/D2596 MFC after: 2 weeks
2015-05-27 22:00:05 +00:00
nfsrv_v4statelimit))
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsrv_freeopenowner(stp, 0, p);
}
}
}
NFSUNLOCKSTATE();
}
/*
* This function checks to see if the credentials are the same.
nfsserver: Fix handling of SP4_NONE For NFSv4.1/4.2, when the client specifies SP4_NONE for state protection in the ExchangeID operation arguments, the server MUST allow the state management operations for any user credentials. (I misread the RFC and thought that SP4_NONE meant "at the server's discression" and not MUST be allowed.) This means that the "sec=XXX" field of the "V4:" exports(5) line only applies to NFSv4.0. This patch fixes the server to always allow state management operations for SP4_NONE, which is the only state management option currently supported. (I have patches that add support for SP4_MACH_CRED to the server. These will be in a future commit.) In practice, this bug does not seem to have caused interoperability problems. MFC after: 2 weeks
2023-01-15 22:07:40 +00:00
* The check for same credentials is needed for state management operations
nfsd: Add support for the SP4_MACH_CRED case in ExchangeID Commit f4179ad46fa4 added support for operation bitmaps for NFSv4.1/4.2. This commit uses those to implement the SP4_MACH_CRED case for the NFSv4.1/4.2 ExchangeID operation since the Linux NFSv4.1/4.2 client is now using this for Kerberized mounts. The Linux Kerberized NFSv4.1/4.2 mounts currently work without support for this because Linux will fall back to SP4_NONE, but there is no guarantee this fallback will work forever. This commit only affects Kerberized NFSv4.1/4.2 mounts from Linux at this time. MFC after: 3 months
2023-04-07 19:49:23 +00:00
* for NFSv4.0 or NFSv4.1/4.2 when SP4_MACH_CRED is configured via
* ExchangeID.
* Returns 1 for not same, 0 otherwise.
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
*/
static int
nfsd: Add support for the SP4_MACH_CRED case in ExchangeID Commit f4179ad46fa4 added support for operation bitmaps for NFSv4.1/4.2. This commit uses those to implement the SP4_MACH_CRED case for the NFSv4.1/4.2 ExchangeID operation since the Linux NFSv4.1/4.2 client is now using this for Kerberized mounts. The Linux Kerberized NFSv4.1/4.2 mounts currently work without support for this because Linux will fall back to SP4_NONE, but there is no guarantee this fallback will work forever. This commit only affects Kerberized NFSv4.1/4.2 mounts from Linux at this time. MFC after: 3 months
2023-04-07 19:49:23 +00:00
nfsrv_notsamecredname(int op, struct nfsrv_descript *nd, struct nfsclient *clp)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
{
nfsd: Add support for the SP4_MACH_CRED case in ExchangeID Commit f4179ad46fa4 added support for operation bitmaps for NFSv4.1/4.2. This commit uses those to implement the SP4_MACH_CRED case for the NFSv4.1/4.2 ExchangeID operation since the Linux NFSv4.1/4.2 client is now using this for Kerberized mounts. The Linux Kerberized NFSv4.1/4.2 mounts currently work without support for this because Linux will fall back to SP4_NONE, but there is no guarantee this fallback will work forever. This commit only affects Kerberized NFSv4.1/4.2 mounts from Linux at this time. MFC after: 3 months
2023-04-07 19:49:23 +00:00
/* Check for the SP4_MACH_CRED case. */
if (op != 0 && nfsrv_checkmachcred(op, nd, clp) != 0)
return (1);
nfsserver: Fix handling of SP4_NONE For NFSv4.1/4.2, when the client specifies SP4_NONE for state protection in the ExchangeID operation arguments, the server MUST allow the state management operations for any user credentials. (I misread the RFC and thought that SP4_NONE meant "at the server's discression" and not MUST be allowed.) This means that the "sec=XXX" field of the "V4:" exports(5) line only applies to NFSv4.0. This patch fixes the server to always allow state management operations for SP4_NONE, which is the only state management option currently supported. (I have patches that add support for SP4_MACH_CRED to the server. These will be in a future commit.) In practice, this bug does not seem to have caused interoperability problems. MFC after: 2 weeks
2023-01-15 22:07:40 +00:00
/* For NFSv4.1/4.2, SP4_NONE always allows this. */
if ((nd->nd_flag & ND_NFSV41) != 0)
return (0);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
if (nd->nd_flag & ND_GSS) {
if (!(clp->lc_flags & LCL_GSS))
return (1);
if (clp->lc_flags & LCL_NAME) {
if (nd->nd_princlen != clp->lc_namelen ||
NFSBCMP(nd->nd_principal, clp->lc_name,
clp->lc_namelen))
return (1);
else
return (0);
}
if (nd->nd_cred->cr_uid == clp->lc_uid)
return (0);
else
return (1);
} else if (clp->lc_flags & LCL_GSS)
return (1);
/*
* For AUTH_SYS, allow the same uid or root. (This is underspecified
* in RFC3530, which talks about principals, but doesn't say anything
* about uids for AUTH_SYS.)
*/
if (nd->nd_cred->cr_uid == clp->lc_uid || nd->nd_cred->cr_uid == 0)
return (0);
else
return (1);
}
/*
* Calculate the lease expiry time.
*/
static time_t
nfsrv_leaseexpiry(void)
{
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
if (NFSD_VNET(nfsrv_stablefirst).nsf_eograce > NFSD_MONOSEC)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
return (NFSD_MONOSEC + 2 * (nfsrv_lease + NFSRV_LEASEDELTA));
return (NFSD_MONOSEC + nfsrv_lease + NFSRV_LEASEDELTA);
}
/*
* Delay the delegation timeout as far as ls_delegtimelimit, as required.
*/
static void
nfsrv_delaydelegtimeout(struct nfsstate *stp)
{
if ((stp->ls_flags & NFSLCK_DELEGRECALL) == 0)
return;
if ((stp->ls_delegtime + 15) > NFSD_MONOSEC &&
stp->ls_delegtime < stp->ls_delegtimelimit) {
stp->ls_delegtime += nfsrv_lease;
if (stp->ls_delegtime > stp->ls_delegtimelimit)
stp->ls_delegtime = stp->ls_delegtimelimit;
}
}
/*
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
* This function checks to see if there is any other state associated
* with the openowner for this Open.
* It returns 1 if there is no other state, 0 otherwise.
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
*/
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
static int
nfsrv_nootherstate(struct nfsstate *stp)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
{
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
struct nfsstate *tstp;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
LIST_FOREACH(tstp, &stp->ls_openowner->ls_open, ls_list) {
if (tstp != stp || !LIST_EMPTY(&tstp->ls_lock))
return (0);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
return (1);
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
/*
* Create a list of lock deltas (changes to local byte range locking
* that can be rolled back using the list) and apply the changes via
* nfsvno_advlock(). Optionally, lock the list. It is expected that either
* the rollback or update function will be called after this.
* It returns an error (and rolls back, as required), if any nfsvno_advlock()
* call fails. If it returns an error, it will unlock the list.
*/
static int
nfsrv_locallock(vnode_t vp, struct nfslockfile *lfp, int flags,
uint64_t first, uint64_t end, struct nfslockconflict *cfp, NFSPROC_T *p)
{
struct nfslock *lop, *nlop;
int error = 0;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
/* Loop through the list of locks. */
lop = LIST_FIRST(&lfp->lf_locallock);
while (first < end && lop != NULL) {
nlop = LIST_NEXT(lop, lo_lckowner);
if (first >= lop->lo_end) {
/* not there yet */
lop = nlop;
} else if (first < lop->lo_first) {
/* new one starts before entry in list */
if (end <= lop->lo_first) {
/* no overlap between old and new */
error = nfsrv_dolocal(vp, lfp, flags,
NFSLCK_UNLOCK, first, end, cfp, p);
if (error != 0)
break;
first = end;
} else {
/* handle fragment overlapped with new one */
error = nfsrv_dolocal(vp, lfp, flags,
NFSLCK_UNLOCK, first, lop->lo_first, cfp,
p);
if (error != 0)
break;
first = lop->lo_first;
}
} else {
/* new one overlaps this entry in list */
if (end <= lop->lo_end) {
/* overlaps all of new one */
error = nfsrv_dolocal(vp, lfp, flags,
lop->lo_flags, first, end, cfp, p);
if (error != 0)
break;
first = end;
} else {
/* handle fragment overlapped with new one */
error = nfsrv_dolocal(vp, lfp, flags,
lop->lo_flags, first, lop->lo_end, cfp, p);
if (error != 0)
break;
first = lop->lo_end;
lop = nlop;
}
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
if (first < end && error == 0)
/* handle fragment past end of list */
error = nfsrv_dolocal(vp, lfp, flags, NFSLCK_UNLOCK, first,
end, cfp, p);
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
NFSEXITCODE(error);
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
return (error);
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
/*
* Local lock unlock. Unlock all byte ranges that are no longer locked
Try and make the nfsrv_localunlock() function in the experimental NFSv4 server more readable. Mostly changes to comments, but a case of >= is changed to >, since == can never happen. Also, I've added a couple of KASSERT()s and a slight optimization, since once the "else if" case happens, subsequent locks in the list can't have any effect. None of these changes fixes any known bug. MFC after: 2 weeks
2010-10-11 23:15:18 +00:00
* by NFSv4. To do this, unlock any subranges of first-->end that
* do not overlap with the byte ranges of any lock in the lfp->lf_lock
* list. This list has all locks for the file held by other
* <clientid, lockowner> tuples. The list is ordered by increasing
* lo_first value, but may have entries that overlap each other, for
* the case of read locks.
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
*/
static void
nfsrv_localunlock(vnode_t vp, struct nfslockfile *lfp, uint64_t init_first,
uint64_t init_end, NFSPROC_T *p)
{
struct nfslock *lop;
Silence newer gcc warnings. Newer versions of gcc generate "set, but not used" warnings in the NFS server. Add __unused macros to silence these warnings. Requested by: mmacy
2018-07-29 21:51:17 +00:00
uint64_t first, end, prevfirst __unused;
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
first = init_first;
end = init_end;
while (first < init_end) {
/* Loop through all nfs locks, adjusting first and end */
Try and make the nfsrv_localunlock() function in the experimental NFSv4 server more readable. Mostly changes to comments, but a case of >= is changed to >, since == can never happen. Also, I've added a couple of KASSERT()s and a slight optimization, since once the "else if" case happens, subsequent locks in the list can't have any effect. None of these changes fixes any known bug. MFC after: 2 weeks
2010-10-11 23:15:18 +00:00
prevfirst = 0;
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
LIST_FOREACH(lop, &lfp->lf_lock, lo_lckfile) {
Try and make the nfsrv_localunlock() function in the experimental NFSv4 server more readable. Mostly changes to comments, but a case of >= is changed to >, since == can never happen. Also, I've added a couple of KASSERT()s and a slight optimization, since once the "else if" case happens, subsequent locks in the list can't have any effect. None of these changes fixes any known bug. MFC after: 2 weeks
2010-10-11 23:15:18 +00:00
KASSERT(prevfirst <= lop->lo_first,
("nfsv4 locks out of order"));
KASSERT(lop->lo_first < lop->lo_end,
("nfsv4 bogus lock"));
prevfirst = lop->lo_first;
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
if (first >= lop->lo_first &&
first < lop->lo_end)
Try and make the nfsrv_localunlock() function in the experimental NFSv4 server more readable. Mostly changes to comments, but a case of >= is changed to >, since == can never happen. Also, I've added a couple of KASSERT()s and a slight optimization, since once the "else if" case happens, subsequent locks in the list can't have any effect. None of these changes fixes any known bug. MFC after: 2 weeks
2010-10-11 23:15:18 +00:00
/*
* Overlaps with initial part, so trim
* off that initial part by moving first past
* it.
*/
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
first = lop->lo_end;
else if (end > lop->lo_first &&
Try and make the nfsrv_localunlock() function in the experimental NFSv4 server more readable. Mostly changes to comments, but a case of >= is changed to >, since == can never happen. Also, I've added a couple of KASSERT()s and a slight optimization, since once the "else if" case happens, subsequent locks in the list can't have any effect. None of these changes fixes any known bug. MFC after: 2 weeks
2010-10-11 23:15:18 +00:00
lop->lo_first > first) {
/*
* This lock defines the end of the
* segment to unlock, so set end to the
* start of it and break out of the loop.
*/
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
end = lop->lo_first;
Try and make the nfsrv_localunlock() function in the experimental NFSv4 server more readable. Mostly changes to comments, but a case of >= is changed to >, since == can never happen. Also, I've added a couple of KASSERT()s and a slight optimization, since once the "else if" case happens, subsequent locks in the list can't have any effect. None of these changes fixes any known bug. MFC after: 2 weeks
2010-10-11 23:15:18 +00:00
break;
}
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
if (first >= end)
Try and make the nfsrv_localunlock() function in the experimental NFSv4 server more readable. Mostly changes to comments, but a case of >= is changed to >, since == can never happen. Also, I've added a couple of KASSERT()s and a slight optimization, since once the "else if" case happens, subsequent locks in the list can't have any effect. None of these changes fixes any known bug. MFC after: 2 weeks
2010-10-11 23:15:18 +00:00
/*
* There is no segment left to do, so
* break out of this loop and then exit
* the outer while() since first will be set
* to end, which must equal init_end here.
*/
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
break;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
if (first < end) {
/* Unlock this segment */
(void) nfsrv_dolocal(vp, lfp, NFSLCK_UNLOCK,
NFSLCK_READ, first, end, NULL, p);
nfsrv_locallock_commit(lfp, NFSLCK_UNLOCK,
first, end);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
Try and make the nfsrv_localunlock() function in the experimental NFSv4 server more readable. Mostly changes to comments, but a case of >= is changed to >, since == can never happen. Also, I've added a couple of KASSERT()s and a slight optimization, since once the "else if" case happens, subsequent locks in the list can't have any effect. None of these changes fixes any known bug. MFC after: 2 weeks
2010-10-11 23:15:18 +00:00
/*
* Now move past this segment and look for any further
* segment in the range, if there is one.
*/
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
first = end;
end = init_end;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
}
/*
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
* Do the local lock operation and update the rollback list, as required.
* Perform the rollback and return the error if nfsvno_advlock() fails.
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
*/
static int
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
nfsrv_dolocal(vnode_t vp, struct nfslockfile *lfp, int flags, int oldflags,
uint64_t first, uint64_t end, struct nfslockconflict *cfp, NFSPROC_T *p)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
{
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
struct nfsrollback *rlp;
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
int error = 0, ltype, oldltype;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
if (flags & NFSLCK_WRITE)
ltype = F_WRLCK;
else if (flags & NFSLCK_READ)
ltype = F_RDLCK;
else
ltype = F_UNLCK;
if (oldflags & NFSLCK_WRITE)
oldltype = F_WRLCK;
else if (oldflags & NFSLCK_READ)
oldltype = F_RDLCK;
else
oldltype = F_UNLCK;
if (ltype == oldltype || (oldltype == F_WRLCK && ltype == F_RDLCK))
/* nothing to do */
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
goto out;
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
error = nfsvno_advlock(vp, ltype, first, end, p);
if (error != 0) {
if (cfp != NULL) {
cfp->cl_clientid.lval[0] = 0;
cfp->cl_clientid.lval[1] = 0;
cfp->cl_first = 0;
cfp->cl_end = NFS64BITSSET;
cfp->cl_flags = NFSLCK_WRITE;
cfp->cl_ownerlen = 5;
NFSBCOPY("LOCAL", cfp->cl_owner, 5);
}
nfsrv_locallock_rollback(vp, lfp, p);
} else if (ltype != F_UNLCK) {
rlp = malloc(sizeof (struct nfsrollback), M_NFSDROLLBACK,
M_WAITOK);
rlp->rlck_first = first;
rlp->rlck_end = end;
rlp->rlck_type = oldltype;
LIST_INSERT_HEAD(&lfp->lf_rollback, rlp, rlck_list);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
Add DEXITCODE plumbing to NFS. Isilon has the concept of an in-memory exit-code ring that saves the last exit code of a function and allows for stack tracing. This is very helpful when debugging tough issues. This patch is essentially a no-op for BSD at this point, until we upstream the dexitcode logic itself. The patch adds DEXITCODE calls to every NFS function that returns an errno error code. A number of code paths were also reorganized to have single exit paths, to reduce code duplication. Submitted by: David Kwan <dkwan@isilon.com> Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:51:09 +00:00
out:
NFSEXITCODE(error);
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
return (error);
}
/*
* Roll back local lock changes and free up the rollback list.
*/
static void
nfsrv_locallock_rollback(vnode_t vp, struct nfslockfile *lfp, NFSPROC_T *p)
{
struct nfsrollback *rlp, *nrlp;
LIST_FOREACH_SAFE(rlp, &lfp->lf_rollback, rlck_list, nrlp) {
(void) nfsvno_advlock(vp, rlp->rlck_type, rlp->rlck_first,
rlp->rlck_end, p);
free(rlp, M_NFSDROLLBACK);
}
LIST_INIT(&lfp->lf_rollback);
}
/*
* Update local lock list and delete rollback list (ie now committed to the
* local locks). Most of the work is done by the internal function.
*/
static void
nfsrv_locallock_commit(struct nfslockfile *lfp, int flags, uint64_t first,
uint64_t end)
{
struct nfsrollback *rlp, *nrlp;
struct nfslock *new_lop, *other_lop;
new_lop = malloc(sizeof (struct nfslock), M_NFSDLOCK, M_WAITOK);
if (flags & (NFSLCK_READ | NFSLCK_WRITE))
other_lop = malloc(sizeof (struct nfslock), M_NFSDLOCK,
M_WAITOK);
else
other_lop = NULL;
new_lop->lo_flags = flags;
new_lop->lo_first = first;
new_lop->lo_end = end;
nfsrv_updatelock(NULL, &new_lop, &other_lop, lfp);
if (new_lop != NULL)
free(new_lop, M_NFSDLOCK);
if (other_lop != NULL)
free(other_lop, M_NFSDLOCK);
/* and get rid of the rollback list */
LIST_FOREACH_SAFE(rlp, &lfp->lf_rollback, rlck_list, nrlp)
free(rlp, M_NFSDROLLBACK);
LIST_INIT(&lfp->lf_rollback);
}
/*
* Lock the struct nfslockfile for local lock updating.
*/
static void
nfsrv_locklf(struct nfslockfile *lfp)
{
int gotlock;
/* lf_usecount ensures *lfp won't be free'd */
lfp->lf_usecount++;
do {
gotlock = nfsv4_lock(&lfp->lf_locallock_lck, 1, NULL,
Fix the new NFS client so that it handles NFSv4 state correctly during a forced dismount. This required that the exclusive and shared (refcnt) sleep lock functions check for MNTK_UMOUNTF before sleeping, so that they won't block while nfscl_umount() is getting rid of the state. As such, a "struct mount *" argument was added to the locking functions. I believe the only remaining case where a forced dismount can get hung in the kernel is when a thread is already attempting to do a TCP connect to a dead server when the krpc client structure called nr_client is NULL. This will only happen just after a "mount -u" with options that force a new TCP connection is done, so it shouldn't be a problem in practice. MFC after: 2 weeks
2011-05-27 22:05:10 +00:00
NFSSTATEMUTEXPTR, NULL);
This patch should fix handling of byte range locks locally on the server for the experimental nfs server. When enabled by setting vfs.newnfs.locallocks_enable to non-zero, the experimental nfs server will now acquire byte range locks on the file on behalf of NFSv4 clients, such that lock conflicts between the NFSv4 clients and processes running locally on the server, will be recognized and handled correctly. MFC after: 2 weeks
2010-03-30 23:11:50 +00:00
} while (gotlock == 0);
lfp->lf_usecount--;
}
/*
* Unlock the struct nfslockfile after local lock updating.
*/
static void
nfsrv_unlocklf(struct nfslockfile *lfp)
{
nfsv4_unlock(&lfp->lf_locallock_lck, 0);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
Add some cleanup code to the module unload operation for the experimental NFS server, so that it doesn't leak memory when unloaded. However, unloading the NFSv4 server is not recommended, since all NFSv4 state will be lost by the unload and clients will have to recover the state after a server reload/restart as if the server crashed/rebooted. MFC after: 2 weeks
2011-04-10 20:43:07 +00:00
/*
* Clear out all state for the NFSv4 server.
* Must be called by a thread that can sleep when no nfsds are running.
*/
void
nfsrv_throwawayallstate(NFSPROC_T *p)
{
struct nfsclient *clp, *nclp;
struct nfslockfile *lfp, *nlfp;
int i;
/*
* For each client, clean out the state and then free the structure.
*/
Make the size of the hash tables used by the NFSv4 server tunable. No appreciable change in performance was observed after increasing the sizes of these tables and then testing with a single client. However, there was an email that indicated high CPU overheads for a heavily loaded NFSv4 and it is hoped that increasing the sizes of the hash tables via these tunables might help. The tables remain the same size by default. Differential Revision: https://reviews.freebsd.org/D2596 MFC after: 2 weeks
2015-05-27 22:00:05 +00:00
for (i = 0; i < nfsrv_clienthashsize; i++) {
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
LIST_FOREACH_SAFE(clp, &NFSD_VNET(nfsclienthash)[i], lc_hash,
nclp) {
Add some cleanup code to the module unload operation for the experimental NFS server, so that it doesn't leak memory when unloaded. However, unloading the NFSv4 server is not recommended, since all NFSv4 state will be lost by the unload and clients will have to recover the state after a server reload/restart as if the server crashed/rebooted. MFC after: 2 weeks
2011-04-10 20:43:07 +00:00
nfsrv_cleanclient(clp, p);
nfsrv_freedeleglist(&clp->lc_deleg);
nfsrv_freedeleglist(&clp->lc_olddeleg);
Make the size of the hash tables used by the NFSv4 server tunable. No appreciable change in performance was observed after increasing the sizes of these tables and then testing with a single client. However, there was an email that indicated high CPU overheads for a heavily loaded NFSv4 and it is hoped that increasing the sizes of the hash tables via these tunables might help. The tables remain the same size by default. Differential Revision: https://reviews.freebsd.org/D2596 MFC after: 2 weeks
2015-05-27 22:00:05 +00:00
free(clp->lc_stateid, M_NFSDCLIENT);
Add some cleanup code to the module unload operation for the experimental NFS server, so that it doesn't leak memory when unloaded. However, unloading the NFSv4 server is not recommended, since all NFSv4 state will be lost by the unload and clients will have to recover the state after a server reload/restart as if the server crashed/rebooted. MFC after: 2 weeks
2011-04-10 20:43:07 +00:00
free(clp, M_NFSDCLIENT);
}
}
/*
* Also, free up any remaining lock file structures.
*/
Make the size of the hash tables used by the NFSv4 server tunable. No appreciable change in performance was observed after increasing the sizes of these tables and then testing with a single client. However, there was an email that indicated high CPU overheads for a heavily loaded NFSv4 and it is hoped that increasing the sizes of the hash tables via these tunables might help. The tables remain the same size by default. Differential Revision: https://reviews.freebsd.org/D2596 MFC after: 2 weeks
2015-05-27 22:00:05 +00:00
for (i = 0; i < nfsrv_lockhashsize; i++) {
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
LIST_FOREACH_SAFE(lfp, &NFSD_VNET(nfslockhash)[i], lf_hash,
nlfp) {
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
printf("nfsd unload: fnd a lock file struct\n");
Add some cleanup code to the module unload operation for the experimental NFS server, so that it doesn't leak memory when unloaded. However, unloading the NFSv4 server is not recommended, since all NFSv4 state will be lost by the unload and clients will have to recover the state after a server reload/restart as if the server crashed/rebooted. MFC after: 2 weeks
2011-04-10 20:43:07 +00:00
nfsrv_freenfslockfile(lfp);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
}
Add some cleanup code to the module unload operation for the experimental NFS server, so that it doesn't leak memory when unloaded. However, unloading the NFSv4 server is not recommended, since all NFSv4 state will be lost by the unload and clients will have to recover the state after a server reload/restart as if the server crashed/rebooted. MFC after: 2 weeks
2011-04-10 20:43:07 +00:00
}
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
/* And get rid of the deviceid structures and layouts. */
nfsrv_freealllayoutsanddevids();
Add some cleanup code to the module unload operation for the experimental NFS server, so that it doesn't leak memory when unloaded. However, unloading the NFSv4 server is not recommended, since all NFSv4 state will be lost by the unload and clients will have to recover the state after a server reload/restart as if the server crashed/rebooted. MFC after: 2 weeks
2011-04-10 20:43:07 +00:00
}
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
/*
* Check the sequence# for the session and slot provided as an argument.
* Also, renew the lease if the session will return NFS_OK.
*/
int
nfsrv_checksequence(struct nfsrv_descript *nd, uint32_t sequenceid,
uint32_t *highest_slotidp, uint32_t *target_highest_slotidp, int cache_this,
uint32_t *sflagsp, NFSPROC_T *p)
{
struct nfsdsession *sep;
struct nfssessionhash *shp;
int error;
shp = NFSSESSIONHASH(nd->nd_sessionid);
NFSLOCKSESSION(shp);
sep = nfsrv_findsession(nd->nd_sessionid);
if (sep == NULL) {
NFSUNLOCKSESSION(shp);
return (NFSERR_BADSESSION);
}
error = nfsv4_seqsession(sequenceid, nd->nd_slotid, *highest_slotidp,
sep->sess_slots, NULL, NFSV4_SLOTS - 1);
if (error != 0) {
NFSUNLOCKSESSION(shp);
return (error);
}
if (cache_this != 0)
nd->nd_flag |= ND_SAVEREPLY;
/* Renew the lease. */
sep->sess_clp->lc_expiry = nfsrv_leaseexpiry();
nd->nd_clientid.qval = sep->sess_clp->lc_clientid.qval;
nd->nd_flag |= ND_IMPLIEDCLID;
nfsd: Add support for the SP4_MACH_CRED case in ExchangeID Commit f4179ad46fa4 added support for operation bitmaps for NFSv4.1/4.2. This commit uses those to implement the SP4_MACH_CRED case for the NFSv4.1/4.2 ExchangeID operation since the Linux NFSv4.1/4.2 client is now using this for Kerberized mounts. The Linux Kerberized NFSv4.1/4.2 mounts currently work without support for this because Linux will fall back to SP4_NONE, but there is no guarantee this fallback will work forever. This commit only affects Kerberized NFSv4.1/4.2 mounts from Linux at this time. MFC after: 3 months
2023-04-07 19:49:23 +00:00
/* Handle the SP4_MECH_CRED case for NFSv4.1/4.2. */
if ((sep->sess_clp->lc_flags & LCL_MACHCRED) != 0 &&
(nd->nd_flag & (ND_GSSINTEGRITY | ND_GSSPRIVACY)) != 0 &&
nd->nd_princlen == sep->sess_clp->lc_namelen &&
!NFSBCMP(sep->sess_clp->lc_name, nd->nd_principal,
nd->nd_princlen)) {
nd->nd_flag |= ND_MACHCRED;
NFSSET_OPBIT(&nd->nd_allowops, &sep->sess_clp->lc_allowops);
}
Add support for NFSv4.2 to the NFS client and server. This patch adds support for NFSv4.2 (RFC-7862) and Extended Attributes (RFC-8276) to the NFS client and server. NFSv4.2 is comprised of several optional features that can be supported in addition to NFSv4.1. This patch adds the following optional features: - posix_fadvise(POSIX_FADV_WILLNEED/POSIX_FADV_DONTNEED) - posix_fallocate() - intra server file range copying via the copy_file_range(2) syscall --> Avoiding data tranfer over the wire to/from the NFS client. - lseek(SEEK_DATA/SEEK_HOLE) - Extended attribute syscalls for "user" namespace attributes as defined by RFC-8276. Although this patch is fairly large, it should not affect support for the other versions of NFS. However it does add two new sysctls that allow a sysadmin to limit which minor versions of NFSv4 a server supports, allowing a sysadmin to disable NFSv4.2. Unfortunately, when the NFS stats structure was last revised, it was assumed that there would be no additional operations added beyond what was specified in RFC-7862. However RFC-8276 did add additional operations, forcing the NFS stats structure to revised again. It now has extra unused entries in all arrays, so that future extensions to NFSv4.2 can be accomodated without revising this structure again. A future commit will update nfsstat(1) to report counts for the new NFSv4.2 specific operations/procedures. This patch affects the internal interface between the nfscommon, nfscl and nfsd modules and, as such, they all must be upgraded simultaneously. I will do a version bump (although arguably not needed), due to this. This code has survived a "make universe" but has not been built with a recent GCC. If you encounter build problems, please email me. Relnotes: yes
2019-12-12 23:22:55 +00:00
/* Save maximum request and reply sizes. */
nd->nd_maxreq = sep->sess_maxreq;
nd->nd_maxresp = sep->sess_maxresp;
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
*sflagsp = 0;
nfsd: do not implicitly bind the back channel for NFSv4.1/4.2 mounts The NFSv4.1 (and 4.2 on 13) server incorrectly binds a new TCP connection to the back channel when first used by an RPC with a Sequence op in it (almost all of them). RFC5661 specifies that only the fore channel should be bound. This was done because early clients (including FreeBSD) did not do the required BindConnectionToSession RPC. Unfortunately, this breaks the Linux client when the "nconnects" mount option is used, since the server may do a callback on the incorrect TCP connection. This patch converts the server behaviour to that required by the RFC. It also makes the server test/indicate failure of the back channel more aggressively. Until this patch is applied to the server, the "nconnects" mount option is not recommended for a Linux NFSv4.1/4.2 client mount to the FreeBSD server. Reported by: bcodding@redhat.com Tested by: bcodding@redhat.com PR: 254560 MFC after: 1 week
2021-03-30 21:31:05 +00:00
if (sep->sess_clp->lc_req.nr_client == NULL ||
(sep->sess_clp->lc_flags & LCL_CBDOWN) != 0)
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
*sflagsp |= NFSV4SEQ_CBPATHDOWN;
NFSUNLOCKSESSION(shp);
if (error == NFSERR_EXPIRED) {
*sflagsp |= NFSV4SEQ_EXPIREDALLSTATEREVOKED;
error = 0;
} else if (error == NFSERR_ADMINREVOKED) {
*sflagsp |= NFSV4SEQ_ADMINSTATEREVOKED;
error = 0;
}
*highest_slotidp = *target_highest_slotidp = NFSV4_SLOTS - 1;
return (0);
}
/*
* Check/set reclaim complete for this session/clientid.
*/
int
Modify the NFSv4.1 server so that it allows ReclaimComplete as done by ESXi 6.7. I believe that a ReclaimComplete with rca_one_fs == TRUE is only to be used after a file system has been transferred to a different file server. However, RFC5661 is somewhat vague w.r.t. this and the ESXi 6.7 client does both a ReclaimComplete with rca_one_fs == TRUE and one with ReclaimComplete with rca_one_fs == FALSE. Therefore, just ignore the rca_one_fs == TRUE operation and return NFS_OK without doing anything instead of replying NFS4ERR_NOTSUPP. This allows the ESXi 6.7 NFSv4.1 client to do a mount. After discussion on the NFSv4 IETF working group mailing list, doing this along with setting a flag to note that a ReclaimComplete with rca_one_fs TRUE was an appropriate way to handle this. The flag that indicates that a ReclaimComplete with rca_one_fs == TRUE was done may be used to disable replies of NFS4ERR_GRACE for non-reclaim state operations in a future commit. This patch along with r332790, r334492 and r336357 allow ESXi 6.7 NFSv4.1 mounts work ok. ESX 6.5 NFSv4.1 mounts do not work well, due to what I believe are violations of RFC-5661 and should not be used. Reported by: andreas.nagy@frequentis.com Tested by: andreas.nagy@frequentis.com, daniel@ftml.net (earlier version) MFC after: 2 weeks Relnotes: yes
2018-07-28 20:21:04 +00:00
nfsrv_checkreclaimcomplete(struct nfsrv_descript *nd, int onefs)
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
{
struct nfsdsession *sep;
struct nfssessionhash *shp;
int error = 0;
shp = NFSSESSIONHASH(nd->nd_sessionid);
NFSLOCKSTATE();
NFSLOCKSESSION(shp);
sep = nfsrv_findsession(nd->nd_sessionid);
if (sep == NULL) {
NFSUNLOCKSESSION(shp);
NFSUNLOCKSTATE();
return (NFSERR_BADSESSION);
}
Modify the NFSv4.1 server so that it allows ReclaimComplete as done by ESXi 6.7. I believe that a ReclaimComplete with rca_one_fs == TRUE is only to be used after a file system has been transferred to a different file server. However, RFC5661 is somewhat vague w.r.t. this and the ESXi 6.7 client does both a ReclaimComplete with rca_one_fs == TRUE and one with ReclaimComplete with rca_one_fs == FALSE. Therefore, just ignore the rca_one_fs == TRUE operation and return NFS_OK without doing anything instead of replying NFS4ERR_NOTSUPP. This allows the ESXi 6.7 NFSv4.1 client to do a mount. After discussion on the NFSv4 IETF working group mailing list, doing this along with setting a flag to note that a ReclaimComplete with rca_one_fs TRUE was an appropriate way to handle this. The flag that indicates that a ReclaimComplete with rca_one_fs == TRUE was done may be used to disable replies of NFS4ERR_GRACE for non-reclaim state operations in a future commit. This patch along with r332790, r334492 and r336357 allow ESXi 6.7 NFSv4.1 mounts work ok. ESX 6.5 NFSv4.1 mounts do not work well, due to what I believe are violations of RFC-5661 and should not be used. Reported by: andreas.nagy@frequentis.com Tested by: andreas.nagy@frequentis.com, daniel@ftml.net (earlier version) MFC after: 2 weeks Relnotes: yes
2018-07-28 20:21:04 +00:00
if (onefs != 0)
sep->sess_clp->lc_flags |= LCL_RECLAIMONEFS;
/* Check to see if reclaim complete has already happened. */
else if ((sep->sess_clp->lc_flags & LCL_RECLAIMCOMPLETE) != 0)
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
error = NFSERR_COMPLETEALREADY;
End grace for the NFSv4 server if all mounts do ReclaimComplete. The NFSv4 protocol requires that the server only allow reclaim of state and not issue any new open/lock state for a grace period after booting. The NFSv4.0 protocol required this grace period to be greater than the lease duration (over 2minutes). For NFSv4.1, the client tells the server that it has done reclaiming state by doing a ReclaimComplete operation. If all NFSv4 clients are NFSv4.1, the grace period can end once all the clients have done ReclaimComplete, shortening the time period considerably. This patch does this. If there are any NFSv4.0 mounts, the grace period will still be over 2minutes. This change is only an optimization and does not affect correct operation. Tested by: andreas.nagy@frequentis.com MFC after: 2 months
2018-05-15 20:28:50 +00:00
else {
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
sep->sess_clp->lc_flags |= LCL_RECLAIMCOMPLETE;
End grace for the NFSv4 server if all mounts do ReclaimComplete. The NFSv4 protocol requires that the server only allow reclaim of state and not issue any new open/lock state for a grace period after booting. The NFSv4.0 protocol required this grace period to be greater than the lease duration (over 2minutes). For NFSv4.1, the client tells the server that it has done reclaiming state by doing a ReclaimComplete operation. If all NFSv4 clients are NFSv4.1, the grace period can end once all the clients have done ReclaimComplete, shortening the time period considerably. This patch does this. If there are any NFSv4.0 mounts, the grace period will still be over 2minutes. This change is only an optimization and does not affect correct operation. Tested by: andreas.nagy@frequentis.com MFC after: 2 months
2018-05-15 20:28:50 +00:00
nfsrv_markreclaim(sep->sess_clp);
}
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
NFSUNLOCKSESSION(shp);
NFSUNLOCKSTATE();
return (error);
}
/*
* Cache the reply in a session slot.
*/
void
Fix noisy NFSv4 server printf. Peter reported that his dmesg was getting cluttered with nfsrv_cache_session: no session messages when he rebooted his NFS server and they did not seem useful. He was correct, in that these messages are "normal" and expected when NFSv4.1 or NFSv4.2 are mounted and the server is rebooted. This patch silences the printf() during the grace period after a reboot. It also adds the client IP address to the printf(), so that the message is more useful if/when it occurs. If this happens outside of the server's grace period, it does indicate something is not working correctly. Instead of adding yet another nd_XXX argument, the arguments for nfsrv_cache_session() were simplified to take a "struct nfsrv_descript *". Reported by: pen@lysator.liu.se MFC after: 2 weeks
2020-04-06 23:21:39 +00:00
nfsrv_cache_session(struct nfsrv_descript *nd, struct mbuf **m)
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
{
struct nfsdsession *sep;
struct nfssessionhash *shp;
Fix noisy NFSv4 server printf. Peter reported that his dmesg was getting cluttered with nfsrv_cache_session: no session messages when he rebooted his NFS server and they did not seem useful. He was correct, in that these messages are "normal" and expected when NFSv4.1 or NFSv4.2 are mounted and the server is rebooted. This patch silences the printf() during the grace period after a reboot. It also adds the client IP address to the printf(), so that the message is more useful if/when it occurs. If this happens outside of the server's grace period, it does indicate something is not working correctly. Instead of adding yet another nd_XXX argument, the arguments for nfsrv_cache_session() were simplified to take a "struct nfsrv_descript *". Reported by: pen@lysator.liu.se MFC after: 2 weeks
2020-04-06 23:21:39 +00:00
char *buf, *cp;
#ifdef INET
struct sockaddr_in *sin;
#endif
#ifdef INET6
struct sockaddr_in6 *sin6;
#endif
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
Fix noisy NFSv4 server printf. Peter reported that his dmesg was getting cluttered with nfsrv_cache_session: no session messages when he rebooted his NFS server and they did not seem useful. He was correct, in that these messages are "normal" and expected when NFSv4.1 or NFSv4.2 are mounted and the server is rebooted. This patch silences the printf() during the grace period after a reboot. It also adds the client IP address to the printf(), so that the message is more useful if/when it occurs. If this happens outside of the server's grace period, it does indicate something is not working correctly. Instead of adding yet another nd_XXX argument, the arguments for nfsrv_cache_session() were simplified to take a "struct nfsrv_descript *". Reported by: pen@lysator.liu.se MFC after: 2 weeks
2020-04-06 23:21:39 +00:00
shp = NFSSESSIONHASH(nd->nd_sessionid);
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
NFSLOCKSESSION(shp);
Fix noisy NFSv4 server printf. Peter reported that his dmesg was getting cluttered with nfsrv_cache_session: no session messages when he rebooted his NFS server and they did not seem useful. He was correct, in that these messages are "normal" and expected when NFSv4.1 or NFSv4.2 are mounted and the server is rebooted. This patch silences the printf() during the grace period after a reboot. It also adds the client IP address to the printf(), so that the message is more useful if/when it occurs. If this happens outside of the server's grace period, it does indicate something is not working correctly. Instead of adding yet another nd_XXX argument, the arguments for nfsrv_cache_session() were simplified to take a "struct nfsrv_descript *". Reported by: pen@lysator.liu.se MFC after: 2 weeks
2020-04-06 23:21:39 +00:00
sep = nfsrv_findsession(nd->nd_sessionid);
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
if (sep == NULL) {
NFSUNLOCKSESSION(shp);
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
if ((NFSD_VNET(nfsrv_stablefirst).nsf_flags &
NFSNSF_GRACEOVER) != 0) {
Fix noisy NFSv4 server printf. Peter reported that his dmesg was getting cluttered with nfsrv_cache_session: no session messages when he rebooted his NFS server and they did not seem useful. He was correct, in that these messages are "normal" and expected when NFSv4.1 or NFSv4.2 are mounted and the server is rebooted. This patch silences the printf() during the grace period after a reboot. It also adds the client IP address to the printf(), so that the message is more useful if/when it occurs. If this happens outside of the server's grace period, it does indicate something is not working correctly. Instead of adding yet another nd_XXX argument, the arguments for nfsrv_cache_session() were simplified to take a "struct nfsrv_descript *". Reported by: pen@lysator.liu.se MFC after: 2 weeks
2020-04-06 23:21:39 +00:00
buf = malloc(INET6_ADDRSTRLEN, M_TEMP, M_WAITOK);
switch (nd->nd_nam->sa_family) {
#ifdef INET
case AF_INET:
sin = (struct sockaddr_in *)nd->nd_nam;
cp = inet_ntop(sin->sin_family,
&sin->sin_addr.s_addr, buf,
INET6_ADDRSTRLEN);
break;
#endif
#ifdef INET6
case AF_INET6:
sin6 = (struct sockaddr_in6 *)nd->nd_nam;
cp = inet_ntop(sin6->sin6_family,
&sin6->sin6_addr, buf, INET6_ADDRSTRLEN);
break;
#endif
default:
cp = NULL;
}
if (cp != NULL)
printf("nfsrv_cache_session: no session "
nfsd: Update console message for no session found The NFSv4.1/4.2 server generates a console message that indicates that there is no session. I was until recently perplexed w.r.t. how this could occur. It turns out that the common cause is multiple NFS clients with the same /etc/hostid. The host uuid is used by the FreeBSD NFSv4.1/4.2 client as a unique identifier for the client. If multiple clients use the same host uuid, this indicates to the NFSv4.1/4.2 server that they are the same client and confusion occurs. This trivial patch modifies the console message to suggest that the client's /etc/hostid needs to be checked for uniqueness. Reviewed by: asomers MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D36377
2022-08-28 01:31:20 +00:00
"IPaddr=%s, check NFS clients for unique "
"/etc/hostid's\n", cp);
Fix noisy NFSv4 server printf. Peter reported that his dmesg was getting cluttered with nfsrv_cache_session: no session messages when he rebooted his NFS server and they did not seem useful. He was correct, in that these messages are "normal" and expected when NFSv4.1 or NFSv4.2 are mounted and the server is rebooted. This patch silences the printf() during the grace period after a reboot. It also adds the client IP address to the printf(), so that the message is more useful if/when it occurs. If this happens outside of the server's grace period, it does indicate something is not working correctly. Instead of adding yet another nd_XXX argument, the arguments for nfsrv_cache_session() were simplified to take a "struct nfsrv_descript *". Reported by: pen@lysator.liu.se MFC after: 2 weeks
2020-04-06 23:21:39 +00:00
else
nfsd: Update console message for no session found The NFSv4.1/4.2 server generates a console message that indicates that there is no session. I was until recently perplexed w.r.t. how this could occur. It turns out that the common cause is multiple NFS clients with the same /etc/hostid. The host uuid is used by the FreeBSD NFSv4.1/4.2 client as a unique identifier for the client. If multiple clients use the same host uuid, this indicates to the NFSv4.1/4.2 server that they are the same client and confusion occurs. This trivial patch modifies the console message to suggest that the client's /etc/hostid needs to be checked for uniqueness. Reviewed by: asomers MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D36377
2022-08-28 01:31:20 +00:00
printf("nfsrv_cache_session: no session, "
"check NFS clients for unique "
"/etc/hostid's\n");
Fix noisy NFSv4 server printf. Peter reported that his dmesg was getting cluttered with nfsrv_cache_session: no session messages when he rebooted his NFS server and they did not seem useful. He was correct, in that these messages are "normal" and expected when NFSv4.1 or NFSv4.2 are mounted and the server is rebooted. This patch silences the printf() during the grace period after a reboot. It also adds the client IP address to the printf(), so that the message is more useful if/when it occurs. If this happens outside of the server's grace period, it does indicate something is not working correctly. Instead of adding yet another nd_XXX argument, the arguments for nfsrv_cache_session() were simplified to take a "struct nfsrv_descript *". Reported by: pen@lysator.liu.se MFC after: 2 weeks
2020-04-06 23:21:39 +00:00
free(buf, M_TEMP);
}
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
m_freem(*m);
return;
}
Fix noisy NFSv4 server printf. Peter reported that his dmesg was getting cluttered with nfsrv_cache_session: no session messages when he rebooted his NFS server and they did not seem useful. He was correct, in that these messages are "normal" and expected when NFSv4.1 or NFSv4.2 are mounted and the server is rebooted. This patch silences the printf() during the grace period after a reboot. It also adds the client IP address to the printf(), so that the message is more useful if/when it occurs. If this happens outside of the server's grace period, it does indicate something is not working correctly. Instead of adding yet another nd_XXX argument, the arguments for nfsrv_cache_session() were simplified to take a "struct nfsrv_descript *". Reported by: pen@lysator.liu.se MFC after: 2 weeks
2020-04-06 23:21:39 +00:00
nfsv4_seqsess_cacherep(nd->nd_slotid, sep->sess_slots, nd->nd_repstat,
m);
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
NFSUNLOCKSESSION(shp);
}
/*
* Search for a session that matches the sessionid.
*/
static struct nfsdsession *
nfsrv_findsession(uint8_t *sessionid)
{
struct nfsdsession *sep;
struct nfssessionhash *shp;
shp = NFSSESSIONHASH(sessionid);
LIST_FOREACH(sep, &shp->list, sess_hash) {
if (!NFSBCMP(sessionid, sep->sess_sessionid, NFSX_V4SESSIONID))
break;
}
return (sep);
}
/*
* Destroy a session.
*/
int
nfsrv_destroysession(struct nfsrv_descript *nd, uint8_t *sessionid)
{
Strengthen locking for the NFSv4.1 server DestroySession operation. If a client did a DestroySession on a session while it was still in use, the server might try to use the session structure after it is free'd. I think the client has violated RFC5661 if it does this, but this patch makes DestroySession block all other nfsd threads so no thread could be using the session when it is free'd. After the DestroySession, nfsd threads will not be able to find the session. The patch also adds a check for nd_sessionid being set, although if that was not the case it would have been all 0s and unlikely to have a false match. This might fix the crashes described in PR#228497 for the FreeNAS server. PR: 228497 MFC after: 1 week
2018-05-30 20:16:17 +00:00
int error, igotlock, samesess;
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
samesess = 0;
Strengthen locking for the NFSv4.1 server DestroySession operation. If a client did a DestroySession on a session while it was still in use, the server might try to use the session structure after it is free'd. I think the client has violated RFC5661 if it does this, but this patch makes DestroySession block all other nfsd threads so no thread could be using the session when it is free'd. After the DestroySession, nfsd threads will not be able to find the session. The patch also adds a check for nd_sessionid being set, although if that was not the case it would have been all 0s and unlikely to have a false match. This might fix the crashes described in PR#228497 for the FreeNAS server. PR: 228497 MFC after: 1 week
2018-05-30 20:16:17 +00:00
if (!NFSBCMP(sessionid, nd->nd_sessionid, NFSX_V4SESSIONID) &&
(nd->nd_flag & ND_HASSEQUENCE) != 0) {
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
samesess = 1;
if ((nd->nd_flag & ND_LASTOP) == 0)
return (NFSERR_BADSESSION);
}
Strengthen locking for the NFSv4.1 server DestroySession operation. If a client did a DestroySession on a session while it was still in use, the server might try to use the session structure after it is free'd. I think the client has violated RFC5661 if it does this, but this patch makes DestroySession block all other nfsd threads so no thread could be using the session when it is free'd. After the DestroySession, nfsd threads will not be able to find the session. The patch also adds a check for nd_sessionid being set, although if that was not the case it would have been all 0s and unlikely to have a false match. This might fix the crashes described in PR#228497 for the FreeNAS server. PR: 228497 MFC after: 1 week
2018-05-30 20:16:17 +00:00
/* Lock out other nfsd threads */
NFSLOCKV4ROOTMUTEX();
nfsv4_relref(&nfsv4rootfs_lock);
do {
igotlock = nfsv4_lock(&nfsv4rootfs_lock, 1, NULL,
NFSV4ROOTLOCKMUTEXPTR, NULL);
} while (igotlock == 0);
NFSUNLOCKV4ROOTMUTEX();
nfsd: Add support for the SP4_MACH_CRED case in ExchangeID Commit f4179ad46fa4 added support for operation bitmaps for NFSv4.1/4.2. This commit uses those to implement the SP4_MACH_CRED case for the NFSv4.1/4.2 ExchangeID operation since the Linux NFSv4.1/4.2 client is now using this for Kerberized mounts. The Linux Kerberized NFSv4.1/4.2 mounts currently work without support for this because Linux will fall back to SP4_NONE, but there is no guarantee this fallback will work forever. This commit only affects Kerberized NFSv4.1/4.2 mounts from Linux at this time. MFC after: 3 months
2023-04-07 19:49:23 +00:00
error = nfsrv_freesession(nd, NULL, sessionid);
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
if (error == 0 && samesess != 0)
nd->nd_flag &= ~ND_HASSEQUENCE;
Strengthen locking for the NFSv4.1 server DestroySession operation. If a client did a DestroySession on a session while it was still in use, the server might try to use the session structure after it is free'd. I think the client has violated RFC5661 if it does this, but this patch makes DestroySession block all other nfsd threads so no thread could be using the session when it is free'd. After the DestroySession, nfsd threads will not be able to find the session. The patch also adds a check for nd_sessionid being set, although if that was not the case it would have been all 0s and unlikely to have a false match. This might fix the crashes described in PR#228497 for the FreeNAS server. PR: 228497 MFC after: 1 week
2018-05-30 20:16:17 +00:00
NFSLOCKV4ROOTMUTEX();
nfsv4_unlock(&nfsv4rootfs_lock, 1);
NFSUNLOCKV4ROOTMUTEX();
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
return (error);
}
Add the BindConnectiontoSession operation to the NFSv4.1 server. Under some fairly unusual circumstances, the Linux NFSv4.1 client is doing a BindConnectiontoSession operation for TCP connections. It is also used by the ESXi6.5 NFSv4.1 client. This patch adds this operation to the NFSv4.1 server. Reported by: andreas.nagy@frequentis.com Tested by: andreas.nagy@frequentis.com MFC after: 2 weeks
2018-06-01 19:47:41 +00:00
/*
* Bind a connection to a session.
* For now, only certain variants are supported, since the current session
* structure can only handle a single backchannel entry, which will be
* applied to all connections if it is set.
*/
int
nfsrv_bindconnsess(struct nfsrv_descript *nd, uint8_t *sessionid, int *foreaftp)
{
struct nfssessionhash *shp;
struct nfsdsession *sep;
struct nfsclient *clp;
SVCXPRT *savxprt;
int error;
error = 0;
Fix a LOR between the NFS server and server side krpc. Recent testing of the NFS-over-TLS code found a LOR between the mutex lock used for sessions and the sleep lock used for server side krpc socket structures in nfsrv_checksequence(). This was fixed by r365789. A similar bug exists in nfsrv_bindconnsess(), where SVC_RELEASE() is called while mutexes are held. This patch applies a fix similar to r365789, moving the SVC_RELEASE() call down to after the mutexes are released. This patch fixes the problem by moving the SVC_RELEASE() call in nfsrv_checksequence() down a few lines to below where the mutex is released. MFC after: 1 week
2020-09-18 23:52:56 +00:00
savxprt = NULL;
Add the BindConnectiontoSession operation to the NFSv4.1 server. Under some fairly unusual circumstances, the Linux NFSv4.1 client is doing a BindConnectiontoSession operation for TCP connections. It is also used by the ESXi6.5 NFSv4.1 client. This patch adds this operation to the NFSv4.1 server. Reported by: andreas.nagy@frequentis.com Tested by: andreas.nagy@frequentis.com MFC after: 2 weeks
2018-06-01 19:47:41 +00:00
shp = NFSSESSIONHASH(sessionid);
NFSLOCKSTATE();
NFSLOCKSESSION(shp);
sep = nfsrv_findsession(sessionid);
if (sep != NULL) {
clp = sep->sess_clp;
nfsd: Add support for the SP4_MACH_CRED case in ExchangeID Commit f4179ad46fa4 added support for operation bitmaps for NFSv4.1/4.2. This commit uses those to implement the SP4_MACH_CRED case for the NFSv4.1/4.2 ExchangeID operation since the Linux NFSv4.1/4.2 client is now using this for Kerberized mounts. The Linux Kerberized NFSv4.1/4.2 mounts currently work without support for this because Linux will fall back to SP4_NONE, but there is no guarantee this fallback will work forever. This commit only affects Kerberized NFSv4.1/4.2 mounts from Linux at this time. MFC after: 3 months
2023-04-07 19:49:23 +00:00
error = nfsrv_checkmachcred(NFSV4OP_BINDCONNTOSESS, nd, clp);
if (error != 0)
goto out;
Add the BindConnectiontoSession operation to the NFSv4.1 server. Under some fairly unusual circumstances, the Linux NFSv4.1 client is doing a BindConnectiontoSession operation for TCP connections. It is also used by the ESXi6.5 NFSv4.1 client. This patch adds this operation to the NFSv4.1 server. Reported by: andreas.nagy@frequentis.com Tested by: andreas.nagy@frequentis.com MFC after: 2 weeks
2018-06-01 19:47:41 +00:00
if (*foreaftp == NFSCDFC4_BACK ||
*foreaftp == NFSCDFC4_BACK_OR_BOTH ||
*foreaftp == NFSCDFC4_FORE_OR_BOTH) {
/* Try to set up a backchannel. */
if (clp->lc_req.nr_client == NULL) {
NFSD_DEBUG(2, "nfsrv_bindconnsess: acquire "
"backchannel\n");
clp->lc_req.nr_client = (struct __rpc_client *)
clnt_bck_create(nd->nd_xprt->xp_socket,
sep->sess_cbprogram, NFSV4_CBVERS);
}
if (clp->lc_req.nr_client != NULL) {
NFSD_DEBUG(2, "nfsrv_bindconnsess: set up "
"backchannel\n");
savxprt = sep->sess_cbsess.nfsess_xprt;
SVC_ACQUIRE(nd->nd_xprt);
krpc: Acquire ref count of CLIENT for backchannel use Michael Dexter <editor@callfortesting.org> reported a crash in FreeNAS, where the first argument to clnt_bck_svccall() was no longer valid. This argument is a pointer to the callback CLIENT structure, which is free'd when the associated NFSv4 ClientID is free'd. This appears to have occurred because a callback reply was still in the socket receive queue when the CLIENT structure was free'd. This patch acquires a reference count on the CLIENT that is not CLNT_RELEASE()'d until the socket structure is destroyed. This should guarantee that the CLIENT structure is still valid when clnt_bck_svccall() is called. It also adds a check for closed or closing to clnt_bck_svccall() so that it will not process the callback RPC reply message after the ClientID is free'd. Comments by: mav MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D30153
2021-06-11 23:57:14 +00:00
CLNT_ACQUIRE(clp->lc_req.nr_client);
nd->nd_xprt->xp_p2 = clp->lc_req.nr_client;
Add the BindConnectiontoSession operation to the NFSv4.1 server. Under some fairly unusual circumstances, the Linux NFSv4.1 client is doing a BindConnectiontoSession operation for TCP connections. It is also used by the ESXi6.5 NFSv4.1 client. This patch adds this operation to the NFSv4.1 server. Reported by: andreas.nagy@frequentis.com Tested by: andreas.nagy@frequentis.com MFC after: 2 weeks
2018-06-01 19:47:41 +00:00
/* Disable idle timeout. */
nd->nd_xprt->xp_idletimeout = 0;
sep->sess_cbsess.nfsess_xprt = nd->nd_xprt;
sep->sess_crflags |= NFSV4CRSESS_CONNBACKCHAN;
nfsd: do not implicitly bind the back channel for NFSv4.1/4.2 mounts The NFSv4.1 (and 4.2 on 13) server incorrectly binds a new TCP connection to the back channel when first used by an RPC with a Sequence op in it (almost all of them). RFC5661 specifies that only the fore channel should be bound. This was done because early clients (including FreeBSD) did not do the required BindConnectionToSession RPC. Unfortunately, this breaks the Linux client when the "nconnects" mount option is used, since the server may do a callback on the incorrect TCP connection. This patch converts the server behaviour to that required by the RFC. It also makes the server test/indicate failure of the back channel more aggressively. Until this patch is applied to the server, the "nconnects" mount option is not recommended for a Linux NFSv4.1/4.2 client mount to the FreeBSD server. Reported by: bcodding@redhat.com Tested by: bcodding@redhat.com PR: 254560 MFC after: 1 week
2021-03-30 21:31:05 +00:00
clp->lc_flags |= LCL_DONEBINDCONN |
LCL_NEEDSCBNULL;
nfsd: fix BindConnectionToSession so that it clears "cb path down" Commit 01ae8969a9ee stopped the NFSv4.1/4.2 server from implicitly binding the back channel to a new TCP connection so that it conforms to RFC5661, for NFSv4.1/4.2. An effect of this for the Linux NFS client is that it will do a BindConnectionToSession when it sees NFSV4SEQ_CBPATHDOWN set in a sequence reply. It will do this for every RPC reply until it no longer sees the flag. Without that patch, this will happen until the client does an Open, which will clear LCL_CBDOWN. This patch clears LCL_CBDOWN right away, so that NFSV4SEQ_CBPATHDOWN will no longer be sent to the client in Sequence replies and the Linux client will not repeat the BindConnectionToSession RPCs. This is not critical for correct behaviour, but reduces RPC overheads for cases where the Open will not be done for a while. MFC after: 2 weeks
2021-04-04 22:05:39 +00:00
clp->lc_flags &= ~LCL_CBDOWN;
Add the BindConnectiontoSession operation to the NFSv4.1 server. Under some fairly unusual circumstances, the Linux NFSv4.1 client is doing a BindConnectiontoSession operation for TCP connections. It is also used by the ESXi6.5 NFSv4.1 client. This patch adds this operation to the NFSv4.1 server. Reported by: andreas.nagy@frequentis.com Tested by: andreas.nagy@frequentis.com MFC after: 2 weeks
2018-06-01 19:47:41 +00:00
if (*foreaftp == NFSCDFS4_BACK)
*foreaftp = NFSCDFS4_BACK;
else
*foreaftp = NFSCDFS4_BOTH;
} else if (*foreaftp != NFSCDFC4_BACK) {
NFSD_DEBUG(2, "nfsrv_bindconnsess: can't set "
"up backchannel\n");
sep->sess_crflags &= ~NFSV4CRSESS_CONNBACKCHAN;
clp->lc_flags |= LCL_DONEBINDCONN;
*foreaftp = NFSCDFS4_FORE;
} else {
error = NFSERR_NOTSUPP;
printf("nfsrv_bindconnsess: Can't add "
"backchannel\n");
}
} else {
NFSD_DEBUG(2, "nfsrv_bindconnsess: Set forechannel\n");
clp->lc_flags |= LCL_DONEBINDCONN;
*foreaftp = NFSCDFS4_FORE;
}
} else
error = NFSERR_BADSESSION;
nfsd: Add support for the SP4_MACH_CRED case in ExchangeID Commit f4179ad46fa4 added support for operation bitmaps for NFSv4.1/4.2. This commit uses those to implement the SP4_MACH_CRED case for the NFSv4.1/4.2 ExchangeID operation since the Linux NFSv4.1/4.2 client is now using this for Kerberized mounts. The Linux Kerberized NFSv4.1/4.2 mounts currently work without support for this because Linux will fall back to SP4_NONE, but there is no guarantee this fallback will work forever. This commit only affects Kerberized NFSv4.1/4.2 mounts from Linux at this time. MFC after: 3 months
2023-04-07 19:49:23 +00:00
out:
Add the BindConnectiontoSession operation to the NFSv4.1 server. Under some fairly unusual circumstances, the Linux NFSv4.1 client is doing a BindConnectiontoSession operation for TCP connections. It is also used by the ESXi6.5 NFSv4.1 client. This patch adds this operation to the NFSv4.1 server. Reported by: andreas.nagy@frequentis.com Tested by: andreas.nagy@frequentis.com MFC after: 2 weeks
2018-06-01 19:47:41 +00:00
NFSUNLOCKSESSION(shp);
NFSUNLOCKSTATE();
Fix a LOR between the NFS server and server side krpc. Recent testing of the NFS-over-TLS code found a LOR between the mutex lock used for sessions and the sleep lock used for server side krpc socket structures in nfsrv_checksequence(). This was fixed by r365789. A similar bug exists in nfsrv_bindconnsess(), where SVC_RELEASE() is called while mutexes are held. This patch applies a fix similar to r365789, moving the SVC_RELEASE() call down to after the mutexes are released. This patch fixes the problem by moving the SVC_RELEASE() call in nfsrv_checksequence() down a few lines to below where the mutex is released. MFC after: 1 week
2020-09-18 23:52:56 +00:00
if (savxprt != NULL)
SVC_RELEASE(savxprt);
Add the BindConnectiontoSession operation to the NFSv4.1 server. Under some fairly unusual circumstances, the Linux NFSv4.1 client is doing a BindConnectiontoSession operation for TCP connections. It is also used by the ESXi6.5 NFSv4.1 client. This patch adds this operation to the NFSv4.1 server. Reported by: andreas.nagy@frequentis.com Tested by: andreas.nagy@frequentis.com MFC after: 2 weeks
2018-06-01 19:47:41 +00:00
return (error);
}
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
/*
* Free up a session structure.
*/
static int
nfsd: Add support for the SP4_MACH_CRED case in ExchangeID Commit f4179ad46fa4 added support for operation bitmaps for NFSv4.1/4.2. This commit uses those to implement the SP4_MACH_CRED case for the NFSv4.1/4.2 ExchangeID operation since the Linux NFSv4.1/4.2 client is now using this for Kerberized mounts. The Linux Kerberized NFSv4.1/4.2 mounts currently work without support for this because Linux will fall back to SP4_NONE, but there is no guarantee this fallback will work forever. This commit only affects Kerberized NFSv4.1/4.2 mounts from Linux at this time. MFC after: 3 months
2023-04-07 19:49:23 +00:00
nfsrv_freesession(struct nfsrv_descript *nd, struct nfsdsession *sep,
uint8_t *sessionid)
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
{
struct nfssessionhash *shp;
int i;
Fix a LOR in the NFSv4.1 server. The ordering of acquisition of the state and session mutexes was reversed in two cases executed when an NFSv4.1 client created/freed a session. Since clients will typically do this only when mounting and dismounting, the likelyhood of causing a deadlock was low but possible. This can only occur for NFSv4.1 mounts, since the others do not use sessions. This was detected while testing the pNFS server/client where the client crashed during dismounting. The patch also reorders the unlocks, although that isn't necessary for correct operation. MFC after: 2 weeks
2016-04-23 01:22:04 +00:00
NFSLOCKSTATE();
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
if (sep == NULL) {
shp = NFSSESSIONHASH(sessionid);
NFSLOCKSESSION(shp);
sep = nfsrv_findsession(sessionid);
} else {
shp = NFSSESSIONHASH(sep->sess_sessionid);
NFSLOCKSESSION(shp);
}
if (sep != NULL) {
nfsd: Add support for the SP4_MACH_CRED case in ExchangeID Commit f4179ad46fa4 added support for operation bitmaps for NFSv4.1/4.2. This commit uses those to implement the SP4_MACH_CRED case for the NFSv4.1/4.2 ExchangeID operation since the Linux NFSv4.1/4.2 client is now using this for Kerberized mounts. The Linux Kerberized NFSv4.1/4.2 mounts currently work without support for this because Linux will fall back to SP4_NONE, but there is no guarantee this fallback will work forever. This commit only affects Kerberized NFSv4.1/4.2 mounts from Linux at this time. MFC after: 3 months
2023-04-07 19:49:23 +00:00
/* Check for the SP4_MACH_CRED case. */
if (nd != NULL && nfsrv_checkmachcred(NFSV4OP_DESTROYSESSION,
nd, sep->sess_clp) != 0) {
NFSUNLOCKSESSION(shp);
NFSUNLOCKSTATE();
return (NFSERR_AUTHERR | AUTH_TOOWEAK);
}
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
sep->sess_refcnt--;
if (sep->sess_refcnt > 0) {
NFSUNLOCKSESSION(shp);
Fix a LOR in the NFSv4.1 server. The ordering of acquisition of the state and session mutexes was reversed in two cases executed when an NFSv4.1 client created/freed a session. Since clients will typically do this only when mounting and dismounting, the likelyhood of causing a deadlock was low but possible. This can only occur for NFSv4.1 mounts, since the others do not use sessions. This was detected while testing the pNFS server/client where the client crashed during dismounting. The patch also reorders the unlocks, although that isn't necessary for correct operation. MFC after: 2 weeks
2016-04-23 01:22:04 +00:00
NFSUNLOCKSTATE();
The NFSv4.1 server should return NFSERR_BACKCHANBUSY instead of NFS_OK. When an NFSv4.1 session is busy due to a callback being in progress, nfsrv_freesession() should return NFSERR_BACKCHANBUSY instead of NFS_OK. The only effect this has is that the DestroySession operation will report the failure for this case and this probably has little or no effect on a client. Spotted by inspection and no failures related to this have been reported. MFC after: 2 months
2018-05-13 12:29:09 +00:00
return (NFSERR_BACKCHANBUSY);
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
}
LIST_REMOVE(sep, sess_hash);
LIST_REMOVE(sep, sess_list);
}
NFSUNLOCKSESSION(shp);
Fix a LOR in the NFSv4.1 server. The ordering of acquisition of the state and session mutexes was reversed in two cases executed when an NFSv4.1 client created/freed a session. Since clients will typically do this only when mounting and dismounting, the likelyhood of causing a deadlock was low but possible. This can only occur for NFSv4.1 mounts, since the others do not use sessions. This was detected while testing the pNFS server/client where the client crashed during dismounting. The patch also reorders the unlocks, although that isn't necessary for correct operation. MFC after: 2 weeks
2016-04-23 01:22:04 +00:00
NFSUNLOCKSTATE();
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
if (sep == NULL)
return (NFSERR_BADSESSION);
for (i = 0; i < NFSV4_SLOTS; i++)
if (sep->sess_slots[i].nfssl_reply != NULL)
m_freem(sep->sess_slots[i].nfssl_reply);
if (sep->sess_cbsess.nfsess_xprt != NULL)
SVC_RELEASE(sep->sess_cbsess.nfsess_xprt);
free(sep, M_NFSDSESSION);
return (0);
}
/*
* Free a stateid.
* RFC5661 says that it should fail when there are associated opens, locks
* or delegations. Since stateids represent opens, I don't see how you can
* free an open stateid (it will be free'd when closed), so this function
* only works for lock stateids (freeing the lock_owner) or delegations.
*/
int
nfsrv_freestateid(struct nfsrv_descript *nd, nfsv4stateid_t *stateidp,
NFSPROC_T *p)
{
struct nfsclient *clp;
struct nfsstate *stp;
int error;
NFSLOCKSTATE();
/*
* Look up the stateid
*/
error = nfsrv_getclient((nfsquad_t)((u_quad_t)0), CLOPS_RENEW, &clp,
NULL, (nfsquad_t)((u_quad_t)0), 0, nd, p);
if (error == 0) {
/* First, check for a delegation. */
LIST_FOREACH(stp, &clp->lc_deleg, ls_list) {
if (!NFSBCMP(stp->ls_stateid.other, stateidp->other,
NFSX_STATEIDOTHER))
break;
}
if (stp != NULL) {
nfsrv_freedeleg(stp);
NFSUNLOCKSTATE();
return (error);
}
}
/* Not a delegation, try for a lock_owner. */
if (error == 0)
error = nfsrv_getstate(clp, stateidp, 0, &stp);
if (error == 0 && ((stp->ls_flags & (NFSLCK_OPEN | NFSLCK_DELEGREAD |
NFSLCK_DELEGWRITE)) != 0 || (stp->ls_flags & NFSLCK_LOCK) == 0))
/* Not a lock_owner stateid. */
error = NFSERR_LOCKSHELD;
if (error == 0 && !LIST_EMPTY(&stp->ls_lock))
error = NFSERR_LOCKSHELD;
if (error == 0)
nfsrv_freelockowner(stp, NULL, 0, p);
NFSUNLOCKSTATE();
return (error);
}
Add support for the TestStateID operation to the NFSv4.1 server. The Linux client now uses the TestStateID operation, so this patch adds support for it to the NFSv4.1 server. The FreeBSD client never uses this operation, so it should not be affected. MFC after: 2 months
2018-05-11 22:16:23 +00:00
/*
* Test a stateid.
*/
int
nfsrv_teststateid(struct nfsrv_descript *nd, nfsv4stateid_t *stateidp,
NFSPROC_T *p)
{
struct nfsclient *clp;
struct nfsstate *stp;
int error;
NFSLOCKSTATE();
/*
* Look up the stateid
*/
error = nfsrv_getclient((nfsquad_t)((u_quad_t)0), CLOPS_RENEW, &clp,
NULL, (nfsquad_t)((u_quad_t)0), 0, nd, p);
if (error == 0)
error = nfsrv_getstate(clp, stateidp, 0, &stp);
if (error == 0 && stateidp->seqid != 0 &&
SEQ_LT(stateidp->seqid, stp->ls_stateid.seqid))
error = NFSERR_OLDSTATEID;
NFSUNLOCKSTATE();
return (error);
}
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
/*
* Generate the xdr for an NFSv4.1 CBSequence Operation.
*/
static int
nfsv4_setcbsequence(struct nfsrv_descript *nd, struct nfsclient *clp,
nfsd: fix session slot handling for failed callbacks When the NFSv4.1/4.2 server does a callback to a client on the back channel, it will use a session slot in the back channel session. If the back channel has failed, the callback will fail and, without this patch, the session slot will not be released. As more callbacks are attempted, all session slots can become busy and then the nfsd thread gets stuck waiting for a back channel session slot. This patch frees the session slot upon callback failure to avoid this problem. Without this patch, the problem can be avoided by leaving delegations disabled in the NFS server. MFC after: 2 weeks
2021-04-23 22:24:47 +00:00
int dont_replycache, struct nfsdsession **sepp, int *slotposp)
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
{
struct nfsdsession *sep;
uint32_t *tl, slotseq = 0;
nfsd: fix session slot handling for failed callbacks When the NFSv4.1/4.2 server does a callback to a client on the back channel, it will use a session slot in the back channel session. If the back channel has failed, the callback will fail and, without this patch, the session slot will not be released. As more callbacks are attempted, all session slots can become busy and then the nfsd thread gets stuck waiting for a back channel session slot. This patch frees the session slot upon callback failure to avoid this problem. Without this patch, the problem can be avoided by leaving delegations disabled in the NFS server. MFC after: 2 weeks
2021-04-23 22:24:47 +00:00
int maxslot;
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
uint8_t sessionid[NFSX_V4SESSIONID];
int error;
error = nfsv4_getcbsession(clp, sepp);
if (error != 0)
return (error);
sep = *sepp;
nfscl: Add optional support for slots marked bad This patch adds support for session slots marked bad to nfsv4_sequencelookup(). An additional boolean argument indicates if the check for slots marked bad should be done. The "cred" argument added to nfscl_reqstart() by commit 326bcf9394c7 is now passed into nfsv4_setquence() so that it can optionally set the boolean argument for nfsv4_sequencelookup(). When optionally enabled, nfsv4_setsequence() will do a DestroySession when all slots are marked bad. Since the code that marks slots bad is not yet committed, this patch should not result in a semantics change. PR: 260011 MFC after: 2 weeks
2022-07-09 21:43:16 +00:00
nfsv4_sequencelookup(NULL, &sep->sess_cbsess, slotposp, &maxslot,
&slotseq, sessionid, true);
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
KASSERT(maxslot >= 0, ("nfsv4_setcbsequence neg maxslot"));
/* Build the Sequence arguments. */
NFSM_BUILD(tl, uint32_t *, NFSX_V4SESSIONID + 5 * NFSX_UNSIGNED);
bcopy(sessionid, tl, NFSX_V4SESSIONID);
tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
nd->nd_slotseq = tl;
nfsd: Fix session slot freeing for NFSv4.1/4.2 Without this patch the NFSv4.1/4.2 server erroneously always frees session slot zero for callbacks. This only affects 4.1/4.2 mounts if the server has delegations enabled or is a pNFS configuration. Even for those cases, the effect is mainly to only use slot 0 for callbacks, serializing all of them. There is a slight chance that callbacks will fail if the client performs them in a different order than received on the TCP connection. If this bug affects your server, you will see console messages like: newnfs_request: Bad session slot This patch fixes the problem. Found during a recent IETF NFSv4 testing event. PR: 263728 MFC after: 2 weeks
2022-05-02 19:45:42 +00:00
nd->nd_slotid = *slotposp;
nd->nd_flag |= ND_HASSLOTID;
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
*tl++ = txdr_unsigned(slotseq);
nfsd: fix session slot handling for failed callbacks When the NFSv4.1/4.2 server does a callback to a client on the back channel, it will use a session slot in the back channel session. If the back channel has failed, the callback will fail and, without this patch, the session slot will not be released. As more callbacks are attempted, all session slots can become busy and then the nfsd thread gets stuck waiting for a back channel session slot. This patch frees the session slot upon callback failure to avoid this problem. Without this patch, the problem can be avoided by leaving delegations disabled in the NFS server. MFC after: 2 weeks
2021-04-23 22:24:47 +00:00
*tl++ = txdr_unsigned(*slotposp);
Merge the NFSv4.1 server code in projects/nfsv4.1-server over into head. The code is not believed to have any effect on the semantics of non-NFSv4.1 server behaviour. It is a rather large merge, but I am hoping that there will not be any regressions for the NFS server. MFC after: 1 month
2014-07-01 20:47:16 +00:00
*tl++ = txdr_unsigned(maxslot);
if (dont_replycache == 0)
*tl++ = newnfs_true;
else
*tl++ = newnfs_false;
*tl = 0; /* No referring call list, for now. */
nd->nd_flag |= ND_HASSEQUENCE;
return (0);
}
/*
* Get a session for the callback.
*/
static int
nfsv4_getcbsession(struct nfsclient *clp, struct nfsdsession **sepp)
{
struct nfsdsession *sep;
NFSLOCKSTATE();
LIST_FOREACH(sep, &clp->lc_session, sess_list) {
if ((sep->sess_crflags & NFSV4CRSESS_CONNBACKCHAN) != 0)
break;
}
if (sep == NULL) {
NFSUNLOCKSTATE();
return (NFSERR_BADSESSION);
}
sep->sess_refcnt++;
*sepp = sep;
NFSUNLOCKSTATE();
return (0);
}
When the nfsd threads are terminated, the NFSv4 server state (opens, locks, etc) is retained, which I believe is correct behaviour. However, for NFSv4.1, the server also retained a reference to the xprt (RPC transport socket structure) for the backchannel. This caused svcpool_destroy() to not call SVC_DESTROY() for the xprt and allowed a socket upcall to occur after the mutexes in the svcpool were destroyed, causing a crash. This patch fixes the code so that the backchannel xprt structure is dereferenced just before svcpool_destroy() is called, so the code does do an SVC_DESTROY() on the xprt, which shuts down the socket upcall. Tested by: g_amanakis@yahoo.com PR: 204340 MFC after: 2 weeks
2015-11-21 23:55:46 +00:00
/*
* Free up all backchannel xprts. This needs to be done when the nfsd threads
* exit, since those transports will all be going away.
* This is only called after all the nfsd threads are done performing RPCs,
* so locking shouldn't be an issue.
*/
nfs: Remove APPLESTATIC macro It is no longer useful. Reviewed by: rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24811
2020-05-12 13:23:25 +00:00
void
When the nfsd threads are terminated, the NFSv4 server state (opens, locks, etc) is retained, which I believe is correct behaviour. However, for NFSv4.1, the server also retained a reference to the xprt (RPC transport socket structure) for the backchannel. This caused svcpool_destroy() to not call SVC_DESTROY() for the xprt and allowed a socket upcall to occur after the mutexes in the svcpool were destroyed, causing a crash. This patch fixes the code so that the backchannel xprt structure is dereferenced just before svcpool_destroy() is called, so the code does do an SVC_DESTROY() on the xprt, which shuts down the socket upcall. Tested by: g_amanakis@yahoo.com PR: 204340 MFC after: 2 weeks
2015-11-21 23:55:46 +00:00
nfsrv_freeallbackchannel_xprts(void)
{
struct nfsdsession *sep;
struct nfsclient *clp;
SVCXPRT *xprt;
int i;
for (i = 0; i < nfsrv_clienthashsize; i++) {
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
LIST_FOREACH(clp, &NFSD_VNET(nfsclienthash)[i], lc_hash) {
When the nfsd threads are terminated, the NFSv4 server state (opens, locks, etc) is retained, which I believe is correct behaviour. However, for NFSv4.1, the server also retained a reference to the xprt (RPC transport socket structure) for the backchannel. This caused svcpool_destroy() to not call SVC_DESTROY() for the xprt and allowed a socket upcall to occur after the mutexes in the svcpool were destroyed, causing a crash. This patch fixes the code so that the backchannel xprt structure is dereferenced just before svcpool_destroy() is called, so the code does do an SVC_DESTROY() on the xprt, which shuts down the socket upcall. Tested by: g_amanakis@yahoo.com PR: 204340 MFC after: 2 weeks
2015-11-21 23:55:46 +00:00
LIST_FOREACH(sep, &clp->lc_session, sess_list) {
xprt = sep->sess_cbsess.nfsess_xprt;
sep->sess_cbsess.nfsess_xprt = NULL;
if (xprt != NULL)
SVC_RELEASE(xprt);
}
}
}
}
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
/*
* Do a layout commit. Actually just call nfsrv_updatemdsattr().
* I have no idea if the rest of these arguments will ever be useful?
*/
int
nfsrv_layoutcommit(struct nfsrv_descript *nd, vnode_t vp, int layouttype,
int hasnewoff, uint64_t newoff, uint64_t offset, uint64_t len,
int hasnewmtime, struct timespec *newmtimep, int reclaim,
nfsv4stateid_t *stateidp, int maxcnt, char *layp, int *hasnewsizep,
uint64_t *newsizep, struct ucred *cred, NFSPROC_T *p)
{
struct nfsvattr na;
int error;
error = nfsrv_updatemdsattr(vp, &na, p);
if (error == 0) {
*hasnewsizep = 1;
*newsizep = na.na_size;
}
return (error);
}
/*
* Try and get a layout.
*/
int
nfsrv_layoutget(struct nfsrv_descript *nd, vnode_t vp, struct nfsexstuff *exp,
int layouttype, int *iomode, uint64_t *offset, uint64_t *len,
uint64_t minlen, nfsv4stateid_t *stateidp, int maxcnt, int *retonclose,
int *layoutlenp, char *layp, struct ucred *cred, NFSPROC_T *p)
{
struct nfslayouthash *lhyp;
struct nfslayout *lyp;
char *devid;
fhandle_t fh, *dsfhp;
int error, mirrorcnt;
if (nfsrv_devidcnt == 0)
return (NFSERR_UNKNLAYOUTTYPE);
if (*offset != 0)
printf("nfsrv_layoutget: off=%ju len=%ju\n", (uintmax_t)*offset,
(uintmax_t)*len);
error = nfsvno_getfh(vp, &fh, p);
NFSD_DEBUG(4, "layoutget getfh=%d\n", error);
if (error != 0)
return (error);
/*
* For now, all layouts are for entire files.
* Only issue Read/Write layouts if requested for a non-readonly fs.
*/
if (NFSVNO_EXRDONLY(exp)) {
if (*iomode == NFSLAYOUTIOMODE_RW)
return (NFSERR_LAYOUTTRYLATER);
*iomode = NFSLAYOUTIOMODE_READ;
}
if (*iomode != NFSLAYOUTIOMODE_RW)
*iomode = NFSLAYOUTIOMODE_READ;
/*
* Check to see if a write layout can be issued for this file.
* This is used during mirror recovery to avoid RW layouts being
* issued for a file while it is being copied to the recovered
* mirror.
*/
if (*iomode == NFSLAYOUTIOMODE_RW && nfsrv_dontlayout(&fh) != 0)
return (NFSERR_LAYOUTTRYLATER);
*retonclose = 0;
*offset = 0;
*len = UINT64_MAX;
/* First, see if a layout already exists and return if found. */
lhyp = NFSLAYOUTHASH(&fh);
NFSLOCKLAYOUT(lhyp);
error = nfsrv_findlayout(&nd->nd_clientid, &fh, layouttype, p, &lyp);
NFSD_DEBUG(4, "layoutget findlay=%d\n", error);
/*
* Not sure if the seqid must be the same, so I won't check it.
*/
if (error == 0 && (stateidp->other[0] != lyp->lay_stateid.other[0] ||
stateidp->other[1] != lyp->lay_stateid.other[1] ||
stateidp->other[2] != lyp->lay_stateid.other[2])) {
if ((lyp->lay_flags & NFSLAY_CALLB) == 0) {
NFSUNLOCKLAYOUT(lhyp);
NFSD_DEBUG(1, "ret bad stateid\n");
return (NFSERR_BADSTATEID);
}
/*
* I believe we get here because there is a race between
* the client processing the CBLAYOUTRECALL and the layout
* being deleted here on the server.
* The client has now done a LayoutGet with a non-layout
* stateid, as it would when there is no layout.
* As such, free this layout and set error == NFSERR_BADSTATEID
* so the code below will create a new layout structure as
* would happen if no layout was found.
* "lyp" will be set before being used below, but set it NULL
* as a safety belt.
*/
nfsrv_freelayout(&lhyp->list, lyp);
lyp = NULL;
error = NFSERR_BADSTATEID;
}
if (error == 0) {
if (lyp->lay_layoutlen > maxcnt) {
NFSUNLOCKLAYOUT(lhyp);
NFSD_DEBUG(1, "ret layout too small\n");
return (NFSERR_TOOSMALL);
}
nfsd: Fix the NFSv4.2 pNFS MDS server for NFSERR_NOSPC via LayoutError If a pNFS server's DS runs out of disk space, it replies NFSERR_NOSPC to the client doing writing. For the Linux client, it then sends a LayoutError RPC to the MDS server to tell it about the error and keeps retrying, doing repeated LayoutGets to the MDS and Write RPCs to the DS. The Linux client is "stuck" until disk space on the DS is free'd up unless a subsequent LayoutGet request is sent a NFSERR_NOSPC reply. The looping problem still occurs for NFSv4.1 mounts, but no fix for this is known at this time. This patch changes the pNFS MDS server to reply to LayoutGet operations with NFSERR_NOSPC once a LayoutError reports the problem, until the DS has available space. This keeps the Linux NFSv4.2 from looping. Found during recent testing because of issues w.r.t. a DS being out of space found during a recent IEFT NFSv4 working group testing event. MFC after: 2 weeks
2021-11-08 23:58:00 +00:00
if (*iomode == NFSLAYOUTIOMODE_RW) {
if ((lyp->lay_flags & NFSLAY_NOSPC) != 0) {
NFSUNLOCKLAYOUT(lhyp);
NFSD_DEBUG(1, "ret layout nospace\n");
return (NFSERR_NOSPC);
}
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
lyp->lay_flags |= NFSLAY_RW;
nfsd: Fix the NFSv4.2 pNFS MDS server for NFSERR_NOSPC via LayoutError If a pNFS server's DS runs out of disk space, it replies NFSERR_NOSPC to the client doing writing. For the Linux client, it then sends a LayoutError RPC to the MDS server to tell it about the error and keeps retrying, doing repeated LayoutGets to the MDS and Write RPCs to the DS. The Linux client is "stuck" until disk space on the DS is free'd up unless a subsequent LayoutGet request is sent a NFSERR_NOSPC reply. The looping problem still occurs for NFSv4.1 mounts, but no fix for this is known at this time. This patch changes the pNFS MDS server to reply to LayoutGet operations with NFSERR_NOSPC once a LayoutError reports the problem, until the DS has available space. This keeps the Linux NFSv4.2 from looping. Found during recent testing because of issues w.r.t. a DS being out of space found during a recent IEFT NFSv4 working group testing event. MFC after: 2 weeks
2021-11-08 23:58:00 +00:00
} else
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
lyp->lay_flags |= NFSLAY_READ;
NFSBCOPY(lyp->lay_xdr, layp, lyp->lay_layoutlen);
*layoutlenp = lyp->lay_layoutlen;
if (++lyp->lay_stateid.seqid == 0)
lyp->lay_stateid.seqid = 1;
stateidp->seqid = lyp->lay_stateid.seqid;
NFSUNLOCKLAYOUT(lhyp);
NFSD_DEBUG(4, "ret fnd layout\n");
return (0);
}
NFSUNLOCKLAYOUT(lhyp);
/* Find the device id and file handle. */
dsfhp = malloc(sizeof(fhandle_t) * NFSDEV_MAXMIRRORS, M_TEMP, M_WAITOK);
devid = malloc(NFSX_V4DEVICEID * NFSDEV_MAXMIRRORS, M_TEMP, M_WAITOK);
error = nfsrv_dsgetdevandfh(vp, p, &mirrorcnt, dsfhp, devid);
NFSD_DEBUG(4, "layoutget devandfh=%d\n", error);
if (error == 0) {
if (layouttype == NFSLAYOUT_NFSV4_1_FILES) {
if (NFSX_V4FILELAYOUT > maxcnt)
error = NFSERR_TOOSMALL;
else
lyp = nfsrv_filelayout(nd, *iomode, &fh, dsfhp,
devid, vp->v_mount->mnt_stat.f_fsid);
} else {
if (NFSX_V4FLEXLAYOUT(mirrorcnt) > maxcnt)
error = NFSERR_TOOSMALL;
else
lyp = nfsrv_flexlayout(nd, *iomode, mirrorcnt,
&fh, dsfhp, devid,
vp->v_mount->mnt_stat.f_fsid);
}
}
free(dsfhp, M_TEMP);
free(devid, M_TEMP);
if (error != 0)
return (error);
/*
* Now, add this layout to the list.
*/
error = nfsrv_addlayout(nd, &lyp, stateidp, layp, layoutlenp, p);
NFSD_DEBUG(4, "layoutget addl=%d\n", error);
/*
* The lyp will be set to NULL by nfsrv_addlayout() if it
* linked the new structure into the lists.
*/
free(lyp, M_NFSDSTATE);
return (error);
}
/*
* Generate a File Layout.
*/
static struct nfslayout *
nfsrv_filelayout(struct nfsrv_descript *nd, int iomode, fhandle_t *fhp,
fhandle_t *dsfhp, char *devid, fsid_t fs)
{
uint32_t *tl;
struct nfslayout *lyp;
uint64_t pattern_offset;
lyp = malloc(sizeof(struct nfslayout) + NFSX_V4FILELAYOUT, M_NFSDSTATE,
M_WAITOK | M_ZERO);
lyp->lay_type = NFSLAYOUT_NFSV4_1_FILES;
if (iomode == NFSLAYOUTIOMODE_RW)
lyp->lay_flags = NFSLAY_RW;
else
lyp->lay_flags = NFSLAY_READ;
NFSBCOPY(fhp, &lyp->lay_fh, sizeof(*fhp));
lyp->lay_clientid.qval = nd->nd_clientid.qval;
lyp->lay_fsid = fs;
nfsd: Fix the NFSv4.2 pNFS MDS server for NFSERR_NOSPC via LayoutError If a pNFS server's DS runs out of disk space, it replies NFSERR_NOSPC to the client doing writing. For the Linux client, it then sends a LayoutError RPC to the MDS server to tell it about the error and keeps retrying, doing repeated LayoutGets to the MDS and Write RPCs to the DS. The Linux client is "stuck" until disk space on the DS is free'd up unless a subsequent LayoutGet request is sent a NFSERR_NOSPC reply. The looping problem still occurs for NFSv4.1 mounts, but no fix for this is known at this time. This patch changes the pNFS MDS server to reply to LayoutGet operations with NFSERR_NOSPC once a LayoutError reports the problem, until the DS has available space. This keeps the Linux NFSv4.2 from looping. Found during recent testing because of issues w.r.t. a DS being out of space found during a recent IEFT NFSv4 working group testing event. MFC after: 2 weeks
2021-11-08 23:58:00 +00:00
NFSBCOPY(devid, lyp->lay_deviceid, NFSX_V4DEVICEID);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
/* Fill in the xdr for the files layout. */
tl = (uint32_t *)lyp->lay_xdr;
NFSBCOPY(devid, tl, NFSX_V4DEVICEID); /* Device ID. */
tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
nfsd: fix stripe size reply for the File Layout pNFS server At a recent testing event I found out that I had misinterpreted RFC5661 where it describes the stripe size in the File Layout's nfl_util field. This patch fixes the pNFS File Layout server so that it returns the correct value to the NFSv4.1/4.2 pNFS enabled client. This affects almost no one, since pNFS server configurations are rare and the extant pNFS aware NFS clients seemed to function correctly despite the erroneous stripe size. It *might* be needed for correct behaviour if a recent Linux client mounts a FreeBSD pNFS server configuration that is using File Layout (non-mirrored configuration). MFC after: 2 weeks
2021-04-20 00:51:07 +00:00
/* Set the stripe size to the maximum I/O size. */
nfsd: Add sysctl to set maximum I/O size up to 1Mbyte Since MAXPHYS now allows the FreeBSD NFS client to do 1Mbyte I/O operations, add a sysctl called vfs.nfsd.srvmaxio so that the maximum NFS server I/O size can be set up to 1Mbyte. The Linux NFS client can also do 1Mbyte I/O operations. The default of 128Kbytes for the maximum I/O size has not been changed for two reasons: - kern.ipc.maxsockbuf must be increased to support 1Mbyte I/O - The limited benchmarking I can do actually shows a drop in I/O rate when the I/O size is above 256Kbytes. However, daveb@spectralogic.com reports seeing an increase in I/O rate for the 1Mbyte I/O size vs 128Kbytes using a Linux client. Reviewed by: asomers MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D30826
2021-07-16 22:01:03 +00:00
*tl++ = txdr_unsigned(nfs_srvmaxio & NFSFLAYUTIL_STRIPE_MASK);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
*tl++ = 0; /* 1st stripe index. */
pattern_offset = 0;
txdr_hyper(pattern_offset, tl); tl += 2; /* Pattern offset. */
*tl++ = txdr_unsigned(1); /* 1 file handle. */
*tl++ = txdr_unsigned(NFSX_V4PNFSFH);
NFSBCOPY(dsfhp, tl, sizeof(*dsfhp));
lyp->lay_layoutlen = NFSX_V4FILELAYOUT;
return (lyp);
}
#define FLEX_OWNERID "999"
#define FLEX_UID0 "0"
/*
* Generate a Flex File Layout.
* The FLEX_OWNERID can be any string of 3 decimal digits. Although this
* string goes on the wire, it isn't supposed to be used by the client,
* since this server uses tight coupling.
* Although not recommended by the spec., if vfs.nfsd.flexlinuxhack=1 use
* a string of "0". This works around the Linux Flex File Layout driver bug
* which uses the synthetic uid/gid strings for the "tightly coupled" case.
*/
static struct nfslayout *
nfsrv_flexlayout(struct nfsrv_descript *nd, int iomode, int mirrorcnt,
fhandle_t *fhp, fhandle_t *dsfhp, char *devid, fsid_t fs)
{
uint32_t *tl;
struct nfslayout *lyp;
uint64_t lenval;
int i;
lyp = malloc(sizeof(struct nfslayout) + NFSX_V4FLEXLAYOUT(mirrorcnt),
M_NFSDSTATE, M_WAITOK | M_ZERO);
lyp->lay_type = NFSLAYOUT_FLEXFILE;
if (iomode == NFSLAYOUTIOMODE_RW)
lyp->lay_flags = NFSLAY_RW;
else
lyp->lay_flags = NFSLAY_READ;
NFSBCOPY(fhp, &lyp->lay_fh, sizeof(*fhp));
lyp->lay_clientid.qval = nd->nd_clientid.qval;
lyp->lay_fsid = fs;
lyp->lay_mirrorcnt = mirrorcnt;
nfsd: Fix the NFSv4.2 pNFS MDS server for NFSERR_NOSPC via LayoutError If a pNFS server's DS runs out of disk space, it replies NFSERR_NOSPC to the client doing writing. For the Linux client, it then sends a LayoutError RPC to the MDS server to tell it about the error and keeps retrying, doing repeated LayoutGets to the MDS and Write RPCs to the DS. The Linux client is "stuck" until disk space on the DS is free'd up unless a subsequent LayoutGet request is sent a NFSERR_NOSPC reply. The looping problem still occurs for NFSv4.1 mounts, but no fix for this is known at this time. This patch changes the pNFS MDS server to reply to LayoutGet operations with NFSERR_NOSPC once a LayoutError reports the problem, until the DS has available space. This keeps the Linux NFSv4.2 from looping. Found during recent testing because of issues w.r.t. a DS being out of space found during a recent IEFT NFSv4 working group testing event. MFC after: 2 weeks
2021-11-08 23:58:00 +00:00
NFSBCOPY(devid, lyp->lay_deviceid, NFSX_V4DEVICEID);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
/* Fill in the xdr for the files layout. */
tl = (uint32_t *)lyp->lay_xdr;
lenval = 0;
txdr_hyper(lenval, tl); tl += 2; /* Stripe unit. */
*tl++ = txdr_unsigned(mirrorcnt); /* # of mirrors. */
for (i = 0; i < mirrorcnt; i++) {
*tl++ = txdr_unsigned(1); /* One stripe. */
NFSBCOPY(devid, tl, NFSX_V4DEVICEID); /* Device ID. */
tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
devid += NFSX_V4DEVICEID;
*tl++ = txdr_unsigned(1); /* Efficiency. */
*tl++ = 0; /* Proxy Stateid. */
*tl++ = 0x55555555;
*tl++ = 0x55555555;
*tl++ = 0x55555555;
*tl++ = txdr_unsigned(1); /* 1 file handle. */
*tl++ = txdr_unsigned(NFSX_V4PNFSFH);
NFSBCOPY(dsfhp, tl, sizeof(*dsfhp));
tl += (NFSM_RNDUP(NFSX_V4PNFSFH) / NFSX_UNSIGNED);
dsfhp++;
if (nfsrv_flexlinuxhack != 0) {
*tl++ = txdr_unsigned(strlen(FLEX_UID0));
*tl = 0; /* 0 pad string. */
NFSBCOPY(FLEX_UID0, tl++, strlen(FLEX_UID0));
*tl++ = txdr_unsigned(strlen(FLEX_UID0));
*tl = 0; /* 0 pad string. */
NFSBCOPY(FLEX_UID0, tl++, strlen(FLEX_UID0));
} else {
*tl++ = txdr_unsigned(strlen(FLEX_OWNERID));
NFSBCOPY(FLEX_OWNERID, tl++, NFSX_UNSIGNED);
*tl++ = txdr_unsigned(strlen(FLEX_OWNERID));
NFSBCOPY(FLEX_OWNERID, tl++, NFSX_UNSIGNED);
}
}
*tl++ = txdr_unsigned(0); /* ff_flags. */
*tl = txdr_unsigned(60); /* Status interval hint. */
lyp->lay_layoutlen = NFSX_V4FLEXLAYOUT(mirrorcnt);
return (lyp);
}
/*
* Parse and process Flex File errors returned via LayoutReturn.
*/
static void
nfsrv_flexlayouterr(struct nfsrv_descript *nd, uint32_t *layp, int maxcnt,
NFSPROC_T *p)
{
uint32_t *tl;
int cnt, errcnt, i, j, opnum, stat;
char devid[NFSX_V4DEVICEID];
tl = layp;
nfsd: Limit parsing of layout errors to maxcnt bytes This patch decrements maxcnt by the appropriate number of bytes during parsing and checks to see if there is data remaining. If not, it just returns from nfsrv_flexlayouterr() without further processing. This prevents the tl pointer from running off the end of the error data pointed at by layp, if there are flaws in the data. Reported by: rtm@lcs.mit.edu Tested by: rtm@lcs.mit.edu PR: 260293 MFC after: 2 weeks
2021-12-13 23:21:31 +00:00
maxcnt -= NFSX_UNSIGNED;
if (maxcnt > 0)
cnt = fxdr_unsigned(int, *tl++);
else
cnt = 0;
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
NFSD_DEBUG(4, "flexlayouterr cnt=%d\n", cnt);
for (i = 0; i < cnt; i++) {
nfsd: Limit parsing of layout errors to maxcnt bytes This patch decrements maxcnt by the appropriate number of bytes during parsing and checks to see if there is data remaining. If not, it just returns from nfsrv_flexlayouterr() without further processing. This prevents the tl pointer from running off the end of the error data pointed at by layp, if there are flaws in the data. Reported by: rtm@lcs.mit.edu Tested by: rtm@lcs.mit.edu PR: 260293 MFC after: 2 weeks
2021-12-13 23:21:31 +00:00
maxcnt -= NFSX_STATEID + 2 * NFSX_HYPER +
NFSX_UNSIGNED;
if (maxcnt <= 0)
break;
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
/* Skip offset, length and stateid for now. */
tl += (4 + NFSX_STATEID / NFSX_UNSIGNED);
errcnt = fxdr_unsigned(int, *tl++);
NFSD_DEBUG(4, "flexlayouterr errcnt=%d\n", errcnt);
for (j = 0; j < errcnt; j++) {
nfsd: Limit parsing of layout errors to maxcnt bytes This patch decrements maxcnt by the appropriate number of bytes during parsing and checks to see if there is data remaining. If not, it just returns from nfsrv_flexlayouterr() without further processing. This prevents the tl pointer from running off the end of the error data pointed at by layp, if there are flaws in the data. Reported by: rtm@lcs.mit.edu Tested by: rtm@lcs.mit.edu PR: 260293 MFC after: 2 weeks
2021-12-13 23:21:31 +00:00
maxcnt -= NFSX_V4DEVICEID + 2 * NFSX_UNSIGNED;
if (maxcnt < 0)
break;
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
NFSBCOPY(tl, devid, NFSX_V4DEVICEID);
tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
stat = fxdr_unsigned(int, *tl++);
opnum = fxdr_unsigned(int, *tl++);
NFSD_DEBUG(4, "flexlayouterr op=%d stat=%d\n", opnum,
stat);
/*
nfsd: Fix the NFSv4 pNFS MDS server for DS NFSERR_NOSPC If a pNFS server's DS runs out of disk space, it replies NFSERR_NOSPC to the client doing writing. For the Linux client, it then sends a LayoutError RPC to the server to tell it about the error and keeps retrying, doing repeated LayoutGet and Write RPCs to the DS. The Linux client is "stuck" until disk space on the DS is free'd up. For a mirrored server configuration, the first mirror that ran out of space was taken offline. This does not make much sense, since the other mirror(s) will run out of space soon and the fix is a manual cleanup up disk space. This patch changes the pNFS server to not disable a mirror for the mirrored case when this occurs. Further work is needed, since the Linux client expects the MDS to reply NFSERR_NOSPC to LayoutGets once the DS is out of space. Without this further change, the above mentioned looping occurs. Found during a recent IEFT NFSv4 working group testing event. MFC after: 2 weeks
2021-11-07 19:43:03 +00:00
* Except for NFSERR_ACCES, NFSERR_STALE and
* NFSERR_NOSPC errors, disable the mirror.
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
*/
nfsd: Fix the NFSv4 pNFS MDS server for DS NFSERR_NOSPC If a pNFS server's DS runs out of disk space, it replies NFSERR_NOSPC to the client doing writing. For the Linux client, it then sends a LayoutError RPC to the server to tell it about the error and keeps retrying, doing repeated LayoutGet and Write RPCs to the DS. The Linux client is "stuck" until disk space on the DS is free'd up. For a mirrored server configuration, the first mirror that ran out of space was taken offline. This does not make much sense, since the other mirror(s) will run out of space soon and the fix is a manual cleanup up disk space. This patch changes the pNFS server to not disable a mirror for the mirrored case when this occurs. Further work is needed, since the Linux client expects the MDS to reply NFSERR_NOSPC to LayoutGets once the DS is out of space. Without this further change, the above mentioned looping occurs. Found during a recent IEFT NFSv4 working group testing event. MFC after: 2 weeks
2021-11-07 19:43:03 +00:00
if (stat != NFSERR_ACCES && stat != NFSERR_STALE &&
stat != NFSERR_NOSPC)
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
nfsrv_delds(devid, p);
nfsd: Fix the NFSv4.2 pNFS MDS server for NFSERR_NOSPC via LayoutError If a pNFS server's DS runs out of disk space, it replies NFSERR_NOSPC to the client doing writing. For the Linux client, it then sends a LayoutError RPC to the MDS server to tell it about the error and keeps retrying, doing repeated LayoutGets to the MDS and Write RPCs to the DS. The Linux client is "stuck" until disk space on the DS is free'd up unless a subsequent LayoutGet request is sent a NFSERR_NOSPC reply. The looping problem still occurs for NFSv4.1 mounts, but no fix for this is known at this time. This patch changes the pNFS MDS server to reply to LayoutGet operations with NFSERR_NOSPC once a LayoutError reports the problem, until the DS has available space. This keeps the Linux NFSv4.2 from looping. Found during recent testing because of issues w.r.t. a DS being out of space found during a recent IEFT NFSv4 working group testing event. MFC after: 2 weeks
2021-11-08 23:58:00 +00:00
/* For NFSERR_NOSPC, mark all devids and layouts. */
if (stat == NFSERR_NOSPC)
nfsrv_marknospc(devid, true);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
}
}
}
/*
* This function removes all flex file layouts which has a mirror with
* a device id that matches the argument.
* Called when the DS represented by the device id has failed.
*/
void
nfsrv_flexmirrordel(char *devid, NFSPROC_T *p)
{
uint32_t *tl;
struct nfslayout *lyp, *nlyp;
struct nfslayouthash *lhyp;
struct nfslayouthead loclyp;
int i, j;
NFSD_DEBUG(4, "flexmirrordel\n");
/* Move all layouts found onto a local list. */
TAILQ_INIT(&loclyp);
for (i = 0; i < nfsrv_layouthashsize; i++) {
lhyp = &nfslayouthash[i];
NFSLOCKLAYOUT(lhyp);
TAILQ_FOREACH_SAFE(lyp, &lhyp->list, lay_list, nlyp) {
if (lyp->lay_type == NFSLAYOUT_FLEXFILE &&
lyp->lay_mirrorcnt > 1) {
NFSD_DEBUG(4, "possible match\n");
tl = lyp->lay_xdr;
tl += 3;
for (j = 0; j < lyp->lay_mirrorcnt; j++) {
tl++;
if (NFSBCMP(devid, tl, NFSX_V4DEVICEID)
== 0) {
/* Found one. */
NFSD_DEBUG(4, "fnd one\n");
TAILQ_REMOVE(&lhyp->list, lyp,
lay_list);
TAILQ_INSERT_HEAD(&loclyp, lyp,
lay_list);
break;
}
tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED +
NFSM_RNDUP(NFSX_V4PNFSFH) /
NFSX_UNSIGNED + 11 * NFSX_UNSIGNED);
}
}
}
NFSUNLOCKLAYOUT(lhyp);
}
/* Now, try to do a Layout recall for each one found. */
TAILQ_FOREACH_SAFE(lyp, &loclyp, lay_list, nlyp) {
NFSD_DEBUG(4, "do layout recall\n");
/*
* The layout stateid.seqid needs to be incremented
* before doing a LAYOUT_RECALL callback.
*/
if (++lyp->lay_stateid.seqid == 0)
lyp->lay_stateid.seqid = 1;
nfsrv_recalllayout(lyp->lay_clientid, &lyp->lay_stateid,
Assorted fixes to handling of LayoutRecall callbacks, mostly error handling. After a re-read of the appropriate section of RFC5661, I decided that a few things should be changed related to LayoutRecall callback handling. Here are the things fixed by this patch. - For two of the three cases that LayoutRecall is done, I now think setting the clora_changed argument false is correct. - All errors other than NFSERR_DELAY returned by LayoutRecall appear permanent, so don't retry for any of them. (NFSERR_DELAY is retried by newnfs_request(), so it is not affected by this patch.) - Instead of waiting "forever" (actually until the process is SIGTERM'd) for Layouts to be returned during a mirror copy, fail and return ENXIO after about 1minute. Waiting for a <ctrl>C made sense when pnfsdscopymr() was done by itself, but did not make sense when done via find(1). This patch only affects the pNFS server.
2018-08-08 20:21:45 +00:00
&lyp->lay_fh, lyp, 1, lyp->lay_type, p);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
nfsrv_freelayout(&loclyp, lyp);
}
}
/*
* Do a recall callback to the client for this layout.
*/
static int
nfsrv_recalllayout(nfsquad_t clid, nfsv4stateid_t *stateidp, fhandle_t *fhp,
Assorted fixes to handling of LayoutRecall callbacks, mostly error handling. After a re-read of the appropriate section of RFC5661, I decided that a few things should be changed related to LayoutRecall callback handling. Here are the things fixed by this patch. - For two of the three cases that LayoutRecall is done, I now think setting the clora_changed argument false is correct. - All errors other than NFSERR_DELAY returned by LayoutRecall appear permanent, so don't retry for any of them. (NFSERR_DELAY is retried by newnfs_request(), so it is not affected by this patch.) - Instead of waiting "forever" (actually until the process is SIGTERM'd) for Layouts to be returned during a mirror copy, fail and return ENXIO after about 1minute. Waiting for a <ctrl>C made sense when pnfsdscopymr() was done by itself, but did not make sense when done via find(1). This patch only affects the pNFS server.
2018-08-08 20:21:45 +00:00
struct nfslayout *lyp, int changed, int laytype, NFSPROC_T *p)
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
{
struct nfsclient *clp;
int error;
NFSD_DEBUG(4, "nfsrv_recalllayout\n");
error = nfsrv_getclient(clid, 0, &clp, NULL, (nfsquad_t)((u_quad_t)0),
0, NULL, p);
NFSD_DEBUG(4, "aft nfsrv_getclient=%d\n", error);
Assorted fixes to handling of LayoutRecall callbacks, mostly error handling. After a re-read of the appropriate section of RFC5661, I decided that a few things should be changed related to LayoutRecall callback handling. Here are the things fixed by this patch. - For two of the three cases that LayoutRecall is done, I now think setting the clora_changed argument false is correct. - All errors other than NFSERR_DELAY returned by LayoutRecall appear permanent, so don't retry for any of them. (NFSERR_DELAY is retried by newnfs_request(), so it is not affected by this patch.) - Instead of waiting "forever" (actually until the process is SIGTERM'd) for Layouts to be returned during a mirror copy, fail and return ENXIO after about 1minute. Waiting for a <ctrl>C made sense when pnfsdscopymr() was done by itself, but did not make sense when done via find(1). This patch only affects the pNFS server.
2018-08-08 20:21:45 +00:00
if (error != 0) {
printf("nfsrv_recalllayout: getclient err=%d\n", error);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
return (error);
Assorted fixes to handling of LayoutRecall callbacks, mostly error handling. After a re-read of the appropriate section of RFC5661, I decided that a few things should be changed related to LayoutRecall callback handling. Here are the things fixed by this patch. - For two of the three cases that LayoutRecall is done, I now think setting the clora_changed argument false is correct. - All errors other than NFSERR_DELAY returned by LayoutRecall appear permanent, so don't retry for any of them. (NFSERR_DELAY is retried by newnfs_request(), so it is not affected by this patch.) - Instead of waiting "forever" (actually until the process is SIGTERM'd) for Layouts to be returned during a mirror copy, fail and return ENXIO after about 1minute. Waiting for a <ctrl>C made sense when pnfsdscopymr() was done by itself, but did not make sense when done via find(1). This patch only affects the pNFS server.
2018-08-08 20:21:45 +00:00
}
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
if ((clp->lc_flags & LCL_NFSV41) != 0) {
error = nfsrv_docallback(clp, NFSV4OP_CBLAYOUTRECALL,
Assorted fixes to handling of LayoutRecall callbacks, mostly error handling. After a re-read of the appropriate section of RFC5661, I decided that a few things should be changed related to LayoutRecall callback handling. Here are the things fixed by this patch. - For two of the three cases that LayoutRecall is done, I now think setting the clora_changed argument false is correct. - All errors other than NFSERR_DELAY returned by LayoutRecall appear permanent, so don't retry for any of them. (NFSERR_DELAY is retried by newnfs_request(), so it is not affected by this patch.) - Instead of waiting "forever" (actually until the process is SIGTERM'd) for Layouts to be returned during a mirror copy, fail and return ENXIO after about 1minute. Waiting for a <ctrl>C made sense when pnfsdscopymr() was done by itself, but did not make sense when done via find(1). This patch only affects the pNFS server.
2018-08-08 20:21:45 +00:00
stateidp, changed, fhp, NULL, NULL, laytype, p);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
/* If lyp != NULL, handle an error return here. */
if (error != 0 && lyp != NULL) {
NFSDRECALLLOCK();
Assorted fixes to handling of LayoutRecall callbacks, mostly error handling. After a re-read of the appropriate section of RFC5661, I decided that a few things should be changed related to LayoutRecall callback handling. Here are the things fixed by this patch. - For two of the three cases that LayoutRecall is done, I now think setting the clora_changed argument false is correct. - All errors other than NFSERR_DELAY returned by LayoutRecall appear permanent, so don't retry for any of them. (NFSERR_DELAY is retried by newnfs_request(), so it is not affected by this patch.) - Instead of waiting "forever" (actually until the process is SIGTERM'd) for Layouts to be returned during a mirror copy, fail and return ENXIO after about 1minute. Waiting for a <ctrl>C made sense when pnfsdscopymr() was done by itself, but did not make sense when done via find(1). This patch only affects the pNFS server.
2018-08-08 20:21:45 +00:00
/*
* Mark it returned, since no layout recall
* has been done.
* All errors seem to be non-recoverable, although
* NFSERR_NOMATCHLAYOUT is a normal event.
*/
if ((lyp->lay_flags & NFSLAY_RECALL) != 0) {
lyp->lay_flags |= NFSLAY_RETURNED;
wakeup(lyp);
}
NFSDRECALLUNLOCK();
if (error != NFSERR_NOMATCHLAYOUT)
printf("nfsrv_recalllayout: err=%d\n", error);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
}
} else
printf("nfsrv_recalllayout: clp not NFSv4.1\n");
return (error);
}
/*
* Find a layout to recall when we exceed our high water mark.
*/
void
nfsrv_recalloldlayout(NFSPROC_T *p)
{
struct nfslayouthash *lhyp;
struct nfslayout *lyp;
nfsquad_t clientid;
nfsv4stateid_t stateid;
fhandle_t fh;
Silence some "might not be initialized" warnings for riscv64. None of these case were actually using the variable(s) uninitialized, but I figured that silencing the warnings via initializing them made sense. Some of these predated r355677.
2019-12-13 21:38:08 +00:00
int error, laytype = 0, ret;
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
lhyp = &nfslayouthash[arc4random() % nfsrv_layouthashsize];
NFSLOCKLAYOUT(lhyp);
TAILQ_FOREACH_REVERSE(lyp, &lhyp->list, nfslayouthead, lay_list) {
if ((lyp->lay_flags & NFSLAY_CALLB) == 0) {
lyp->lay_flags |= NFSLAY_CALLB;
/*
* The layout stateid.seqid needs to be incremented
* before doing a LAYOUT_RECALL callback.
*/
if (++lyp->lay_stateid.seqid == 0)
lyp->lay_stateid.seqid = 1;
clientid = lyp->lay_clientid;
stateid = lyp->lay_stateid;
Copy all bits of a file handle in case there is padding in the structure. At least on x86, fhandle_t is a packed structure, so I believe an assignment will copy all the bits. However, for some current/future architectures, there might be padding in the structure that doesn't get copied via an assignment. Since NFS assumes a file handle is an opaque blob of bits that can be compared via memcmp()/bcmp(), all the bits including any padding must be copied. This patch replaces the assignments with a call to a byte copy function. Spotted during code inspection.
2018-08-05 19:21:50 +00:00
NFSBCOPY(&lyp->lay_fh, &fh, sizeof(fh));
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
laytype = lyp->lay_type;
break;
}
}
NFSUNLOCKLAYOUT(lhyp);
if (lyp != NULL) {
Assorted fixes to handling of LayoutRecall callbacks, mostly error handling. After a re-read of the appropriate section of RFC5661, I decided that a few things should be changed related to LayoutRecall callback handling. Here are the things fixed by this patch. - For two of the three cases that LayoutRecall is done, I now think setting the clora_changed argument false is correct. - All errors other than NFSERR_DELAY returned by LayoutRecall appear permanent, so don't retry for any of them. (NFSERR_DELAY is retried by newnfs_request(), so it is not affected by this patch.) - Instead of waiting "forever" (actually until the process is SIGTERM'd) for Layouts to be returned during a mirror copy, fail and return ENXIO after about 1minute. Waiting for a <ctrl>C made sense when pnfsdscopymr() was done by itself, but did not make sense when done via find(1). This patch only affects the pNFS server.
2018-08-08 20:21:45 +00:00
error = nfsrv_recalllayout(clientid, &stateid, &fh, NULL, 0,
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
laytype, p);
if (error != 0 && error != NFSERR_NOMATCHLAYOUT)
Assorted fixes to handling of LayoutRecall callbacks, mostly error handling. After a re-read of the appropriate section of RFC5661, I decided that a few things should be changed related to LayoutRecall callback handling. Here are the things fixed by this patch. - For two of the three cases that LayoutRecall is done, I now think setting the clora_changed argument false is correct. - All errors other than NFSERR_DELAY returned by LayoutRecall appear permanent, so don't retry for any of them. (NFSERR_DELAY is retried by newnfs_request(), so it is not affected by this patch.) - Instead of waiting "forever" (actually until the process is SIGTERM'd) for Layouts to be returned during a mirror copy, fail and return ENXIO after about 1minute. Waiting for a <ctrl>C made sense when pnfsdscopymr() was done by itself, but did not make sense when done via find(1). This patch only affects the pNFS server.
2018-08-08 20:21:45 +00:00
NFSD_DEBUG(4, "recallold=%d\n", error);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
if (error != 0) {
NFSLOCKLAYOUT(lhyp);
/*
* Since the hash list was unlocked, we need to
* find it again.
*/
ret = nfsrv_findlayout(&clientid, &fh, laytype, p,
&lyp);
if (ret == 0 &&
(lyp->lay_flags & NFSLAY_CALLB) != 0 &&
lyp->lay_stateid.other[0] == stateid.other[0] &&
lyp->lay_stateid.other[1] == stateid.other[1] &&
lyp->lay_stateid.other[2] == stateid.other[2]) {
/*
* The client no longer knows this layout, so
* it can be free'd now.
*/
if (error == NFSERR_NOMATCHLAYOUT)
nfsrv_freelayout(&lhyp->list, lyp);
else {
/*
* Leave it to be tried later by
* clearing NFSLAY_CALLB and moving
* it to the head of the list, so it
* won't be tried again for a while.
*/
lyp->lay_flags &= ~NFSLAY_CALLB;
TAILQ_REMOVE(&lhyp->list, lyp,
lay_list);
TAILQ_INSERT_HEAD(&lhyp->list, lyp,
lay_list);
}
}
NFSUNLOCKLAYOUT(lhyp);
}
}
}
/*
* Try and return layout(s).
*/
int
nfsrv_layoutreturn(struct nfsrv_descript *nd, vnode_t vp,
int layouttype, int iomode, uint64_t offset, uint64_t len, int reclaim,
int kind, nfsv4stateid_t *stateidp, int maxcnt, uint32_t *layp, int *fndp,
struct ucred *cred, NFSPROC_T *p)
{
struct nfsvattr na;
struct nfslayouthash *lhyp;
struct nfslayout *lyp;
fhandle_t fh;
int error = 0;
*fndp = 0;
if (kind == NFSV4LAYOUTRET_FILE) {
error = nfsvno_getfh(vp, &fh, p);
if (error == 0) {
error = nfsrv_updatemdsattr(vp, &na, p);
if (error != 0)
printf("nfsrv_layoutreturn: updatemdsattr"
" failed=%d\n", error);
}
if (error == 0) {
if (reclaim == newnfs_true) {
error = nfsrv_checkgrace(NULL, NULL,
NFSLCK_RECLAIM);
if (error != NFSERR_NOGRACE)
error = 0;
return (error);
}
lhyp = NFSLAYOUTHASH(&fh);
NFSDRECALLLOCK();
NFSLOCKLAYOUT(lhyp);
error = nfsrv_findlayout(&nd->nd_clientid, &fh,
layouttype, p, &lyp);
NFSD_DEBUG(4, "layoutret findlay=%d\n", error);
if (error == 0 &&
stateidp->other[0] == lyp->lay_stateid.other[0] &&
stateidp->other[1] == lyp->lay_stateid.other[1] &&
stateidp->other[2] == lyp->lay_stateid.other[2]) {
NFSD_DEBUG(4, "nfsrv_layoutreturn: stateid %d"
" %x %x %x laystateid %d %x %x %x"
" off=%ju len=%ju flgs=0x%x\n",
stateidp->seqid, stateidp->other[0],
stateidp->other[1], stateidp->other[2],
lyp->lay_stateid.seqid,
lyp->lay_stateid.other[0],
lyp->lay_stateid.other[1],
lyp->lay_stateid.other[2],
(uintmax_t)offset, (uintmax_t)len,
lyp->lay_flags);
if (++lyp->lay_stateid.seqid == 0)
lyp->lay_stateid.seqid = 1;
stateidp->seqid = lyp->lay_stateid.seqid;
if (offset == 0 && len == UINT64_MAX) {
if ((iomode & NFSLAYOUTIOMODE_READ) !=
0)
lyp->lay_flags &= ~NFSLAY_READ;
if ((iomode & NFSLAYOUTIOMODE_RW) != 0)
lyp->lay_flags &= ~NFSLAY_RW;
if ((lyp->lay_flags & (NFSLAY_READ |
NFSLAY_RW)) == 0)
nfsrv_freelayout(&lhyp->list,
lyp);
else
*fndp = 1;
} else
*fndp = 1;
}
NFSUNLOCKLAYOUT(lhyp);
/* Search the nfsrv_recalllist for a match. */
TAILQ_FOREACH(lyp, &nfsrv_recalllisthead, lay_list) {
if (NFSBCMP(&lyp->lay_fh, &fh,
sizeof(fh)) == 0 &&
lyp->lay_clientid.qval ==
nd->nd_clientid.qval &&
stateidp->other[0] ==
lyp->lay_stateid.other[0] &&
stateidp->other[1] ==
lyp->lay_stateid.other[1] &&
stateidp->other[2] ==
lyp->lay_stateid.other[2]) {
lyp->lay_flags |= NFSLAY_RETURNED;
wakeup(lyp);
error = 0;
}
}
NFSDRECALLUNLOCK();
}
nfsd: Add checks for layout errors in LayoutReturn For a LayoutReturn when using the Flexible File Layout, error reports may be provided in the request. Sanity check the size of these error reports and check that they exist before calling nfsrv_flexlayouterr(). Reported by: rtm@lcs.mit.edu Tested by: rtm@lcs.mit.edu PR: 260012 MFC after: 2 weeks
2021-11-26 23:42:32 +00:00
if (layouttype == NFSLAYOUT_FLEXFILE && layp != NULL)
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
nfsrv_flexlayouterr(nd, layp, maxcnt, p);
} else if (kind == NFSV4LAYOUTRET_FSID)
nfsrv_freelayouts(&nd->nd_clientid,
&vp->v_mount->mnt_stat.f_fsid, layouttype, iomode);
else if (kind == NFSV4LAYOUTRET_ALL)
nfsrv_freelayouts(&nd->nd_clientid, NULL, layouttype, iomode);
else
error = NFSERR_INVAL;
if (error == -1)
error = 0;
return (error);
}
/*
* Look for an existing layout.
*/
static int
nfsrv_findlayout(nfsquad_t *clientidp, fhandle_t *fhp, int laytype,
NFSPROC_T *p, struct nfslayout **lypp)
{
struct nfslayouthash *lhyp;
struct nfslayout *lyp;
int ret;
*lypp = NULL;
ret = 0;
lhyp = NFSLAYOUTHASH(fhp);
TAILQ_FOREACH(lyp, &lhyp->list, lay_list) {
if (NFSBCMP(&lyp->lay_fh, fhp, sizeof(*fhp)) == 0 &&
lyp->lay_clientid.qval == clientidp->qval &&
lyp->lay_type == laytype)
break;
}
if (lyp != NULL)
*lypp = lyp;
else
ret = -1;
return (ret);
}
/*
* Add the new layout, as required.
*/
static int
nfsrv_addlayout(struct nfsrv_descript *nd, struct nfslayout **lypp,
nfsv4stateid_t *stateidp, char *layp, int *layoutlenp, NFSPROC_T *p)
{
struct nfsclient *clp;
struct nfslayouthash *lhyp;
struct nfslayout *lyp, *nlyp;
fhandle_t *fhp;
int error;
KASSERT((nd->nd_flag & ND_IMPLIEDCLID) != 0,
("nfsrv_layoutget: no nd_clientid\n"));
lyp = *lypp;
fhp = &lyp->lay_fh;
NFSLOCKSTATE();
error = nfsrv_getclient((nfsquad_t)((u_quad_t)0), CLOPS_RENEW, &clp,
NULL, (nfsquad_t)((u_quad_t)0), 0, nd, p);
if (error != 0) {
NFSUNLOCKSTATE();
return (error);
}
lyp->lay_stateid.seqid = stateidp->seqid = 1;
lyp->lay_stateid.other[0] = stateidp->other[0] =
clp->lc_clientid.lval[0];
lyp->lay_stateid.other[1] = stateidp->other[1] =
clp->lc_clientid.lval[1];
lyp->lay_stateid.other[2] = stateidp->other[2] =
nfsrv_nextstateindex(clp);
NFSUNLOCKSTATE();
lhyp = NFSLAYOUTHASH(fhp);
NFSLOCKLAYOUT(lhyp);
TAILQ_FOREACH(nlyp, &lhyp->list, lay_list) {
if (NFSBCMP(&nlyp->lay_fh, fhp, sizeof(*fhp)) == 0 &&
nlyp->lay_clientid.qval == nd->nd_clientid.qval)
break;
}
if (nlyp != NULL) {
/* A layout already exists, so use it. */
nlyp->lay_flags |= (lyp->lay_flags & (NFSLAY_READ | NFSLAY_RW));
NFSBCOPY(nlyp->lay_xdr, layp, nlyp->lay_layoutlen);
*layoutlenp = nlyp->lay_layoutlen;
if (++nlyp->lay_stateid.seqid == 0)
nlyp->lay_stateid.seqid = 1;
stateidp->seqid = nlyp->lay_stateid.seqid;
stateidp->other[0] = nlyp->lay_stateid.other[0];
stateidp->other[1] = nlyp->lay_stateid.other[1];
stateidp->other[2] = nlyp->lay_stateid.other[2];
NFSUNLOCKLAYOUT(lhyp);
return (0);
}
/* Insert the new layout in the lists. */
*lypp = NULL;
atomic_add_int(&nfsrv_layoutcnt, 1);
nfsd: Wrap nfsstatsv1_p in the NFSD_VNET() macro Commit 7344856e3a6d added a lot of macros that will front end vnet macros so that nfsd(8) can run in vnet prison. The nfsstatsv1_p variable got missed. This patch wraps all uses of nfsstatsv1_p with the NFSD_VNET() macro. The NFSD_VNET() macro is still a null macro. MFC after: 3 months
2023-02-16 01:39:07 +00:00
NFSD_VNET(nfsstatsv1_p)->srvlayouts++;
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
NFSBCOPY(lyp->lay_xdr, layp, lyp->lay_layoutlen);
*layoutlenp = lyp->lay_layoutlen;
TAILQ_INSERT_HEAD(&lhyp->list, lyp, lay_list);
NFSUNLOCKLAYOUT(lhyp);
return (0);
}
/*
* Get the devinfo for a deviceid.
*/
int
nfsrv_getdevinfo(char *devid, int layouttype, uint32_t *maxcnt,
uint32_t *notify, int *devaddrlen, char **devaddr)
{
struct nfsdevice *ds;
if ((layouttype != NFSLAYOUT_NFSV4_1_FILES && layouttype !=
NFSLAYOUT_FLEXFILE) ||
(nfsrv_maxpnfsmirror > 1 && layouttype == NFSLAYOUT_NFSV4_1_FILES))
return (NFSERR_UNKNLAYOUTTYPE);
/*
* Now, search for the device id. Note that the structures won't go
* away, but the order changes in the list. As such, the lock only
* needs to be held during the search through the list.
*/
NFSDDSLOCK();
TAILQ_FOREACH(ds, &nfsrv_devidhead, nfsdev_list) {
if (NFSBCMP(devid, ds->nfsdev_deviceid, NFSX_V4DEVICEID) == 0 &&
ds->nfsdev_nmp != NULL)
break;
}
NFSDDSUNLOCK();
if (ds == NULL)
return (NFSERR_NOENT);
/* If the correct nfsdev_XXXXaddrlen is > 0, we have the device info. */
*devaddrlen = 0;
if (layouttype == NFSLAYOUT_NFSV4_1_FILES) {
*devaddrlen = ds->nfsdev_fileaddrlen;
*devaddr = ds->nfsdev_fileaddr;
} else if (layouttype == NFSLAYOUT_FLEXFILE) {
*devaddrlen = ds->nfsdev_flexaddrlen;
*devaddr = ds->nfsdev_flexaddr;
}
if (*devaddrlen == 0)
return (NFSERR_UNKNLAYOUTTYPE);
/*
* The XDR overhead is 3 unsigned values: layout_type,
* length_of_address and notify bitmap.
* If the notify array is changed to not all zeros, the
* count of unsigned values must be increased.
*/
if (*maxcnt > 0 && *maxcnt < NFSM_RNDUP(*devaddrlen) +
3 * NFSX_UNSIGNED) {
*maxcnt = NFSM_RNDUP(*devaddrlen) + 3 * NFSX_UNSIGNED;
return (NFSERR_TOOSMALL);
}
return (0);
}
/*
* Free a list of layout state structures.
*/
static void
nfsrv_freelayoutlist(nfsquad_t clientid)
{
struct nfslayouthash *lhyp;
struct nfslayout *lyp, *nlyp;
int i;
for (i = 0; i < nfsrv_layouthashsize; i++) {
lhyp = &nfslayouthash[i];
NFSLOCKLAYOUT(lhyp);
TAILQ_FOREACH_SAFE(lyp, &lhyp->list, lay_list, nlyp) {
if (lyp->lay_clientid.qval == clientid.qval)
nfsrv_freelayout(&lhyp->list, lyp);
}
NFSUNLOCKLAYOUT(lhyp);
}
}
/*
* Free up a layout.
*/
static void
nfsrv_freelayout(struct nfslayouthead *lhp, struct nfslayout *lyp)
{
NFSD_DEBUG(4, "Freelayout=%p\n", lyp);
atomic_add_int(&nfsrv_layoutcnt, -1);
nfsd: Wrap nfsstatsv1_p in the NFSD_VNET() macro Commit 7344856e3a6d added a lot of macros that will front end vnet macros so that nfsd(8) can run in vnet prison. The nfsstatsv1_p variable got missed. This patch wraps all uses of nfsstatsv1_p with the NFSD_VNET() macro. The NFSD_VNET() macro is still a null macro. MFC after: 3 months
2023-02-16 01:39:07 +00:00
NFSD_VNET(nfsstatsv1_p)->srvlayouts--;
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
TAILQ_REMOVE(lhp, lyp, lay_list);
free(lyp, M_NFSDSTATE);
}
/*
* Free up a device id.
*/
void
nfsrv_freeonedevid(struct nfsdevice *ds)
{
int i;
atomic_add_int(&nfsrv_devidcnt, -1);
vrele(ds->nfsdev_dvp);
for (i = 0; i < nfsrv_dsdirsize; i++)
if (ds->nfsdev_dsdir[i] != NULL)
vrele(ds->nfsdev_dsdir[i]);
free(ds->nfsdev_fileaddr, M_NFSDSTATE);
free(ds->nfsdev_flexaddr, M_NFSDSTATE);
free(ds->nfsdev_host, M_NFSDSTATE);
free(ds, M_NFSDSTATE);
}
/*
* Free up a device id and its mirrors.
*/
static void
nfsrv_freedevid(struct nfsdevice *ds)
{
TAILQ_REMOVE(&nfsrv_devidhead, ds, nfsdev_list);
nfsrv_freeonedevid(ds);
}
/*
* Free all layouts and device ids.
* Done when the nfsd threads are shut down since there may be a new
* modified device id list created when the nfsd is restarted.
*/
void
nfsrv_freealllayoutsanddevids(void)
{
struct nfsdontlist *mrp, *nmrp;
struct nfslayout *lyp, *nlyp;
/* Get rid of the deviceid structures. */
nfsrv_freealldevids();
TAILQ_INIT(&nfsrv_devidhead);
nfsrv_devidcnt = 0;
/* Get rid of all layouts. */
nfsrv_freealllayouts();
/* Get rid of any nfsdontlist entries. */
LIST_FOREACH_SAFE(mrp, &nfsrv_dontlisthead, nfsmr_list, nmrp)
free(mrp, M_NFSDSTATE);
LIST_INIT(&nfsrv_dontlisthead);
nfsrv_dontlistlen = 0;
/* Free layouts in the recall list. */
TAILQ_FOREACH_SAFE(lyp, &nfsrv_recalllisthead, lay_list, nlyp)
nfsrv_freelayout(&nfsrv_recalllisthead, lyp);
TAILQ_INIT(&nfsrv_recalllisthead);
}
/*
* Free layouts that match the arguments.
*/
static void
nfsrv_freelayouts(nfsquad_t *clid, fsid_t *fs, int laytype, int iomode)
{
struct nfslayouthash *lhyp;
struct nfslayout *lyp, *nlyp;
int i;
for (i = 0; i < nfsrv_layouthashsize; i++) {
lhyp = &nfslayouthash[i];
NFSLOCKLAYOUT(lhyp);
TAILQ_FOREACH_SAFE(lyp, &lhyp->list, lay_list, nlyp) {
if (clid->qval != lyp->lay_clientid.qval)
continue;
Deduplicate fsid comparisons Comparing fsid_t objects requires internal knowledge of the fsid structure and yet this is duplicated across a number of places in the code. Simplify by creating a fsidcmp function (macro). Reviewed by: mjg, rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24749
2020-05-21 01:55:35 +00:00
if (fs != NULL && fsidcmp(fs, &lyp->lay_fsid) != 0)
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
continue;
if (laytype != lyp->lay_type)
continue;
if ((iomode & NFSLAYOUTIOMODE_READ) != 0)
lyp->lay_flags &= ~NFSLAY_READ;
if ((iomode & NFSLAYOUTIOMODE_RW) != 0)
lyp->lay_flags &= ~NFSLAY_RW;
if ((lyp->lay_flags & (NFSLAY_READ | NFSLAY_RW)) == 0)
nfsrv_freelayout(&lhyp->list, lyp);
}
NFSUNLOCKLAYOUT(lhyp);
}
}
/*
* Free all layouts for the argument file.
*/
void
nfsrv_freefilelayouts(fhandle_t *fhp)
{
struct nfslayouthash *lhyp;
struct nfslayout *lyp, *nlyp;
lhyp = NFSLAYOUTHASH(fhp);
NFSLOCKLAYOUT(lhyp);
TAILQ_FOREACH_SAFE(lyp, &lhyp->list, lay_list, nlyp) {
if (NFSBCMP(&lyp->lay_fh, fhp, sizeof(*fhp)) == 0)
nfsrv_freelayout(&lhyp->list, lyp);
}
NFSUNLOCKLAYOUT(lhyp);
}
/*
* Free all layouts.
*/
static void
nfsrv_freealllayouts(void)
{
struct nfslayouthash *lhyp;
struct nfslayout *lyp, *nlyp;
int i;
for (i = 0; i < nfsrv_layouthashsize; i++) {
lhyp = &nfslayouthash[i];
NFSLOCKLAYOUT(lhyp);
TAILQ_FOREACH_SAFE(lyp, &lhyp->list, lay_list, nlyp)
nfsrv_freelayout(&lhyp->list, lyp);
NFSUNLOCKLAYOUT(lhyp);
}
}
/*
* Look up the mount path for the DS server.
*/
static int
Add an optional feature to the pNFS server. Without this patch, the pNFS server distributes the data storage files across all of the specified DSs. A tester noted that it would be nice if a system administrator could control which DSs are used to store the file data for a given exported MDS file system. This patch adds the kernel support to do this. It also makes a slight semantic change to nfsv4_findmirror(), since some uses of it no longer require that the DS being searched for have a current mirror. A patch that will be committed in a few minutes will modify the nfsd daemon to support this feature. The patch should only affect sites using the pNFS server (specified via the "-p" command line option for nfsd. Suggested by: james.rose@framestore.com
2018-07-02 19:21:33 +00:00
nfsrv_setdsserver(char *dspathp, char *mdspathp, NFSPROC_T *p,
struct nfsdevice **dsp)
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
{
struct nameidata nd;
struct nfsdevice *ds;
Add an optional feature to the pNFS server. Without this patch, the pNFS server distributes the data storage files across all of the specified DSs. A tester noted that it would be nice if a system administrator could control which DSs are used to store the file data for a given exported MDS file system. This patch adds the kernel support to do this. It also makes a slight semantic change to nfsv4_findmirror(), since some uses of it no longer require that the DS being searched for have a current mirror. A patch that will be committed in a few minutes will modify the nfsd daemon to support this feature. The patch should only affect sites using the pNFS server (specified via the "-p" command line option for nfsd. Suggested by: james.rose@framestore.com
2018-07-02 19:21:33 +00:00
struct mount *mp;
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
int error, i;
char *dsdirpath;
size_t dsdirsize;
NFSD_DEBUG(4, "setdssrv path=%s\n", dspathp);
*dsp = NULL;
nfsd: Prepare the NFS server code to run in a vnet prison This patch defines null macros that can be used to apply the vnet macros for global variables and SYSCTL flags. It also applies these macros to many of the global variables and some of the SYSCTLs. Since the macros do nothing, these changes should not result in semantics changes, although the changes are large in number. The patch does change several global variables that were arrays or structures to pointers to same. For these variables, modified initialization and cleanup code malloc's and free's the arrays/structures. This was done so that the vnet footprint would be about 300bytes when the macros are defined as vnet macros, allowing nfsd.ko to load dynamically. I believe the comments in D37519 have been addressed, although it has never been reviewed, due in part to the large size of the patch. This is the first of a series of patches that will put D37519 in main. Once everything is in main, the macros will be defined as front end macros to the vnet ones. MFC after: 3 months Differential Revision: https://reviews.freebsd.org/D37519
2023-02-11 23:51:19 +00:00
if (jailed(p->td_ucred)) {
printf("A pNFS nfsd cannot run in a jail\n");
return (EPERM);
}
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF, UIO_SYSSPACE,
vfs: remove the unused thread argument from NDINIT* See b4a58fbf640409a1 ("vfs: remove cn_thread") Bump __FreeBSD_version to 1400043.
2021-11-25 21:43:06 +00:00
dspathp);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
error = namei(&nd);
NFSD_DEBUG(4, "lookup=%d\n", error);
if (error != 0)
return (error);
if (nd.ni_vp->v_type != VDIR) {
vput(nd.ni_vp);
NFSD_DEBUG(4, "dspath not dir\n");
return (ENOTDIR);
}
if (strcmp(nd.ni_vp->v_mount->mnt_vfc->vfc_name, "nfs") != 0) {
vput(nd.ni_vp);
NFSD_DEBUG(4, "dspath not an NFS mount\n");
return (ENXIO);
}
/*
* Allocate a DS server structure with the NFS mounted directory
* vnode reference counted, so that a non-forced dismount will
* fail with EBUSY.
Add an optional feature to the pNFS server. Without this patch, the pNFS server distributes the data storage files across all of the specified DSs. A tester noted that it would be nice if a system administrator could control which DSs are used to store the file data for a given exported MDS file system. This patch adds the kernel support to do this. It also makes a slight semantic change to nfsv4_findmirror(), since some uses of it no longer require that the DS being searched for have a current mirror. A patch that will be committed in a few minutes will modify the nfsd daemon to support this feature. The patch should only affect sites using the pNFS server (specified via the "-p" command line option for nfsd. Suggested by: james.rose@framestore.com
2018-07-02 19:21:33 +00:00
* This structure is always linked into the list, even if an error
* is being returned. The caller will free the entire list upon
* an error return.
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
*/
*dsp = ds = malloc(sizeof(*ds) + nfsrv_dsdirsize * sizeof(vnode_t),
M_NFSDSTATE, M_WAITOK | M_ZERO);
ds->nfsdev_dvp = nd.ni_vp;
ds->nfsdev_nmp = VFSTONFS(nd.ni_vp->v_mount);
vfs: drop the mostly unused flags argument from VOP_UNLOCK Filesystems which want to use it in limited capacity can employ the VOP_UNLOCK_FLAGS macro. Reviewed by: kib (previous version) Differential Revision: https://reviews.freebsd.org/D21427
2020-01-03 22:29:58 +00:00
NFSVOPUNLOCK(nd.ni_vp);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
dsdirsize = strlen(dspathp) + 16;
dsdirpath = malloc(dsdirsize, M_TEMP, M_WAITOK);
/* Now, create the DS directory structures. */
for (i = 0; i < nfsrv_dsdirsize; i++) {
snprintf(dsdirpath, dsdirsize, "%s/ds%d", dspathp, i);
NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF,
vfs: remove the unused thread argument from NDINIT* See b4a58fbf640409a1 ("vfs: remove cn_thread") Bump __FreeBSD_version to 1400043.
2021-11-25 21:43:06 +00:00
UIO_SYSSPACE, dsdirpath);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
error = namei(&nd);
NFSD_DEBUG(4, "dsdirpath=%s lookup=%d\n", dsdirpath, error);
if (error != 0)
break;
if (nd.ni_vp->v_type != VDIR) {
vput(nd.ni_vp);
error = ENOTDIR;
NFSD_DEBUG(4, "dsdirpath not a VDIR\n");
break;
}
if (strcmp(nd.ni_vp->v_mount->mnt_vfc->vfc_name, "nfs") != 0) {
vput(nd.ni_vp);
error = ENXIO;
NFSD_DEBUG(4, "dsdirpath not an NFS mount\n");
break;
}
ds->nfsdev_dsdir[i] = nd.ni_vp;
vfs: drop the mostly unused flags argument from VOP_UNLOCK Filesystems which want to use it in limited capacity can employ the VOP_UNLOCK_FLAGS macro. Reviewed by: kib (previous version) Differential Revision: https://reviews.freebsd.org/D21427
2020-01-03 22:29:58 +00:00
NFSVOPUNLOCK(nd.ni_vp);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
}
free(dsdirpath, M_TEMP);
Add an optional feature to the pNFS server. Without this patch, the pNFS server distributes the data storage files across all of the specified DSs. A tester noted that it would be nice if a system administrator could control which DSs are used to store the file data for a given exported MDS file system. This patch adds the kernel support to do this. It also makes a slight semantic change to nfsv4_findmirror(), since some uses of it no longer require that the DS being searched for have a current mirror. A patch that will be committed in a few minutes will modify the nfsd daemon to support this feature. The patch should only affect sites using the pNFS server (specified via the "-p" command line option for nfsd. Suggested by: james.rose@framestore.com
2018-07-02 19:21:33 +00:00
if (strlen(mdspathp) > 0) {
/*
* This DS stores file for a specific MDS exported file
* system.
*/
NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF,
vfs: remove the unused thread argument from NDINIT* See b4a58fbf640409a1 ("vfs: remove cn_thread") Bump __FreeBSD_version to 1400043.
2021-11-25 21:43:06 +00:00
UIO_SYSSPACE, mdspathp);
Add an optional feature to the pNFS server. Without this patch, the pNFS server distributes the data storage files across all of the specified DSs. A tester noted that it would be nice if a system administrator could control which DSs are used to store the file data for a given exported MDS file system. This patch adds the kernel support to do this. It also makes a slight semantic change to nfsv4_findmirror(), since some uses of it no longer require that the DS being searched for have a current mirror. A patch that will be committed in a few minutes will modify the nfsd daemon to support this feature. The patch should only affect sites using the pNFS server (specified via the "-p" command line option for nfsd. Suggested by: james.rose@framestore.com
2018-07-02 19:21:33 +00:00
error = namei(&nd);
NFSD_DEBUG(4, "mds lookup=%d\n", error);
if (error != 0)
goto out;
if (nd.ni_vp->v_type != VDIR) {
vput(nd.ni_vp);
error = ENOTDIR;
NFSD_DEBUG(4, "mdspath not dir\n");
goto out;
}
mp = nd.ni_vp->v_mount;
if ((mp->mnt_flag & MNT_EXPORTED) == 0) {
vput(nd.ni_vp);
error = ENXIO;
NFSD_DEBUG(4, "mdspath not an exported fs\n");
goto out;
}
ds->nfsdev_mdsfsid = mp->mnt_stat.f_fsid;
ds->nfsdev_mdsisset = 1;
vput(nd.ni_vp);
}
out:
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
TAILQ_INSERT_TAIL(&nfsrv_devidhead, ds, nfsdev_list);
atomic_add_int(&nfsrv_devidcnt, 1);
return (error);
}
/*
* Look up the mount path for the DS server and delete it.
*/
int
Add support for a "forced" pnfsdskill to the pNFS server kernel code. The pnfsdskill(8) command will normally fail if there is no valid mirror for the DS to be disabled. However, a system administrator may need to disable a DS which does not have a valid mirror so that the nfsd threads can be terminated. This patch adds the kernel code needed by pnfsdskill(8) to implement this "forced" case of disabling a DS. This patch only affects the pNFS server.
2018-07-09 19:58:01 +00:00
nfsrv_deldsserver(int op, char *dspathp, NFSPROC_T *p)
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
{
struct mount *mp;
struct nfsmount *nmp;
struct nfsdevice *ds;
int error;
NFSD_DEBUG(4, "deldssrv path=%s\n", dspathp);
/*
* Search for the path in the mount list. Avoid looking the path
* up, since this mount point may be hung, with associated locked
* vnodes, etc.
* Set NFSMNTP_CANCELRPCS so that any forced dismount will be blocked
* until this completes.
* As noted in the man page, this should be done before any forced
* dismount on the mount point, but at least the handshake on
* NFSMNTP_CANCELRPCS should make it safe.
*/
error = 0;
ds = NULL;
nmp = NULL;
mtx_lock(&mountlist_mtx);
TAILQ_FOREACH(mp, &mountlist, mnt_list) {
if (strcmp(mp->mnt_stat.f_mntonname, dspathp) == 0 &&
strcmp(mp->mnt_stat.f_fstypename, "nfs") == 0 &&
mp->mnt_data != NULL) {
nmp = VFSTONFS(mp);
NFSLOCKMNT(nmp);
if ((nmp->nm_privflag & (NFSMNTP_FORCEDISM |
NFSMNTP_CANCELRPCS)) == 0) {
nmp->nm_privflag |= NFSMNTP_CANCELRPCS;
NFSUNLOCKMNT(nmp);
} else {
NFSUNLOCKMNT(nmp);
nmp = NULL;
}
break;
}
}
mtx_unlock(&mountlist_mtx);
if (nmp != NULL) {
Add support for a "forced" pnfsdskill to the pNFS server kernel code. The pnfsdskill(8) command will normally fail if there is no valid mirror for the DS to be disabled. However, a system administrator may need to disable a DS which does not have a valid mirror so that the nfsd threads can be terminated. This patch adds the kernel code needed by pnfsdskill(8) to implement this "forced" case of disabling a DS. This patch only affects the pNFS server.
2018-07-09 19:58:01 +00:00
ds = nfsrv_deldsnmp(op, nmp, p);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
NFSD_DEBUG(4, "deldsnmp=%p\n", ds);
if (ds != NULL) {
nfsrv_killrpcs(nmp);
NFSD_DEBUG(4, "aft killrpcs\n");
} else
error = ENXIO;
NFSLOCKMNT(nmp);
nmp->nm_privflag &= ~NFSMNTP_CANCELRPCS;
wakeup(nmp);
NFSUNLOCKMNT(nmp);
} else
error = EINVAL;
return (error);
}
/*
* Search for and remove a DS entry which matches the "nmp" argument.
* The nfsdevice structure pointer is returned so that the caller can
* free it via nfsrv_freeonedevid().
Add support for a "forced" pnfsdskill to the pNFS server kernel code. The pnfsdskill(8) command will normally fail if there is no valid mirror for the DS to be disabled. However, a system administrator may need to disable a DS which does not have a valid mirror so that the nfsd threads can be terminated. This patch adds the kernel code needed by pnfsdskill(8) to implement this "forced" case of disabling a DS. This patch only affects the pNFS server.
2018-07-09 19:58:01 +00:00
* For the forced case, do not try to do LayoutRecalls, since the server
* must be shut down now anyhow.
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
*/
struct nfsdevice *
Add support for a "forced" pnfsdskill to the pNFS server kernel code. The pnfsdskill(8) command will normally fail if there is no valid mirror for the DS to be disabled. However, a system administrator may need to disable a DS which does not have a valid mirror so that the nfsd threads can be terminated. This patch adds the kernel code needed by pnfsdskill(8) to implement this "forced" case of disabling a DS. This patch only affects the pNFS server.
2018-07-09 19:58:01 +00:00
nfsrv_deldsnmp(int op, struct nfsmount *nmp, NFSPROC_T *p)
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
{
struct nfsdevice *fndds;
NFSD_DEBUG(4, "deldsdvp\n");
NFSDDSLOCK();
Add support for a "forced" pnfsdskill to the pNFS server kernel code. The pnfsdskill(8) command will normally fail if there is no valid mirror for the DS to be disabled. However, a system administrator may need to disable a DS which does not have a valid mirror so that the nfsd threads can be terminated. This patch adds the kernel code needed by pnfsdskill(8) to implement this "forced" case of disabling a DS. This patch only affects the pNFS server.
2018-07-09 19:58:01 +00:00
if (op == PNFSDOP_FORCEDELDS)
fndds = nfsv4_findmirror(nmp);
else
fndds = nfsrv_findmirroredds(nmp);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
if (fndds != NULL)
nfsrv_deleteds(fndds);
NFSDDSUNLOCK();
if (fndds != NULL) {
Add support for a "forced" pnfsdskill to the pNFS server kernel code. The pnfsdskill(8) command will normally fail if there is no valid mirror for the DS to be disabled. However, a system administrator may need to disable a DS which does not have a valid mirror so that the nfsd threads can be terminated. This patch adds the kernel code needed by pnfsdskill(8) to implement this "forced" case of disabling a DS. This patch only affects the pNFS server.
2018-07-09 19:58:01 +00:00
if (op != PNFSDOP_FORCEDELDS)
nfsrv_flexmirrordel(fndds->nfsdev_deviceid, p);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
printf("pNFS server: mirror %s failed\n", fndds->nfsdev_host);
}
return (fndds);
}
/*
* Similar to nfsrv_deldsnmp(), except that the DS is indicated by deviceid.
* This function also calls nfsrv_killrpcs() to unblock RPCs on the mount
* point.
* Also, returns an error instead of the nfsdevice found.
*/
nfs: Remove APPLESTATIC macro It is no longer useful. Reviewed by: rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24811
2020-05-12 13:23:25 +00:00
int
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
nfsrv_delds(char *devid, NFSPROC_T *p)
{
struct nfsdevice *ds, *fndds;
struct nfsmount *nmp;
int fndmirror;
NFSD_DEBUG(4, "delds\n");
/*
* Search the DS server list for a match with devid.
* Remove the DS entry if found and there is a mirror.
*/
fndds = NULL;
nmp = NULL;
fndmirror = 0;
NFSDDSLOCK();
TAILQ_FOREACH(ds, &nfsrv_devidhead, nfsdev_list) {
if (NFSBCMP(ds->nfsdev_deviceid, devid, NFSX_V4DEVICEID) == 0 &&
ds->nfsdev_nmp != NULL) {
NFSD_DEBUG(4, "fnd main ds\n");
fndds = ds;
break;
Add an optional feature to the pNFS server. Without this patch, the pNFS server distributes the data storage files across all of the specified DSs. A tester noted that it would be nice if a system administrator could control which DSs are used to store the file data for a given exported MDS file system. This patch adds the kernel support to do this. It also makes a slight semantic change to nfsv4_findmirror(), since some uses of it no longer require that the DS being searched for have a current mirror. A patch that will be committed in a few minutes will modify the nfsd daemon to support this feature. The patch should only affect sites using the pNFS server (specified via the "-p" command line option for nfsd. Suggested by: james.rose@framestore.com
2018-07-02 19:21:33 +00:00
}
}
if (fndds == NULL) {
NFSDDSUNLOCK();
return (ENXIO);
}
if (fndds->nfsdev_mdsisset == 0 && nfsrv_faildscnt > 0)
fndmirror = 1;
Fix the pNFS server so that it handles the "#mds_path" check for mirrors. The recently added feature of the pNFS server will set an fsid for the MDS file system to define the file system a DS should store files for. For a case where a DS handling all file systems has failed, it was possible for the code to check for a mirror with a specified fs, even though nfsdev_mdsisset was 0, possibly causing a false successful check for a mirror. This patch adds a check for nfsdev_mdsisset != 0 to avoid this. It only affects the pNFS server for a rare case. Found via code inspection.
2018-07-04 19:46:26 +00:00
else if (fndds->nfsdev_mdsisset != 0) {
Add an optional feature to the pNFS server. Without this patch, the pNFS server distributes the data storage files across all of the specified DSs. A tester noted that it would be nice if a system administrator could control which DSs are used to store the file data for a given exported MDS file system. This patch adds the kernel support to do this. It also makes a slight semantic change to nfsv4_findmirror(), since some uses of it no longer require that the DS being searched for have a current mirror. A patch that will be committed in a few minutes will modify the nfsd daemon to support this feature. The patch should only affect sites using the pNFS server (specified via the "-p" command line option for nfsd. Suggested by: james.rose@framestore.com
2018-07-02 19:21:33 +00:00
/* For the fsid is set case, search for a mirror. */
TAILQ_FOREACH(ds, &nfsrv_devidhead, nfsdev_list) {
if (ds != fndds && ds->nfsdev_nmp != NULL &&
ds->nfsdev_mdsisset != 0 &&
Deduplicate fsid comparisons Comparing fsid_t objects requires internal knowledge of the fsid structure and yet this is duplicated across a number of places in the code. Simplify by creating a fsidcmp function (macro). Reviewed by: mjg, rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24749
2020-05-21 01:55:35 +00:00
fsidcmp(&ds->nfsdev_mdsfsid,
&fndds->nfsdev_mdsfsid) == 0) {
Add an optional feature to the pNFS server. Without this patch, the pNFS server distributes the data storage files across all of the specified DSs. A tester noted that it would be nice if a system administrator could control which DSs are used to store the file data for a given exported MDS file system. This patch adds the kernel support to do this. It also makes a slight semantic change to nfsv4_findmirror(), since some uses of it no longer require that the DS being searched for have a current mirror. A patch that will be committed in a few minutes will modify the nfsd daemon to support this feature. The patch should only affect sites using the pNFS server (specified via the "-p" command line option for nfsd. Suggested by: james.rose@framestore.com
2018-07-02 19:21:33 +00:00
fndmirror = 1;
break;
}
}
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
}
Add an optional feature to the pNFS server. Without this patch, the pNFS server distributes the data storage files across all of the specified DSs. A tester noted that it would be nice if a system administrator could control which DSs are used to store the file data for a given exported MDS file system. This patch adds the kernel support to do this. It also makes a slight semantic change to nfsv4_findmirror(), since some uses of it no longer require that the DS being searched for have a current mirror. A patch that will be committed in a few minutes will modify the nfsd daemon to support this feature. The patch should only affect sites using the pNFS server (specified via the "-p" command line option for nfsd. Suggested by: james.rose@framestore.com
2018-07-02 19:21:33 +00:00
if (fndmirror != 0) {
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
nmp = fndds->nfsdev_nmp;
NFSLOCKMNT(nmp);
if ((nmp->nm_privflag & (NFSMNTP_FORCEDISM |
NFSMNTP_CANCELRPCS)) == 0) {
nmp->nm_privflag |= NFSMNTP_CANCELRPCS;
NFSUNLOCKMNT(nmp);
nfsrv_deleteds(fndds);
} else {
NFSUNLOCKMNT(nmp);
nmp = NULL;
}
}
NFSDDSUNLOCK();
Add an optional feature to the pNFS server. Without this patch, the pNFS server distributes the data storage files across all of the specified DSs. A tester noted that it would be nice if a system administrator could control which DSs are used to store the file data for a given exported MDS file system. This patch adds the kernel support to do this. It also makes a slight semantic change to nfsv4_findmirror(), since some uses of it no longer require that the DS being searched for have a current mirror. A patch that will be committed in a few minutes will modify the nfsd daemon to support this feature. The patch should only affect sites using the pNFS server (specified via the "-p" command line option for nfsd. Suggested by: james.rose@framestore.com
2018-07-02 19:21:33 +00:00
if (nmp != NULL) {
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
nfsrv_flexmirrordel(fndds->nfsdev_deviceid, p);
printf("pNFS server: mirror %s failed\n", fndds->nfsdev_host);
nfsrv_killrpcs(nmp);
NFSLOCKMNT(nmp);
nmp->nm_privflag &= ~NFSMNTP_CANCELRPCS;
wakeup(nmp);
NFSUNLOCKMNT(nmp);
return (0);
}
return (ENXIO);
}
/*
* Mark a DS as disabled by setting nfsdev_nmp = NULL.
*/
static void
nfsrv_deleteds(struct nfsdevice *fndds)
{
NFSD_DEBUG(4, "deleteds: deleting a mirror\n");
fndds->nfsdev_nmp = NULL;
Add an optional feature to the pNFS server. Without this patch, the pNFS server distributes the data storage files across all of the specified DSs. A tester noted that it would be nice if a system administrator could control which DSs are used to store the file data for a given exported MDS file system. This patch adds the kernel support to do this. It also makes a slight semantic change to nfsv4_findmirror(), since some uses of it no longer require that the DS being searched for have a current mirror. A patch that will be committed in a few minutes will modify the nfsd daemon to support this feature. The patch should only affect sites using the pNFS server (specified via the "-p" command line option for nfsd. Suggested by: james.rose@framestore.com
2018-07-02 19:21:33 +00:00
if (fndds->nfsdev_mdsisset == 0)
nfsrv_faildscnt--;
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
}
/*
* Fill in the addr structures for the File and Flex File layouts.
*/
static void
nfsrv_allocdevid(struct nfsdevice *ds, char *addr, char *dnshost)
{
uint32_t *tl;
char *netprot;
int addrlen;
static uint64_t new_devid = 0;
if (strchr(addr, ':') != NULL)
netprot = "tcp6";
else
netprot = "tcp";
/* Fill in the device id. */
NFSBCOPY(&nfsdev_time, ds->nfsdev_deviceid, sizeof(nfsdev_time));
new_devid++;
NFSBCOPY(&new_devid, &ds->nfsdev_deviceid[sizeof(nfsdev_time)],
sizeof(new_devid));
/*
* Fill in the file addr (actually the nfsv4_file_layout_ds_addr4
* as defined in RFC5661) in XDR.
*/
addrlen = NFSM_RNDUP(strlen(addr)) + NFSM_RNDUP(strlen(netprot)) +
6 * NFSX_UNSIGNED;
NFSD_DEBUG(4, "hn=%s addr=%s netprot=%s\n", dnshost, addr, netprot);
ds->nfsdev_fileaddrlen = addrlen;
tl = malloc(addrlen, M_NFSDSTATE, M_WAITOK | M_ZERO);
ds->nfsdev_fileaddr = (char *)tl;
*tl++ = txdr_unsigned(1); /* One stripe with index 0. */
*tl++ = 0;
*tl++ = txdr_unsigned(1); /* One multipath list */
*tl++ = txdr_unsigned(1); /* with one entry in it. */
/* The netaddr for this one entry. */
*tl++ = txdr_unsigned(strlen(netprot));
NFSBCOPY(netprot, tl, strlen(netprot));
tl += (NFSM_RNDUP(strlen(netprot)) / NFSX_UNSIGNED);
*tl++ = txdr_unsigned(strlen(addr));
NFSBCOPY(addr, tl, strlen(addr));
/*
* Fill in the flex file addr (actually the ff_device_addr4
* as defined for Flexible File Layout) in XDR.
*/
addrlen = NFSM_RNDUP(strlen(addr)) + NFSM_RNDUP(strlen(netprot)) +
Add support for NFSv4.2 to the NFS client and server. This patch adds support for NFSv4.2 (RFC-7862) and Extended Attributes (RFC-8276) to the NFS client and server. NFSv4.2 is comprised of several optional features that can be supported in addition to NFSv4.1. This patch adds the following optional features: - posix_fadvise(POSIX_FADV_WILLNEED/POSIX_FADV_DONTNEED) - posix_fallocate() - intra server file range copying via the copy_file_range(2) syscall --> Avoiding data tranfer over the wire to/from the NFS client. - lseek(SEEK_DATA/SEEK_HOLE) - Extended attribute syscalls for "user" namespace attributes as defined by RFC-8276. Although this patch is fairly large, it should not affect support for the other versions of NFS. However it does add two new sysctls that allow a sysadmin to limit which minor versions of NFSv4 a server supports, allowing a sysadmin to disable NFSv4.2. Unfortunately, when the NFS stats structure was last revised, it was assumed that there would be no additional operations added beyond what was specified in RFC-7862. However RFC-8276 did add additional operations, forcing the NFS stats structure to revised again. It now has extra unused entries in all arrays, so that future extensions to NFSv4.2 can be accomodated without revising this structure again. A future commit will update nfsstat(1) to report counts for the new NFSv4.2 specific operations/procedures. This patch affects the internal interface between the nfscommon, nfscl and nfsd modules and, as such, they all must be upgraded simultaneously. I will do a version bump (although arguably not needed), due to this. This code has survived a "make universe" but has not been built with a recent GCC. If you encounter build problems, please email me. Relnotes: yes
2019-12-12 23:22:55 +00:00
14 * NFSX_UNSIGNED;
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
ds->nfsdev_flexaddrlen = addrlen;
tl = malloc(addrlen, M_NFSDSTATE, M_WAITOK | M_ZERO);
ds->nfsdev_flexaddr = (char *)tl;
*tl++ = txdr_unsigned(1); /* One multipath entry. */
/* The netaddr for this one entry. */
*tl++ = txdr_unsigned(strlen(netprot));
NFSBCOPY(netprot, tl, strlen(netprot));
tl += (NFSM_RNDUP(strlen(netprot)) / NFSX_UNSIGNED);
*tl++ = txdr_unsigned(strlen(addr));
NFSBCOPY(addr, tl, strlen(addr));
tl += (NFSM_RNDUP(strlen(addr)) / NFSX_UNSIGNED);
Add support for NFSv4.2 to the NFS client and server. This patch adds support for NFSv4.2 (RFC-7862) and Extended Attributes (RFC-8276) to the NFS client and server. NFSv4.2 is comprised of several optional features that can be supported in addition to NFSv4.1. This patch adds the following optional features: - posix_fadvise(POSIX_FADV_WILLNEED/POSIX_FADV_DONTNEED) - posix_fallocate() - intra server file range copying via the copy_file_range(2) syscall --> Avoiding data tranfer over the wire to/from the NFS client. - lseek(SEEK_DATA/SEEK_HOLE) - Extended attribute syscalls for "user" namespace attributes as defined by RFC-8276. Although this patch is fairly large, it should not affect support for the other versions of NFS. However it does add two new sysctls that allow a sysadmin to limit which minor versions of NFSv4 a server supports, allowing a sysadmin to disable NFSv4.2. Unfortunately, when the NFS stats structure was last revised, it was assumed that there would be no additional operations added beyond what was specified in RFC-7862. However RFC-8276 did add additional operations, forcing the NFS stats structure to revised again. It now has extra unused entries in all arrays, so that future extensions to NFSv4.2 can be accomodated without revising this structure again. A future commit will update nfsstat(1) to report counts for the new NFSv4.2 specific operations/procedures. This patch affects the internal interface between the nfscommon, nfscl and nfsd modules and, as such, they all must be upgraded simultaneously. I will do a version bump (although arguably not needed), due to this. This code has survived a "make universe" but has not been built with a recent GCC. If you encounter build problems, please email me. Relnotes: yes
2019-12-12 23:22:55 +00:00
*tl++ = txdr_unsigned(2); /* Two NFS Versions. */
*tl++ = txdr_unsigned(NFS_VER4); /* NFSv4. */
*tl++ = txdr_unsigned(NFSV42_MINORVERSION); /* Minor version 2. */
nfsd: Add sysctl to set maximum I/O size up to 1Mbyte Since MAXPHYS now allows the FreeBSD NFS client to do 1Mbyte I/O operations, add a sysctl called vfs.nfsd.srvmaxio so that the maximum NFS server I/O size can be set up to 1Mbyte. The Linux NFS client can also do 1Mbyte I/O operations. The default of 128Kbytes for the maximum I/O size has not been changed for two reasons: - kern.ipc.maxsockbuf must be increased to support 1Mbyte I/O - The limited benchmarking I can do actually shows a drop in I/O rate when the I/O size is above 256Kbytes. However, daveb@spectralogic.com reports seeing an increase in I/O rate for the 1Mbyte I/O size vs 128Kbytes using a Linux client. Reviewed by: asomers MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D30826
2021-07-16 22:01:03 +00:00
*tl++ = txdr_unsigned(nfs_srvmaxio); /* DS max rsize. */
*tl++ = txdr_unsigned(nfs_srvmaxio); /* DS max wsize. */
Add support for NFSv4.2 to the NFS client and server. This patch adds support for NFSv4.2 (RFC-7862) and Extended Attributes (RFC-8276) to the NFS client and server. NFSv4.2 is comprised of several optional features that can be supported in addition to NFSv4.1. This patch adds the following optional features: - posix_fadvise(POSIX_FADV_WILLNEED/POSIX_FADV_DONTNEED) - posix_fallocate() - intra server file range copying via the copy_file_range(2) syscall --> Avoiding data tranfer over the wire to/from the NFS client. - lseek(SEEK_DATA/SEEK_HOLE) - Extended attribute syscalls for "user" namespace attributes as defined by RFC-8276. Although this patch is fairly large, it should not affect support for the other versions of NFS. However it does add two new sysctls that allow a sysadmin to limit which minor versions of NFSv4 a server supports, allowing a sysadmin to disable NFSv4.2. Unfortunately, when the NFS stats structure was last revised, it was assumed that there would be no additional operations added beyond what was specified in RFC-7862. However RFC-8276 did add additional operations, forcing the NFS stats structure to revised again. It now has extra unused entries in all arrays, so that future extensions to NFSv4.2 can be accomodated without revising this structure again. A future commit will update nfsstat(1) to report counts for the new NFSv4.2 specific operations/procedures. This patch affects the internal interface between the nfscommon, nfscl and nfsd modules and, as such, they all must be upgraded simultaneously. I will do a version bump (although arguably not needed), due to this. This code has survived a "make universe" but has not been built with a recent GCC. If you encounter build problems, please email me. Relnotes: yes
2019-12-12 23:22:55 +00:00
*tl++ = newnfs_true; /* Tightly coupled. */
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
*tl++ = txdr_unsigned(NFS_VER4); /* NFSv4. */
*tl++ = txdr_unsigned(NFSV41_MINORVERSION); /* Minor version 1. */
nfsd: Add sysctl to set maximum I/O size up to 1Mbyte Since MAXPHYS now allows the FreeBSD NFS client to do 1Mbyte I/O operations, add a sysctl called vfs.nfsd.srvmaxio so that the maximum NFS server I/O size can be set up to 1Mbyte. The Linux NFS client can also do 1Mbyte I/O operations. The default of 128Kbytes for the maximum I/O size has not been changed for two reasons: - kern.ipc.maxsockbuf must be increased to support 1Mbyte I/O - The limited benchmarking I can do actually shows a drop in I/O rate when the I/O size is above 256Kbytes. However, daveb@spectralogic.com reports seeing an increase in I/O rate for the 1Mbyte I/O size vs 128Kbytes using a Linux client. Reviewed by: asomers MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D30826
2021-07-16 22:01:03 +00:00
*tl++ = txdr_unsigned(nfs_srvmaxio); /* DS max rsize. */
*tl++ = txdr_unsigned(nfs_srvmaxio); /* DS max wsize. */
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
*tl = newnfs_true; /* Tightly coupled. */
ds->nfsdev_hostnamelen = strlen(dnshost);
ds->nfsdev_host = malloc(ds->nfsdev_hostnamelen + 1, M_NFSDSTATE,
M_WAITOK);
NFSBCOPY(dnshost, ds->nfsdev_host, ds->nfsdev_hostnamelen + 1);
}
/*
* Create the device id list.
* Return 0 if the nfsd threads are to run and ENXIO if the "-p" argument
* is misconfigured.
*/
int
nfsrv_createdevids(struct nfsd_nfsd_args *args, NFSPROC_T *p)
{
struct nfsdevice *ds;
Add an optional feature to the pNFS server. Without this patch, the pNFS server distributes the data storage files across all of the specified DSs. A tester noted that it would be nice if a system administrator could control which DSs are used to store the file data for a given exported MDS file system. This patch adds the kernel support to do this. It also makes a slight semantic change to nfsv4_findmirror(), since some uses of it no longer require that the DS being searched for have a current mirror. A patch that will be committed in a few minutes will modify the nfsd daemon to support this feature. The patch should only affect sites using the pNFS server (specified via the "-p" command line option for nfsd. Suggested by: james.rose@framestore.com
2018-07-02 19:21:33 +00:00
char *addrp, *dnshostp, *dspathp, *mdspathp;
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
int error, i;
addrp = args->addr;
dnshostp = args->dnshost;
dspathp = args->dspath;
Add an optional feature to the pNFS server. Without this patch, the pNFS server distributes the data storage files across all of the specified DSs. A tester noted that it would be nice if a system administrator could control which DSs are used to store the file data for a given exported MDS file system. This patch adds the kernel support to do this. It also makes a slight semantic change to nfsv4_findmirror(), since some uses of it no longer require that the DS being searched for have a current mirror. A patch that will be committed in a few minutes will modify the nfsd daemon to support this feature. The patch should only affect sites using the pNFS server (specified via the "-p" command line option for nfsd. Suggested by: james.rose@framestore.com
2018-07-02 19:21:33 +00:00
mdspathp = args->mdspath;
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
nfsrv_maxpnfsmirror = args->mirrorcnt;
Add an optional feature to the pNFS server. Without this patch, the pNFS server distributes the data storage files across all of the specified DSs. A tester noted that it would be nice if a system administrator could control which DSs are used to store the file data for a given exported MDS file system. This patch adds the kernel support to do this. It also makes a slight semantic change to nfsv4_findmirror(), since some uses of it no longer require that the DS being searched for have a current mirror. A patch that will be committed in a few minutes will modify the nfsd daemon to support this feature. The patch should only affect sites using the pNFS server (specified via the "-p" command line option for nfsd. Suggested by: james.rose@framestore.com
2018-07-02 19:21:33 +00:00
if (addrp == NULL || dnshostp == NULL || dspathp == NULL ||
mdspathp == NULL)
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
return (0);
/*
* Loop around for each nul-terminated string in args->addr,
Add an optional feature to the pNFS server. Without this patch, the pNFS server distributes the data storage files across all of the specified DSs. A tester noted that it would be nice if a system administrator could control which DSs are used to store the file data for a given exported MDS file system. This patch adds the kernel support to do this. It also makes a slight semantic change to nfsv4_findmirror(), since some uses of it no longer require that the DS being searched for have a current mirror. A patch that will be committed in a few minutes will modify the nfsd daemon to support this feature. The patch should only affect sites using the pNFS server (specified via the "-p" command line option for nfsd. Suggested by: james.rose@framestore.com
2018-07-02 19:21:33 +00:00
* args->dnshost, args->dnspath and args->mdspath.
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
*/
while (addrp < (args->addr + args->addrlen) &&
dnshostp < (args->dnshost + args->dnshostlen) &&
Add an optional feature to the pNFS server. Without this patch, the pNFS server distributes the data storage files across all of the specified DSs. A tester noted that it would be nice if a system administrator could control which DSs are used to store the file data for a given exported MDS file system. This patch adds the kernel support to do this. It also makes a slight semantic change to nfsv4_findmirror(), since some uses of it no longer require that the DS being searched for have a current mirror. A patch that will be committed in a few minutes will modify the nfsd daemon to support this feature. The patch should only affect sites using the pNFS server (specified via the "-p" command line option for nfsd. Suggested by: james.rose@framestore.com
2018-07-02 19:21:33 +00:00
dspathp < (args->dspath + args->dspathlen) &&
mdspathp < (args->mdspath + args->mdspathlen)) {
error = nfsrv_setdsserver(dspathp, mdspathp, p, &ds);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
if (error != 0) {
/* Free all DS servers. */
nfsrv_freealldevids();
nfsrv_devidcnt = 0;
return (ENXIO);
}
nfsrv_allocdevid(ds, addrp, dnshostp);
addrp += (strlen(addrp) + 1);
dnshostp += (strlen(dnshostp) + 1);
dspathp += (strlen(dspathp) + 1);
Add an optional feature to the pNFS server. Without this patch, the pNFS server distributes the data storage files across all of the specified DSs. A tester noted that it would be nice if a system administrator could control which DSs are used to store the file data for a given exported MDS file system. This patch adds the kernel support to do this. It also makes a slight semantic change to nfsv4_findmirror(), since some uses of it no longer require that the DS being searched for have a current mirror. A patch that will be committed in a few minutes will modify the nfsd daemon to support this feature. The patch should only affect sites using the pNFS server (specified via the "-p" command line option for nfsd. Suggested by: james.rose@framestore.com
2018-07-02 19:21:33 +00:00
mdspathp += (strlen(mdspathp) + 1);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
}
if (nfsrv_devidcnt < nfsrv_maxpnfsmirror) {
/* Free all DS servers. */
nfsrv_freealldevids();
nfsrv_devidcnt = 0;
nfsrv_maxpnfsmirror = 1;
return (ENXIO);
}
Add a counter to limit the number of disabled DSs for a mirrored pNFS MDS. This patch adds a counter that limits the number of disabled mirrored DSs to mirror level - 1. It also makes a small change that keeps a Write that has failed with EACCES when attempted by a client to a DS from disabling the DS. This patch only affects the pNFS server.
2018-06-22 00:55:39 +00:00
/* We can fail at most one less DS than the mirror level. */
nfsrv_faildscnt = nfsrv_maxpnfsmirror - 1;
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
/*
* Allocate the nfslayout hash table now, since this is a pNFS server.
* Make it 1% of the high water mark and at least 100.
*/
if (nfslayouthash == NULL) {
nfsrv_layouthashsize = nfsrv_layouthighwater / 100;
if (nfsrv_layouthashsize < 100)
nfsrv_layouthashsize = 100;
nfslayouthash = mallocarray(nfsrv_layouthashsize,
sizeof(struct nfslayouthash), M_NFSDSESSION, M_WAITOK |
M_ZERO);
for (i = 0; i < nfsrv_layouthashsize; i++) {
mtx_init(&nfslayouthash[i].mtx, "nfslm", NULL, MTX_DEF);
TAILQ_INIT(&nfslayouthash[i].list);
}
}
return (0);
}
/*
* Free all device ids.
*/
static void
nfsrv_freealldevids(void)
{
struct nfsdevice *ds, *nds;
TAILQ_FOREACH_SAFE(ds, &nfsrv_devidhead, nfsdev_list, nds)
nfsrv_freedevid(ds);
}
/*
* Check to see if there is a Read/Write Layout plus either:
* - A Write Delegation
* or
* - An Open with Write_access.
* Return 1 if this is the case and 0 otherwise.
* This function is used by nfsrv_proxyds() to decide if doing a Proxy
* Getattr RPC to the Data Server (DS) is necessary.
*/
#define NFSCLIDVECSIZE 6
nfs: Remove APPLESTATIC macro It is no longer useful. Reviewed by: rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24811
2020-05-12 13:23:25 +00:00
int
Delete the unused "nd" argument for nfsrv_checkdsattr(). The "nd" argument for nfsrv_checkdsattr() is no longer used by the function. This patch deletes it. This allows subsequent patches to delete the "nd" argument from nfsrv_proxyds(), since it's only use of "nd" was to pass it to nfsrv_checkdsattr(). The same will then be true for nfsvno_getattr(), which passes "nd" to nfsrv_proxyds(). Getting rid of the "nd" argument from nfsvno_getattr() avoids confusion over why it might need "nd". This patch is trivial and does not have any semantic effect. Found by inspection while working on the NFSv4.2 server.
2019-09-04 22:37:28 +00:00
nfsrv_checkdsattr(vnode_t vp, NFSPROC_T *p)
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
{
fhandle_t fh, *tfhp;
struct nfsstate *stp;
struct nfslayout *lyp;
struct nfslayouthash *lhyp;
struct nfslockhashhead *hp;
struct nfslockfile *lfp;
nfsquad_t clid[NFSCLIDVECSIZE];
int clidcnt, ret;
ret = nfsvno_getfh(vp, &fh, p);
if (ret != 0)
return (0);
/* First check for a Read/Write Layout. */
clidcnt = 0;
lhyp = NFSLAYOUTHASH(&fh);
NFSLOCKLAYOUT(lhyp);
TAILQ_FOREACH(lyp, &lhyp->list, lay_list) {
if (NFSBCMP(&lyp->lay_fh, &fh, sizeof(fh)) == 0 &&
((lyp->lay_flags & NFSLAY_RW) != 0 ||
((lyp->lay_flags & NFSLAY_READ) != 0 &&
nfsrv_pnfsatime != 0))) {
if (clidcnt < NFSCLIDVECSIZE)
clid[clidcnt].qval = lyp->lay_clientid.qval;
clidcnt++;
}
}
NFSUNLOCKLAYOUT(lhyp);
if (clidcnt == 0) {
/* None found, so return 0. */
return (0);
}
/* Get the nfslockfile for this fh. */
NFSLOCKSTATE();
hp = NFSLOCKHASH(&fh);
LIST_FOREACH(lfp, hp, lf_hash) {
tfhp = &lfp->lf_fh;
if (NFSVNO_CMPFH(&fh, tfhp))
break;
}
if (lfp == NULL) {
/* None found, so return 0. */
NFSUNLOCKSTATE();
return (0);
}
/* Now, look for a Write delegation for this clientid. */
LIST_FOREACH(stp, &lfp->lf_deleg, ls_file) {
if ((stp->ls_flags & NFSLCK_DELEGWRITE) != 0 &&
nfsrv_fndclid(clid, stp->ls_clp->lc_clientid, clidcnt) != 0)
break;
}
if (stp != NULL) {
/* Found one, so return 1. */
NFSUNLOCKSTATE();
return (1);
}
/* No Write delegation, so look for an Open with Write_access. */
LIST_FOREACH(stp, &lfp->lf_open, ls_file) {
KASSERT((stp->ls_flags & NFSLCK_OPEN) != 0,
("nfsrv_checkdsattr: Non-open in Open list\n"));
if ((stp->ls_flags & NFSLCK_WRITEACCESS) != 0 &&
nfsrv_fndclid(clid, stp->ls_clp->lc_clientid, clidcnt) != 0)
break;
}
NFSUNLOCKSTATE();
if (stp != NULL)
return (1);
return (0);
}
/*
* Look for a matching clientid in the vector. Return 1 if one might match.
*/
static int
nfsrv_fndclid(nfsquad_t *clidvec, nfsquad_t clid, int clidcnt)
{
int i;
/* If too many for the vector, return 1 since there might be a match. */
if (clidcnt > NFSCLIDVECSIZE)
return (1);
for (i = 0; i < clidcnt; i++)
if (clidvec[i].qval == clid.qval)
return (1);
return (0);
}
/*
* Check the don't list for "vp" and see if issuing an rw layout is allowed.
* Return 1 if issuing an rw layout isn't allowed, 0 otherwise.
*/
static int
nfsrv_dontlayout(fhandle_t *fhp)
{
struct nfsdontlist *mrp;
int ret;
if (nfsrv_dontlistlen == 0)
return (0);
ret = 0;
NFSDDONTLISTLOCK();
LIST_FOREACH(mrp, &nfsrv_dontlisthead, nfsmr_list) {
if (NFSBCMP(fhp, &mrp->nfsmr_fh, sizeof(*fhp)) == 0 &&
(mrp->nfsmr_flags & NFSMR_DONTLAYOUT) != 0) {
ret = 1;
break;
}
}
NFSDDONTLISTUNLOCK();
return (ret);
}
#define PNFSDS_COPYSIZ 65536
/*
* Create a new file on a DS and copy the contents of an extant DS file to it.
* This can be used for recovery of a DS file onto a recovered DS.
* The steps are:
* - When called, the MDS file's vnode is locked, blocking LayoutGet operations.
* - Disable issuing of read/write layouts for the file via the nfsdontlist,
* so that they will be disabled after the MDS file's vnode is unlocked.
* - Set up the nfsrv_recalllist so that recall of read/write layouts can
* be done.
* - Unlock the MDS file's vnode, so that the client(s) can perform proxied
* writes, LayoutCommits and LayoutReturns for the file when completing the
* LayoutReturn requested by the LayoutRecall callback.
* - Issue a LayoutRecall callback for all read/write layouts and wait for
* them to be returned. (If the LayoutRecall callback replies
* NFSERR_NOMATCHLAYOUT, they are gone and no LayoutReturn is needed.)
* - Exclusively lock the MDS file's vnode. This ensures that no proxied
* writes are in progress or can occur during the DS file copy.
* It also blocks Setattr operations.
* - Create the file on the recovered mirror.
* - Copy the file from the operational DS.
* - Copy any ACL from the MDS file to the new DS file.
* - Set the modify time of the new DS file to that of the MDS file.
* - Update the extended attribute for the MDS file.
* - Enable issuing of rw layouts by deleting the nfsdontlist entry.
* - The caller will unlock the MDS file's vnode allowing operations
* to continue normally, since it is now on the mirror again.
*/
int
nfsrv_copymr(vnode_t vp, vnode_t fvp, vnode_t dvp, struct nfsdevice *ds,
struct pnfsdsfile *pf, struct pnfsdsfile *wpf, int mirrorcnt,
struct ucred *cred, NFSPROC_T *p)
{
struct nfsdontlist *mrp, *nmrp;
struct nfslayouthash *lhyp;
struct nfslayout *lyp, *nlyp;
struct nfslayouthead thl;
Fix LORs between vn_start_write() and vn_lock() in nfsrv_copymr(). When coding the pNFS server, I added vn_start_write() calls in nfsrv_copymr() done while the vnodes were locked, not realizing I had introduced LORs and possible deadlock when an exported file system on the MDS is suspended. This patch fixes the LORs by moving the vn_start_write() calls up to before where the vnodes are locked. For "tvp", the vn_start_write() probaby isn't necessary, because NFS mounts can't be suspended. However, I think doing so is harmless. Thanks go to kib@ for letting me know that I had introduced these LORs. This patch only affects the behaviour of the pNFS server when pnfsdscopymr(8) is used to recover a mirrored DS.
2018-08-18 19:14:06 +00:00
struct mount *mp, *tvmp;
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
struct acl *aclp;
struct vattr va;
struct timespec mtime;
fhandle_t fh;
vnode_t tvp;
off_t rdpos, wrpos;
ssize_t aresid;
char *dat;
int didprintf, ret, retacl, xfer;
ASSERT_VOP_LOCKED(fvp, "nfsrv_copymr fvp");
ASSERT_VOP_LOCKED(vp, "nfsrv_copymr vp");
/*
* Allocate a nfsdontlist entry and set the NFSMR_DONTLAYOUT flag
* so that no more RW layouts will get issued.
*/
ret = nfsvno_getfh(vp, &fh, p);
if (ret != 0) {
NFSD_DEBUG(4, "nfsrv_copymr: getfh=%d\n", ret);
return (ret);
}
nmrp = malloc(sizeof(*nmrp), M_NFSDSTATE, M_WAITOK);
nmrp->nfsmr_flags = NFSMR_DONTLAYOUT;
NFSBCOPY(&fh, &nmrp->nfsmr_fh, sizeof(fh));
NFSDDONTLISTLOCK();
LIST_FOREACH(mrp, &nfsrv_dontlisthead, nfsmr_list) {
if (NFSBCMP(&fh, &mrp->nfsmr_fh, sizeof(fh)) == 0)
break;
}
if (mrp == NULL) {
LIST_INSERT_HEAD(&nfsrv_dontlisthead, nmrp, nfsmr_list);
mrp = nmrp;
nmrp = NULL;
nfsrv_dontlistlen++;
NFSD_DEBUG(4, "nfsrv_copymr: in dontlist\n");
} else {
NFSDDONTLISTUNLOCK();
free(nmrp, M_NFSDSTATE);
NFSD_DEBUG(4, "nfsrv_copymr: dup dontlist\n");
return (ENXIO);
}
NFSDDONTLISTUNLOCK();
/*
* Search for all RW layouts for this file. Move them to the
* recall list, so they can be recalled and their return noted.
*/
lhyp = NFSLAYOUTHASH(&fh);
NFSDRECALLLOCK();
NFSLOCKLAYOUT(lhyp);
TAILQ_FOREACH_SAFE(lyp, &lhyp->list, lay_list, nlyp) {
if (NFSBCMP(&lyp->lay_fh, &fh, sizeof(fh)) == 0 &&
(lyp->lay_flags & NFSLAY_RW) != 0) {
TAILQ_REMOVE(&lhyp->list, lyp, lay_list);
TAILQ_INSERT_HEAD(&nfsrv_recalllisthead, lyp, lay_list);
lyp->lay_trycnt = 0;
}
}
NFSUNLOCKLAYOUT(lhyp);
NFSDRECALLUNLOCK();
ret = 0;
Fix LORs between vn_start_write() and vn_lock() in nfsrv_copymr(). When coding the pNFS server, I added vn_start_write() calls in nfsrv_copymr() done while the vnodes were locked, not realizing I had introduced LORs and possible deadlock when an exported file system on the MDS is suspended. This patch fixes the LORs by moving the vn_start_write() calls up to before where the vnodes are locked. For "tvp", the vn_start_write() probaby isn't necessary, because NFS mounts can't be suspended. However, I think doing so is harmless. Thanks go to kib@ for letting me know that I had introduced these LORs. This patch only affects the behaviour of the pNFS server when pnfsdscopymr(8) is used to recover a mirrored DS.
2018-08-18 19:14:06 +00:00
mp = tvmp = NULL;
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
didprintf = 0;
TAILQ_INIT(&thl);
/* Unlock the MDS vp, so that a LayoutReturn can be done on it. */
vfs: drop the mostly unused flags argument from VOP_UNLOCK Filesystems which want to use it in limited capacity can employ the VOP_UNLOCK_FLAGS macro. Reviewed by: kib (previous version) Differential Revision: https://reviews.freebsd.org/D21427
2020-01-03 22:29:58 +00:00
NFSVOPUNLOCK(vp);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
/* Now, do a recall for all layouts not yet recalled. */
tryagain:
NFSDRECALLLOCK();
TAILQ_FOREACH(lyp, &nfsrv_recalllisthead, lay_list) {
if (NFSBCMP(&lyp->lay_fh, &fh, sizeof(fh)) == 0 &&
(lyp->lay_flags & NFSLAY_RECALL) == 0) {
lyp->lay_flags |= NFSLAY_RECALL;
/*
* The layout stateid.seqid needs to be incremented
* before doing a LAYOUT_RECALL callback.
*/
if (++lyp->lay_stateid.seqid == 0)
lyp->lay_stateid.seqid = 1;
NFSDRECALLUNLOCK();
nfsrv_recalllayout(lyp->lay_clientid, &lyp->lay_stateid,
Assorted fixes to handling of LayoutRecall callbacks, mostly error handling. After a re-read of the appropriate section of RFC5661, I decided that a few things should be changed related to LayoutRecall callback handling. Here are the things fixed by this patch. - For two of the three cases that LayoutRecall is done, I now think setting the clora_changed argument false is correct. - All errors other than NFSERR_DELAY returned by LayoutRecall appear permanent, so don't retry for any of them. (NFSERR_DELAY is retried by newnfs_request(), so it is not affected by this patch.) - Instead of waiting "forever" (actually until the process is SIGTERM'd) for Layouts to be returned during a mirror copy, fail and return ENXIO after about 1minute. Waiting for a <ctrl>C made sense when pnfsdscopymr() was done by itself, but did not make sense when done via find(1). This patch only affects the pNFS server.
2018-08-08 20:21:45 +00:00
&lyp->lay_fh, lyp, 0, lyp->lay_type, p);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
NFSD_DEBUG(4, "nfsrv_copymr: recalled layout\n");
goto tryagain;
}
}
/* Now wait for them to be returned. */
tryagain2:
TAILQ_FOREACH(lyp, &nfsrv_recalllisthead, lay_list) {
if (NFSBCMP(&lyp->lay_fh, &fh, sizeof(fh)) == 0) {
if ((lyp->lay_flags & NFSLAY_RETURNED) != 0) {
TAILQ_REMOVE(&nfsrv_recalllisthead, lyp,
lay_list);
TAILQ_INSERT_HEAD(&thl, lyp, lay_list);
NFSD_DEBUG(4,
"nfsrv_copymr: layout returned\n");
} else {
Assorted fixes to handling of LayoutRecall callbacks, mostly error handling. After a re-read of the appropriate section of RFC5661, I decided that a few things should be changed related to LayoutRecall callback handling. Here are the things fixed by this patch. - For two of the three cases that LayoutRecall is done, I now think setting the clora_changed argument false is correct. - All errors other than NFSERR_DELAY returned by LayoutRecall appear permanent, so don't retry for any of them. (NFSERR_DELAY is retried by newnfs_request(), so it is not affected by this patch.) - Instead of waiting "forever" (actually until the process is SIGTERM'd) for Layouts to be returned during a mirror copy, fail and return ENXIO after about 1minute. Waiting for a <ctrl>C made sense when pnfsdscopymr() was done by itself, but did not make sense when done via find(1). This patch only affects the pNFS server.
2018-08-08 20:21:45 +00:00
lyp->lay_trycnt++;
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
ret = mtx_sleep(lyp, NFSDRECALLMUTEXPTR,
PVFS | PCATCH, "nfsmrl", hz);
NFSD_DEBUG(4, "nfsrv_copymr: aft sleep=%d\n",
ret);
if (ret == EINTR || ret == ERESTART)
break;
Assorted fixes to handling of LayoutRecall callbacks, mostly error handling. After a re-read of the appropriate section of RFC5661, I decided that a few things should be changed related to LayoutRecall callback handling. Here are the things fixed by this patch. - For two of the three cases that LayoutRecall is done, I now think setting the clora_changed argument false is correct. - All errors other than NFSERR_DELAY returned by LayoutRecall appear permanent, so don't retry for any of them. (NFSERR_DELAY is retried by newnfs_request(), so it is not affected by this patch.) - Instead of waiting "forever" (actually until the process is SIGTERM'd) for Layouts to be returned during a mirror copy, fail and return ENXIO after about 1minute. Waiting for a <ctrl>C made sense when pnfsdscopymr() was done by itself, but did not make sense when done via find(1). This patch only affects the pNFS server.
2018-08-08 20:21:45 +00:00
if ((lyp->lay_flags & NFSLAY_RETURNED) == 0) {
/*
* Give up after 60sec and return
* ENXIO, failing the copymr.
* This layout will remain on the
* recalllist. It can only be cleared
* by restarting the nfsd.
* This seems the safe way to handle
* it, since it cannot be safely copied
* with an outstanding RW layout.
*/
if (lyp->lay_trycnt >= 60) {
ret = ENXIO;
break;
}
if (didprintf == 0) {
printf("nfsrv_copymr: layout "
"not returned\n");
didprintf = 1;
}
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
}
}
goto tryagain2;
}
}
NFSDRECALLUNLOCK();
/* We can now get rid of the layouts that have been returned. */
TAILQ_FOREACH_SAFE(lyp, &thl, lay_list, nlyp)
nfsrv_freelayout(&thl, lyp);
Fix LORs between vn_start_write() and vn_lock() in nfsrv_copymr(). When coding the pNFS server, I added vn_start_write() calls in nfsrv_copymr() done while the vnodes were locked, not realizing I had introduced LORs and possible deadlock when an exported file system on the MDS is suspended. This patch fixes the LORs by moving the vn_start_write() calls up to before where the vnodes are locked. For "tvp", the vn_start_write() probaby isn't necessary, because NFS mounts can't be suspended. However, I think doing so is harmless. Thanks go to kib@ for letting me know that I had introduced these LORs. This patch only affects the behaviour of the pNFS server when pnfsdscopymr(8) is used to recover a mirrored DS.
2018-08-18 19:14:06 +00:00
/*
* Do the vn_start_write() calls here, before the MDS vnode is
* locked and the tvp is created (locked) in the NFS file system
* that dvp is in.
* For tvmp, this probably isn't necessary, since it will be an
* NFS mount and they are not suspendable at this time.
*/
if (ret == 0)
vfs: introduce V_PCATCH to stop abusing PCATCH
2022-09-17 15:22:04 +00:00
ret = vn_start_write(vp, &mp, V_WAIT | V_PCATCH);
Fix LORs between vn_start_write() and vn_lock() in nfsrv_copymr(). When coding the pNFS server, I added vn_start_write() calls in nfsrv_copymr() done while the vnodes were locked, not realizing I had introduced LORs and possible deadlock when an exported file system on the MDS is suspended. This patch fixes the LORs by moving the vn_start_write() calls up to before where the vnodes are locked. For "tvp", the vn_start_write() probaby isn't necessary, because NFS mounts can't be suspended. However, I think doing so is harmless. Thanks go to kib@ for letting me know that I had introduced these LORs. This patch only affects the behaviour of the pNFS server when pnfsdscopymr(8) is used to recover a mirrored DS.
2018-08-18 19:14:06 +00:00
if (ret == 0) {
tvmp = dvp->v_mount;
vfs: introduce V_PCATCH to stop abusing PCATCH
2022-09-17 15:22:04 +00:00
ret = vn_start_write(NULL, &tvmp, V_WAIT | V_PCATCH);
Fix LORs between vn_start_write() and vn_lock() in nfsrv_copymr(). When coding the pNFS server, I added vn_start_write() calls in nfsrv_copymr() done while the vnodes were locked, not realizing I had introduced LORs and possible deadlock when an exported file system on the MDS is suspended. This patch fixes the LORs by moving the vn_start_write() calls up to before where the vnodes are locked. For "tvp", the vn_start_write() probaby isn't necessary, because NFS mounts can't be suspended. However, I think doing so is harmless. Thanks go to kib@ for letting me know that I had introduced these LORs. This patch only affects the behaviour of the pNFS server when pnfsdscopymr(8) is used to recover a mirrored DS.
2018-08-18 19:14:06 +00:00
}
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
/*
* LK_EXCLUSIVE lock the MDS vnode, so that any
* proxied writes through the MDS will be blocked until we have
* completed the copy and update of the extended attributes.
* This will also ensure that any attributes and ACL will not be
* changed until the copy is complete.
*/
NFSVOPLOCK(vp, LK_EXCLUSIVE | LK_RETRY);
vfs: introduce v_irflag and make v_type smaller The current vnode layout is not smp-friendly by having frequently read data avoidably sharing cachelines with very frequently modified fields. In particular v_iflag inspected for VI_DOOMED can be found in the same line with v_usecount. Instead make it available in the same cacheline as the v_op, v_data and v_type which all get read all the time. v_type is avoidably 4 bytes while the necessary data will easily fit in 1. Shrinking it frees up 3 bytes, 2 of which get used here to introduce a new flag field with a new value: VIRF_DOOMED. Reviewed by: kib, jeff Differential Revision: https://reviews.freebsd.org/D22715
2019-12-08 21:30:04 +00:00
if (ret == 0 && VN_IS_DOOMED(vp)) {
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
NFSD_DEBUG(4, "nfsrv_copymr: lk_exclusive doomed\n");
ret = ESTALE;
}
/* Create the data file on the recovered DS. */
if (ret == 0)
ret = nfsrv_createdsfile(vp, &fh, pf, dvp, ds, cred, p, &tvp);
/* Copy the DS file, if created successfully. */
if (ret == 0) {
/*
* Get any NFSv4 ACL on the MDS file, so that it can be set
* on the new DS file.
*/
aclp = acl_alloc(M_WAITOK | M_ZERO);
retacl = VOP_GETACL(vp, ACL_TYPE_NFS4, aclp, cred, p);
if (retacl != 0 && retacl != ENOATTR)
NFSD_DEBUG(1, "nfsrv_copymr: vop_getacl=%d\n", retacl);
dat = malloc(PNFSDS_COPYSIZ, M_TEMP, M_WAITOK);
Fix the kernel part of pnfsdscopymr() to handle holes in the file being copied. If a mirrored DS is being recovered that has a lot of large sparse files, pnfsdscopymr(8) would use a lot of space on the recovered mirror since it would write the "holes" in the file being mirrored. This patch adds code to check for a "hole" and skip doing the write. The check is done on a "per PNFSDS_COPYSIZ size block", which is currently 64K. I think that most file server file systems will be using a blocksize at least this large. If the file server is using a smaller blocksize and smaller holes need to be preserved, PNFSDS_COPYSIZ could be decreased. The block of 0s is malloc()d, since pnfsdcopymr(8) should be an infrequent occurrence.
2018-07-08 18:15:55 +00:00
/* Malloc a block of 0s used to check for holes. */
if (nfsrv_zeropnfsdat == NULL)
nfsrv_zeropnfsdat = malloc(PNFSDS_COPYSIZ, M_TEMP,
M_WAITOK | M_ZERO);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
rdpos = wrpos = 0;
Fix LORs between vn_start_write() and vn_lock() in nfsrv_copymr(). When coding the pNFS server, I added vn_start_write() calls in nfsrv_copymr() done while the vnodes were locked, not realizing I had introduced LORs and possible deadlock when an exported file system on the MDS is suspended. This patch fixes the LORs by moving the vn_start_write() calls up to before where the vnodes are locked. For "tvp", the vn_start_write() probaby isn't necessary, because NFS mounts can't be suspended. However, I think doing so is harmless. Thanks go to kib@ for letting me know that I had introduced these LORs. This patch only affects the behaviour of the pNFS server when pnfsdscopymr(8) is used to recover a mirrored DS.
2018-08-18 19:14:06 +00:00
ret = VOP_GETATTR(fvp, &va, cred);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
aresid = 0;
while (ret == 0 && aresid == 0) {
ret = vn_rdwr(UIO_READ, fvp, dat, PNFSDS_COPYSIZ,
rdpos, UIO_SYSSPACE, IO_NODELOCKED, cred, NULL,
&aresid, p);
xfer = PNFSDS_COPYSIZ - aresid;
if (ret == 0 && xfer > 0) {
rdpos += xfer;
Fix the kernel part of pnfsdscopymr() to handle holes in the file being copied. If a mirrored DS is being recovered that has a lot of large sparse files, pnfsdscopymr(8) would use a lot of space on the recovered mirror since it would write the "holes" in the file being mirrored. This patch adds code to check for a "hole" and skip doing the write. The check is done on a "per PNFSDS_COPYSIZ size block", which is currently 64K. I think that most file server file systems will be using a blocksize at least this large. If the file server is using a smaller blocksize and smaller holes need to be preserved, PNFSDS_COPYSIZ could be decreased. The block of 0s is malloc()d, since pnfsdcopymr(8) should be an infrequent occurrence.
2018-07-08 18:15:55 +00:00
/*
* Skip the write for holes, except for the
* last block.
*/
if (xfer < PNFSDS_COPYSIZ || rdpos ==
va.va_size || NFSBCMP(dat,
nfsrv_zeropnfsdat, PNFSDS_COPYSIZ) != 0)
ret = vn_rdwr(UIO_WRITE, tvp, dat, xfer,
wrpos, UIO_SYSSPACE, IO_NODELOCKED,
cred, NULL, NULL, p);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
if (ret == 0)
wrpos += xfer;
}
}
/* If there is an ACL and the copy succeeded, set the ACL. */
if (ret == 0 && retacl == 0) {
ret = VOP_SETACL(tvp, ACL_TYPE_NFS4, aclp, cred, p);
/*
* Don't consider these as errors, since VOP_GETACL()
* can return an ACL when they are not actually
* supported. For example, for UFS, VOP_GETACL()
* will return a trivial ACL based on the uid/gid/mode
* when there is no ACL on the file.
* This case should be recognized as a trivial ACL
* by UFS's VOP_SETACL() and succeed, but...
*/
if (ret == ENOATTR || ret == EOPNOTSUPP || ret == EPERM)
ret = 0;
}
if (ret == 0)
ret = VOP_FSYNC(tvp, MNT_WAIT, p);
/* Set the DS data file's modify time that of the MDS file. */
if (ret == 0)
ret = VOP_GETATTR(vp, &va, cred);
if (ret == 0) {
mtime = va.va_mtime;
VATTR_NULL(&va);
va.va_mtime = mtime;
ret = VOP_SETATTR(tvp, &va, cred);
}
vput(tvp);
acl_free(aclp);
free(dat, M_TEMP);
}
Fix LORs between vn_start_write() and vn_lock() in nfsrv_copymr(). When coding the pNFS server, I added vn_start_write() calls in nfsrv_copymr() done while the vnodes were locked, not realizing I had introduced LORs and possible deadlock when an exported file system on the MDS is suspended. This patch fixes the LORs by moving the vn_start_write() calls up to before where the vnodes are locked. For "tvp", the vn_start_write() probaby isn't necessary, because NFS mounts can't be suspended. However, I think doing so is harmless. Thanks go to kib@ for letting me know that I had introduced these LORs. This patch only affects the behaviour of the pNFS server when pnfsdscopymr(8) is used to recover a mirrored DS.
2018-08-18 19:14:06 +00:00
if (tvmp != NULL)
vn_finished_write(tvmp);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
/* Update the extended attributes for the newly created DS file. */
Fix LORs between vn_start_write() and vn_lock() in nfsrv_copymr(). When coding the pNFS server, I added vn_start_write() calls in nfsrv_copymr() done while the vnodes were locked, not realizing I had introduced LORs and possible deadlock when an exported file system on the MDS is suspended. This patch fixes the LORs by moving the vn_start_write() calls up to before where the vnodes are locked. For "tvp", the vn_start_write() probaby isn't necessary, because NFS mounts can't be suspended. However, I think doing so is harmless. Thanks go to kib@ for letting me know that I had introduced these LORs. This patch only affects the behaviour of the pNFS server when pnfsdscopymr(8) is used to recover a mirrored DS.
2018-08-18 19:14:06 +00:00
if (ret == 0)
ret = vn_extattr_set(vp, IO_NODELOCKED,
EXTATTR_NAMESPACE_SYSTEM, "pnfsd.dsfile",
sizeof(*wpf) * mirrorcnt, (char *)wpf, p);
if (mp != NULL)
vn_finished_write(mp);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
/* Get rid of the dontlist entry, so that Layouts can be issued. */
NFSDDONTLISTLOCK();
LIST_REMOVE(mrp, nfsmr_list);
NFSDDONTLISTUNLOCK();
free(mrp, M_NFSDSTATE);
return (ret);
}
/*
* Create a data storage file on the recovered DS.
*/
static int
nfsrv_createdsfile(vnode_t vp, fhandle_t *fhp, struct pnfsdsfile *pf,
vnode_t dvp, struct nfsdevice *ds, struct ucred *cred, NFSPROC_T *p,
vnode_t *tvpp)
{
struct vattr va, nva;
int error;
/* Make data file name based on FH. */
error = VOP_GETATTR(vp, &va, cred);
if (error == 0) {
/* Set the attributes for "vp" to Setattr the DS vp. */
VATTR_NULL(&nva);
nva.va_uid = va.va_uid;
nva.va_gid = va.va_gid;
nva.va_mode = va.va_mode;
nva.va_size = 0;
VATTR_NULL(&va);
va.va_type = VREG;
va.va_mode = nva.va_mode;
NFSD_DEBUG(4, "nfsrv_dscreatefile: dvp=%p pf=%p\n", dvp, pf);
error = nfsrv_dscreate(dvp, &va, &nva, fhp, pf, NULL,
pf->dsf_filename, cred, p, tvpp);
}
return (error);
}
/*
* Look up the MDS file shared locked, and then get the extended attribute
* to find the extant DS file to be copied to the new mirror.
* If successful, *vpp is set to the MDS file's vp and *nvpp is
* set to a DS data file for the MDS file, both exclusively locked.
* The "buf" argument has the pnfsdsfile structure from the MDS file
* in it and buflen is set to its length.
*/
int
nfsrv_mdscopymr(char *mdspathp, char *dspathp, char *curdspathp, char *buf,
int *buflenp, char *fname, NFSPROC_T *p, struct vnode **vpp,
struct vnode **nvpp, struct pnfsdsfile **pfp, struct nfsdevice **dsp,
struct nfsdevice **fdsp)
{
struct nameidata nd;
struct vnode *vp, *curvp;
struct pnfsdsfile *pf;
struct nfsmount *nmp, *curnmp;
int dsdir, error, mirrorcnt, ippos;
vp = NULL;
curvp = NULL;
curnmp = NULL;
*dsp = NULL;
*fdsp = NULL;
if (dspathp == NULL && curdspathp != NULL)
return (EPERM);
/*
* Look up the MDS file shared locked. The lock will be upgraded
* to an exclusive lock after any rw layouts have been returned.
*/
NFSD_DEBUG(4, "mdsopen path=%s\n", mdspathp);
NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF, UIO_SYSSPACE,
vfs: remove the unused thread argument from NDINIT* See b4a58fbf640409a1 ("vfs: remove cn_thread") Bump __FreeBSD_version to 1400043.
2021-11-25 21:43:06 +00:00
mdspathp);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
error = namei(&nd);
NFSD_DEBUG(4, "lookup=%d\n", error);
if (error != 0)
return (error);
if (nd.ni_vp->v_type != VREG) {
vput(nd.ni_vp);
NFSD_DEBUG(4, "mdspath not reg\n");
return (EISDIR);
}
vp = nd.ni_vp;
if (curdspathp != NULL) {
/*
* Look up the current DS path and find the nfsdev structure for
* it.
*/
NFSD_DEBUG(4, "curmdsdev path=%s\n", curdspathp);
NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF,
vfs: remove the unused thread argument from NDINIT* See b4a58fbf640409a1 ("vfs: remove cn_thread") Bump __FreeBSD_version to 1400043.
2021-11-25 21:43:06 +00:00
UIO_SYSSPACE, curdspathp);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
error = namei(&nd);
NFSD_DEBUG(4, "ds lookup=%d\n", error);
if (error != 0) {
vput(vp);
return (error);
}
if (nd.ni_vp->v_type != VDIR) {
vput(nd.ni_vp);
vput(vp);
NFSD_DEBUG(4, "curdspath not dir\n");
return (ENOTDIR);
}
if (strcmp(nd.ni_vp->v_mount->mnt_vfc->vfc_name, "nfs") != 0) {
vput(nd.ni_vp);
vput(vp);
NFSD_DEBUG(4, "curdspath not an NFS mount\n");
return (ENXIO);
}
curnmp = VFSTONFS(nd.ni_vp->v_mount);
fs: clean up empty lines in .c and .h files
2020-09-01 21:18:40 +00:00
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
/* Search the nfsdev list for a match. */
NFSDDSLOCK();
*fdsp = nfsv4_findmirror(curnmp);
NFSDDSUNLOCK();
if (*fdsp == NULL)
curnmp = NULL;
if (curnmp == NULL) {
vput(nd.ni_vp);
vput(vp);
NFSD_DEBUG(4, "mdscopymr: no current ds\n");
return (ENXIO);
}
curvp = nd.ni_vp;
}
if (dspathp != NULL) {
/* Look up the nfsdev path and find the nfsdev structure. */
NFSD_DEBUG(4, "mdsdev path=%s\n", dspathp);
NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF,
vfs: remove the unused thread argument from NDINIT* See b4a58fbf640409a1 ("vfs: remove cn_thread") Bump __FreeBSD_version to 1400043.
2021-11-25 21:43:06 +00:00
UIO_SYSSPACE, dspathp);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
error = namei(&nd);
NFSD_DEBUG(4, "ds lookup=%d\n", error);
if (error != 0) {
vput(vp);
if (curvp != NULL)
vput(curvp);
return (error);
}
if (nd.ni_vp->v_type != VDIR || nd.ni_vp == curvp) {
vput(nd.ni_vp);
vput(vp);
if (curvp != NULL)
vput(curvp);
NFSD_DEBUG(4, "dspath not dir\n");
if (nd.ni_vp == curvp)
return (EPERM);
return (ENOTDIR);
}
if (strcmp(nd.ni_vp->v_mount->mnt_vfc->vfc_name, "nfs") != 0) {
vput(nd.ni_vp);
vput(vp);
if (curvp != NULL)
vput(curvp);
NFSD_DEBUG(4, "dspath not an NFS mount\n");
return (ENXIO);
}
nmp = VFSTONFS(nd.ni_vp->v_mount);
fs: clean up empty lines in .c and .h files
2020-09-01 21:18:40 +00:00
Add an optional feature to the pNFS server. Without this patch, the pNFS server distributes the data storage files across all of the specified DSs. A tester noted that it would be nice if a system administrator could control which DSs are used to store the file data for a given exported MDS file system. This patch adds the kernel support to do this. It also makes a slight semantic change to nfsv4_findmirror(), since some uses of it no longer require that the DS being searched for have a current mirror. A patch that will be committed in a few minutes will modify the nfsd daemon to support this feature. The patch should only affect sites using the pNFS server (specified via the "-p" command line option for nfsd. Suggested by: james.rose@framestore.com
2018-07-02 19:21:33 +00:00
/*
* Search the nfsdevice list for a match. If curnmp == NULL,
* this is a recovery and there must be a mirror.
*/
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
NFSDDSLOCK();
Add an optional feature to the pNFS server. Without this patch, the pNFS server distributes the data storage files across all of the specified DSs. A tester noted that it would be nice if a system administrator could control which DSs are used to store the file data for a given exported MDS file system. This patch adds the kernel support to do this. It also makes a slight semantic change to nfsv4_findmirror(), since some uses of it no longer require that the DS being searched for have a current mirror. A patch that will be committed in a few minutes will modify the nfsd daemon to support this feature. The patch should only affect sites using the pNFS server (specified via the "-p" command line option for nfsd. Suggested by: james.rose@framestore.com
2018-07-02 19:21:33 +00:00
if (curnmp == NULL)
*dsp = nfsrv_findmirroredds(nmp);
else
*dsp = nfsv4_findmirror(nmp);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
NFSDDSUNLOCK();
if (*dsp == NULL) {
vput(nd.ni_vp);
vput(vp);
if (curvp != NULL)
vput(curvp);
NFSD_DEBUG(4, "mdscopymr: no ds\n");
return (ENXIO);
}
} else {
nd.ni_vp = NULL;
nmp = NULL;
}
/*
* Get a vp for an available DS data file using the extended
* attribute on the MDS file.
* If there is a valid entry for the new DS in the extended attribute
* on the MDS file (as checked via the nmp argument),
* nfsrv_dsgetsockmnt() returns EEXIST, so no copying will occur.
*/
error = nfsrv_dsgetsockmnt(vp, 0, buf, buflenp, &mirrorcnt, p,
NULL, NULL, NULL, fname, nvpp, &nmp, curnmp, &ippos, &dsdir);
if (curvp != NULL)
vput(curvp);
if (nd.ni_vp == NULL) {
if (error == 0 && nmp != NULL) {
/* Search the nfsdev list for a match. */
NFSDDSLOCK();
Add an optional feature to the pNFS server. Without this patch, the pNFS server distributes the data storage files across all of the specified DSs. A tester noted that it would be nice if a system administrator could control which DSs are used to store the file data for a given exported MDS file system. This patch adds the kernel support to do this. It also makes a slight semantic change to nfsv4_findmirror(), since some uses of it no longer require that the DS being searched for have a current mirror. A patch that will be committed in a few minutes will modify the nfsd daemon to support this feature. The patch should only affect sites using the pNFS server (specified via the "-p" command line option for nfsd. Suggested by: james.rose@framestore.com
2018-07-02 19:21:33 +00:00
*dsp = nfsrv_findmirroredds(nmp);
Merge the pNFS server code from projects/pnfs-planb-server into head. This code merge adds a pNFS service to the NFSv4.1 server. Although it is a large commit it should not affect behaviour for a non-pNFS NFS server. Some documentation on how this works can be found at: http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt and will hopefully be turned into a proper document soon. This is a merge of the kernel code. Userland and man page changes will come soon, once the dust settles on this merge. It has passed a "make universe", so I hope it will not cause build problems. It also adds NFSv4.1 server support for the "current stateid". Here is a brief overview of the pNFS service: A pNFS service separates the Read/Write oeprations from all the other NFSv4.1 Metadata operations. It is hoped that this separation allows a pNFS service to be configured that exceeds the limits of a single NFS server for either storage capacity and/or I/O bandwidth. It is possible to configure mirroring within the data servers (DSs) so that the data storage file for an MDS file will be mirrored on two or more of the DSs. When this is used, failure of a DS will not stop the pNFS service and a failed DS can be recovered once repaired while the pNFS service continues to operate. Although two way mirroring would be the norm, it is possible to set a mirroring level of up to four or the number of DSs, whichever is less. The Metadata server will always be a single point of failure, just as a single NFS server is. A Plan B pNFS service consists of a single MetaData Server (MDS) and K Data Servers (DS), all of which are recent FreeBSD systems. Clients will mount the MDS as they would a single NFS server. When files are created, the MDS creates a file tree identical to what a single NFS server creates, except that all the regular (VREG) files will be empty. As such, if you look at the exported tree on the MDS directly on the MDS server (not via an NFS mount), the files will all be of size 0. Each of these files will also have two extended attributes in the system attribute name space: pnfsd.dsfile - This extended attrbute stores the information that the MDS needs to find the data storage file(s) on DS(s) for this file. pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime and Change attributes for the file, so that the MDS doesn't need to acquire the attributes from the DS for every Getattr operation. For each regular (VREG) file, the MDS creates a data storage file on one (or more if mirroring is enabled) of the DSs in one of the "dsNN" subdirectories. The name of this file is the file handle of the file on the MDS in hexadecimal so that the name is unique. The DSs use subdirectories named "ds0" to "dsN" so that no one directory gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize on the MDS, with the default being 20. For production servers that will store a lot of files, this value should probably be much larger. It can be increased when the "nfsd" daemon is not running on the MDS, once the "dsK" directories are created. For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces of information to the client that allows it to do I/O directly to the DS. DeviceInfo - This is relatively static information that defines what a DS is. The critical bits of information returned by the FreeBSD server is the IP address of the DS and, for the Flexible File layout, that NFSv4.1 is to be used and that it is "tightly coupled". There is a "deviceid" which identifies the DeviceInfo. Layout - This is per file and can be recalled by the server when it is no longer valid. For the FreeBSD server, there is support for two types of layout, call File and Flexible File layout. Both allow the client to do I/O on the DS via NFSv4.1 I/O operations. The Flexible File layout is a more recent variant that allows specification of mirrors, where the client is expected to do writes to all mirrors to maintain them in a consistent state. The Flexible File layout also allows the client to report I/O errors for a DS back to the MDS. The Flexible File layout supports two variants referred to as "tightly coupled" vs "loosely coupled". The FreeBSD server always uses the "tightly coupled" variant where the client uses the same credentials to do I/O on the DS as it would on the MDS. For the "loosely coupled" variant, the layout specifies a synthetic user/group that the client uses to do I/O on the DS. The FreeBSD server does not do striping and always returns layouts for the entire file. The critical information in a layout is Read vs Read/Writea and DeviceID(s) that identify which DS(s) the data is stored on. At this time, the MDS generates File Layout layouts to NFSv4.1 clients that know how to do pNFS for the non-mirrored DS case unless the sysctl vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File layouts are generated. The mirrored DS configuration always generates Flexible File layouts. For NFS clients that do not support NFSv4.1 pNFS, all I/O operations are done against the MDS which acts as a proxy for the appropriate DS(s). When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy. If the DS is on the same machine, the MDS/DS will do the RPC on the DS as a proxy and so on, until the machine runs out of some resource, such as session slots or mbufs. As such, DSs must be separate systems from the MDS. Tested by: james.rose@framestore.com Relnotes: yes
2018-06-12 19:36:32 +00:00
NFSDDSUNLOCK();
}
if (error == 0 && (nmp == NULL || *dsp == NULL)) {
if (nvpp != NULL && *nvpp != NULL) {
vput(*nvpp);
*nvpp = NULL;
}
error = ENXIO;
}
} else
vput(nd.ni_vp);
/*
* When dspathp != NULL and curdspathp == NULL, this is a recovery
* and is only allowed if there is a 0.0.0.0 IP address entry.
* When curdspathp != NULL, the ippos will be set to that entry.
*/
if (error == 0 && dspathp != NULL && ippos == -1) {
if (nvpp != NULL && *nvpp != NULL) {
vput(*nvpp);
*nvpp = NULL;
}
error = ENXIO;
}
if (error == 0) {
*vpp = vp;
pf = (struct pnfsdsfile *)buf;
if (ippos == -1) {
/* If no zeroip pnfsdsfile, add one. */
ippos = *buflenp / sizeof(*pf);
*buflenp += sizeof(*pf);
pf += ippos;
pf->dsf_dir = dsdir;
strlcpy(pf->dsf_filename, fname,
sizeof(pf->dsf_filename));
} else
pf += ippos;
*pfp = pf;
} else
vput(vp);
return (error);
}
Add an optional feature to the pNFS server. Without this patch, the pNFS server distributes the data storage files across all of the specified DSs. A tester noted that it would be nice if a system administrator could control which DSs are used to store the file data for a given exported MDS file system. This patch adds the kernel support to do this. It also makes a slight semantic change to nfsv4_findmirror(), since some uses of it no longer require that the DS being searched for have a current mirror. A patch that will be committed in a few minutes will modify the nfsd daemon to support this feature. The patch should only affect sites using the pNFS server (specified via the "-p" command line option for nfsd. Suggested by: james.rose@framestore.com
2018-07-02 19:21:33 +00:00
/*
* Search for a matching pnfsd mirror device structure, base on the nmp arg.
* Return one if found, NULL otherwise.
*/
static struct nfsdevice *
nfsrv_findmirroredds(struct nfsmount *nmp)
{
struct nfsdevice *ds, *fndds;
int fndmirror;
mtx_assert(NFSDDSMUTEXPTR, MA_OWNED);
/*
* Search the DS server list for a match with nmp.
* Remove the DS entry if found and there is a mirror.
*/
fndds = NULL;
fndmirror = 0;
if (nfsrv_devidcnt == 0)
return (fndds);
TAILQ_FOREACH(ds, &nfsrv_devidhead, nfsdev_list) {
if (ds->nfsdev_nmp == nmp) {
NFSD_DEBUG(4, "nfsrv_findmirroredds: fnd main ds\n");
fndds = ds;
break;
}
}
if (fndds == NULL)
return (fndds);
if (fndds->nfsdev_mdsisset == 0 && nfsrv_faildscnt > 0)
fndmirror = 1;
Fix the pNFS server so that it handles the "#mds_path" check for mirrors. The recently added feature of the pNFS server will set an fsid for the MDS file system to define the file system a DS should store files for. For a case where a DS handling all file systems has failed, it was possible for the code to check for a mirror with a specified fs, even though nfsdev_mdsisset was 0, possibly causing a false successful check for a mirror. This patch adds a check for nfsdev_mdsisset != 0 to avoid this. It only affects the pNFS server for a rare case. Found via code inspection.
2018-07-04 19:46:26 +00:00
else if (fndds->nfsdev_mdsisset != 0) {
Add an optional feature to the pNFS server. Without this patch, the pNFS server distributes the data storage files across all of the specified DSs. A tester noted that it would be nice if a system administrator could control which DSs are used to store the file data for a given exported MDS file system. This patch adds the kernel support to do this. It also makes a slight semantic change to nfsv4_findmirror(), since some uses of it no longer require that the DS being searched for have a current mirror. A patch that will be committed in a few minutes will modify the nfsd daemon to support this feature. The patch should only affect sites using the pNFS server (specified via the "-p" command line option for nfsd. Suggested by: james.rose@framestore.com
2018-07-02 19:21:33 +00:00
/* For the fsid is set case, search for a mirror. */
TAILQ_FOREACH(ds, &nfsrv_devidhead, nfsdev_list) {
if (ds != fndds && ds->nfsdev_nmp != NULL &&
ds->nfsdev_mdsisset != 0 &&
Deduplicate fsid comparisons Comparing fsid_t objects requires internal knowledge of the fsid structure and yet this is duplicated across a number of places in the code. Simplify by creating a fsidcmp function (macro). Reviewed by: mjg, rmacklem Approved by: mav (mentor) MFC after: 1 week Sponsored by: iXsystems, Inc. Differential Revision: https://reviews.freebsd.org/D24749
2020-05-21 01:55:35 +00:00
fsidcmp(&ds->nfsdev_mdsfsid,
&fndds->nfsdev_mdsfsid) == 0) {
Add an optional feature to the pNFS server. Without this patch, the pNFS server distributes the data storage files across all of the specified DSs. A tester noted that it would be nice if a system administrator could control which DSs are used to store the file data for a given exported MDS file system. This patch adds the kernel support to do this. It also makes a slight semantic change to nfsv4_findmirror(), since some uses of it no longer require that the DS being searched for have a current mirror. A patch that will be committed in a few minutes will modify the nfsd daemon to support this feature. The patch should only affect sites using the pNFS server (specified via the "-p" command line option for nfsd. Suggested by: james.rose@framestore.com
2018-07-02 19:21:33 +00:00
fndmirror = 1;
break;
}
}
}
if (fndmirror == 0) {
NFSD_DEBUG(4, "nfsrv_findmirroredds: no mirror for DS\n");
return (NULL);
}
return (fndds);
}
nfsd: Fix the NFSv4.2 pNFS MDS server for NFSERR_NOSPC via LayoutError If a pNFS server's DS runs out of disk space, it replies NFSERR_NOSPC to the client doing writing. For the Linux client, it then sends a LayoutError RPC to the MDS server to tell it about the error and keeps retrying, doing repeated LayoutGets to the MDS and Write RPCs to the DS. The Linux client is "stuck" until disk space on the DS is free'd up unless a subsequent LayoutGet request is sent a NFSERR_NOSPC reply. The looping problem still occurs for NFSv4.1 mounts, but no fix for this is known at this time. This patch changes the pNFS MDS server to reply to LayoutGet operations with NFSERR_NOSPC once a LayoutError reports the problem, until the DS has available space. This keeps the Linux NFSv4.2 from looping. Found during recent testing because of issues w.r.t. a DS being out of space found during a recent IEFT NFSv4 working group testing event. MFC after: 2 weeks
2021-11-08 23:58:00 +00:00
/*
* Mark the appropriate devid and all associated layout as "out of space".
*/
void
nfsrv_marknospc(char *devid, bool setit)
{
struct nfsdevice *ds;
struct nfslayout *lyp;
struct nfslayouthash *lhyp;
int i;
NFSDDSLOCK();
TAILQ_FOREACH(ds, &nfsrv_devidhead, nfsdev_list) {
if (NFSBCMP(ds->nfsdev_deviceid, devid, NFSX_V4DEVICEID) == 0) {
NFSD_DEBUG(1, "nfsrv_marknospc: devid %d\n", setit);
ds->nfsdev_nospc = setit;
}
}
NFSDDSUNLOCK();
for (i = 0; i < nfsrv_layouthashsize; i++) {
lhyp = &nfslayouthash[i];
NFSLOCKLAYOUT(lhyp);
TAILQ_FOREACH(lyp, &lhyp->list, lay_list) {
if (NFSBCMP(lyp->lay_deviceid, devid,
NFSX_V4DEVICEID) == 0) {
NFSD_DEBUG(1, "nfsrv_marknospc: layout %d\n",
setit);
if (setit)
lyp->lay_flags |= NFSLAY_NOSPC;
else
lyp->lay_flags &= ~NFSLAY_NOSPC;
}
}
NFSUNLOCKLAYOUT(lhyp);
}
}
nfsd: Add support for the SP4_MACH_CRED case in ExchangeID Commit f4179ad46fa4 added support for operation bitmaps for NFSv4.1/4.2. This commit uses those to implement the SP4_MACH_CRED case for the NFSv4.1/4.2 ExchangeID operation since the Linux NFSv4.1/4.2 client is now using this for Kerberized mounts. The Linux Kerberized NFSv4.1/4.2 mounts currently work without support for this because Linux will fall back to SP4_NONE, but there is no guarantee this fallback will work forever. This commit only affects Kerberized NFSv4.1/4.2 mounts from Linux at this time. MFC after: 3 months
2023-04-07 19:49:23 +00:00
/*
* Check to see if SP4_MACH_CRED is in use and, if it is, check that the
* correct machine credential is being used.
*/
static int
nfsrv_checkmachcred(int op, struct nfsrv_descript *nd, struct nfsclient *clp)
{
if ((clp->lc_flags & LCL_MACHCRED) == 0 ||
!NFSISSET_OPBIT(&clp->lc_mustops, op))
return (0);
KASSERT((nd->nd_flag & ND_NFSV41) != 0,
("nfsrv_checkmachcred: MachCred for NFSv4.0"));
if ((nd->nd_flag & (ND_GSSINTEGRITY | ND_GSSPRIVACY)) != 0 &&
nd->nd_princlen == clp->lc_namelen &&
!NFSBCMP(nd->nd_principal, clp->lc_name, nd->nd_princlen))
return (0);
return (NFSERR_AUTHERR | AUTH_TOOWEAK);
}
Reference in New Issue Copy Permalink