d/freebsd-dev
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
29c8295312
freebsd-dev/sys/fs/nfsserver/nfs_nfsdport.c

6763 lines
181 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: further adoption of SPDX licensing ID tags. Mainly focus on files that use BSD 3-Clause license. 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. Special thanks to Wind River for providing access to "The Duke of Highlander" tool: an older (2014) run over FreeBSD tree was useful as a starting point.
2017-11-20 19:43:44 +00:00
* SPDX-License-Identifier: BSD-3-Clause
*
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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) 1989, 1993
* The Regents of the University of California. All rights reserved.
*
* This code is derived from software contributed to Berkeley by
* Rick Macklem at The University of Guelph.
*
* 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.
Renumber copyright clause 4 Renumber cluase 4 to 3, per what everybody else did when BSD granted them permission to remove clause 3. My insistance on keeping the same numbering for legal reasons is too pedantic, so give up on that point. Submitted by: Jan Schaumann <jschauma@stevens.edu> Pull Request: https://github.com/freebsd/freebsd/pull/96
2017-02-28 23:42:47 +00:00
* 3. Neither the name of the University nor the names of its contributors
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS 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 REGENTS 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("$FreeBSD$");
Update kernel inclusions of capability.h to use capsicum.h instead; some further refinement is required as some device drivers intended to be portable over FreeBSD versions rely on __FreeBSD_version to decide whether to include capability.h. MFC after: 3 weeks
2014-03-16 10:55:57 +00:00
#include <sys/capsicum.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>
Second-to-last commit implementing Capsicum capabilities in the FreeBSD kernel for FreeBSD 9.0: Add a new capability mask argument to fget(9) and friends, allowing system call code to declare what capabilities are required when an integer file descriptor is converted into an in-kernel struct file *. With options CAPABILITIES compiled into the kernel, this enforces capability protection; without, this change is effectively a no-op. Some cases require special handling, such as mmap(2), which must preserve information about the maximum rights at the time of mapping in the memory map so that they can later be enforced in mprotect(2) -- this is done by narrowing the rights in the existing max_protection field used for similar purposes with file permissions. In namei(9), we assert that the code is not reached from within capability mode, as we're not yet ready to enforce namespace capabilities there. This will follow in a later commit. Update two capability names: CAP_EVENT and CAP_KEVENT become CAP_POST_KEVENT and CAP_POLL_KEVENT to more accurately indicate what they represent. Approved by: re (bz) Submitted by: jonathan Sponsored by: Google Inc
2011-08-11 12:30:23 +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
/*
* Functions that perform the vfs operations required by the routines in
* nfsd_serv.c. It is hoped that this change will make the server more
* portable.
*/
#include <fs/nfs/nfsport.h>
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
#include <security/mac/mac_framework.h>
#include <sys/filio.h>
Modify the experimental NFSv4 server's file handle hash function to use the generic hash32_buf() function. Although adding the bytes seemed sufficient for UFS and ZFS, since most of the bytes are the same for file handles on the same volume, this might not be sufficient for other file systems. Use of a generic function also seems preferable to one specific to NFSv4. Suggested by: gleb.kurtsou at gmail.com MFC after: 10 days
2010-10-23 22:28:29 +00:00
#include <sys/hash.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 <sys/sysctl.h>
Modify sys/fs/nfsserver/nfs_nfsdport.c to use nlm_acquire_next_sysid() to set the l_sysid for locks correctly. Approved by: kib (mentor)
2009-05-21 01:50:27 +00:00
#include <nlm/nlm_prot.h>
#include <nlm/nlm.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
Add some FEATURE macros for various features (AUDIT/CAM/IPC/KTR/MAC/NFS/NTP/ PMC/SYSV/...). No FreeBSD version bump, the userland application to query the features will be committed last and can serve as an indication of the availablility if needed. Sponsored by: Google Summer of Code 2010 Submitted by: kibab Reviewed by: arch@ (parts by rwatson, trasz, jhb) X-MFC after: to be determined in last commit with code from this project
2011-02-25 10:11:01 +00:00
FEATURE(nfsd, "NFSv4 server");
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 u_int32_t newnfs_true, newnfs_false, newnfs_xdrneg1;
extern int nfsrv_useacl;
extern int newnfs_numnfsd;
extern struct mount nfsv4root_mnt;
extern struct nfsrv_stablefirst nfsrv_stablefirst;
extern void (*nfsd_call_servertimer)(void);
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
extern SVCPOOL *nfsrvd_pool;
Add two new options to the nfssvc(2) syscall that allow processes running as root to suspend/resume execution of the kernel nfsd threads. An earlier version of this patch was tested by Vincent Hoffman (vince at unsane.co.uk) and John Hickey (jh at deterlab.net). Reviewed by: kib MFC after: 2 weeks
2012-10-14 22:33:17 +00:00
extern struct nfsv4lock nfsd_suspend_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
extern struct nfsclienthashhead *nfsclienthash;
extern struct nfslockhashhead *nfslockhash;
extern struct nfssessionhash *nfssessionhash;
extern int nfsrv_sessionhashsize;
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
extern struct nfsstatsv1 nfsstatsv1;
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 nfslayouthash *nfslayouthash;
extern int nfsrv_layouthashsize;
extern struct mtx nfsrv_dslock_mtx;
extern int nfs_pnfsiothreads;
extern struct nfsdontlisthead nfsrv_dontlisthead;
extern volatile int nfsrv_dontlistlen;
extern volatile int nfsrv_devidcnt;
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
struct vfsoptlist nfsv4root_opt, nfsv4root_newopt;
NFSDLOCKMUTEX;
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
NFSSTATESPINLOCK;
Fix several performance related issues in the new NFS server's DRC for NFS over TCP. - Increase the size of the hash tables. - Create a separate mutex for each hash list of the TCP hash table. - Single thread the code that deletes stale cache entries. - Add a tunable called vfs.nfsd.tcphighwater, which can be increased to allow the cache to grow larger, avoiding the overhead of frequent scans to delete stale cache entries. (The default value will result in frequent scans to delete stale cache entries, analagous to what the pre-patched code does.) - Add a tunable called vfs.nfsd.cachetcp that can be used to disable DRC caching for NFS over TCP, since the old NFS server didn't DRC cache TCP. It also adjusts the size of nfsrc_floodlevel dynamically, so that it is always greater than vfs.nfsd.tcphighwater. For UDP the algorithm remains the same as the pre-patched code, but the tunable vfs.nfsd.udphighwater can be used to allow the cache to grow larger and reduce the overhead caused by frequent scans for stale entries. UDP also uses a larger hash table size than the pre-patched code. Reported by: wollman Tested by: wollman (earlier version of patch) Submitted by: ivoras (earlier patch) Reviewed by: jhb (earlier version of patch) MFC after: 1 month
2013-08-14 21:11:26 +00:00
struct nfsrchash_bucket nfsrchash_table[NFSRVCACHE_HASHSIZE];
Rework NFS Duplicate Request Cache cleanup logic. - Introduce additional hash to group requests by hash of sockref. This allows to process TCP acknowledgements without looping though all the cache, and as result allows to do it every time. - Indroduce additional callbacks to notify application layer about sockets disconnection. Without this last few requests processed just before socket disconnection never processed their ACKs and stuck in cache for many hours. - Implement transport-specific method for tracking reply acknowledgements. New implementation does not cross multiple stack layers to get the data and does not have race conditions that previously made some requests stuck in cache. This could be done more efficiently at sockbuf layer, but that would broke some KBIs, while I don't know other consumers for it aside NFS. - Instead of traversing all DRC twice per request, run cleaning only once per request, and except in some conditions traverse only single hash slot at a time. Together this limits NFS DRC growth only to situations of real connectivity problems. If network is working well, and so all replies are acknowledged, cache remains almost empty even after hours of heavy load. Without this change on the same test cache was growing to many thousand requests even with perfectly working local network. As another result this reduces CPU time spent on the DRC handling during SPEC NFS benchmark from about 10% to 0.5%. Sponsored by: iXsystems, Inc.
2014-01-03 15:09:59 +00:00
struct nfsrchash_bucket nfsrcahash_table[NFSRVCACHE_HASHSIZE];
Fix several performance related issues in the new NFS server's DRC for NFS over TCP. - Increase the size of the hash tables. - Create a separate mutex for each hash list of the TCP hash table. - Single thread the code that deletes stale cache entries. - Add a tunable called vfs.nfsd.tcphighwater, which can be increased to allow the cache to grow larger, avoiding the overhead of frequent scans to delete stale cache entries. (The default value will result in frequent scans to delete stale cache entries, analagous to what the pre-patched code does.) - Add a tunable called vfs.nfsd.cachetcp that can be used to disable DRC caching for NFS over TCP, since the old NFS server didn't DRC cache TCP. It also adjusts the size of nfsrc_floodlevel dynamically, so that it is always greater than vfs.nfsd.tcphighwater. For UDP the algorithm remains the same as the pre-patched code, but the tunable vfs.nfsd.udphighwater can be used to allow the cache to grow larger and reduce the overhead caused by frequent scans for stale entries. UDP also uses a larger hash table size than the pre-patched code. Reported by: wollman Tested by: wollman (earlier version of patch) Submitted by: ivoras (earlier patch) Reviewed by: jhb (earlier version of patch) MFC after: 1 month
2013-08-14 21:11:26 +00:00
struct mtx nfsrc_udpmtx;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 mtx nfs_v4root_mutex;
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 mtx nfsrv_dontlistlock_mtx;
struct mtx nfsrv_recalllock_mtx;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 nfsrvfh nfs_rootfh, nfs_pubfh;
int nfs_pubfhset = 0, nfs_rootfhset = 0;
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
struct proc *nfsd_master_proc = 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
int nfsd_debuglevel = 0;
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
static pid_t nfsd_master_pid = (pid_t)-1;
static char nfsd_master_comm[MAXCOMLEN + 1];
static struct timeval nfsd_master_start;
Modify sys/fs/nfsserver/nfs_nfsdport.c to use nlm_acquire_next_sysid() to set the l_sysid for locks correctly. Approved by: kib (mentor)
2009-05-21 01:50:27 +00:00
static uint32_t nfsv4_sysid = 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
static fhandle_t zerofh;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
Added a SYSCTL to sys/fs/nfsserver/nfs_nfsdport.c so that the value of nfsrv_dolocallocks can be changed via sysctl. I also added some non-empty descriptor strings and reformatted some overly long lines. Approved by: kib (mentor)
2009-05-17 17:54:01 +00:00
static int nfssvc_srvcall(struct thread *, struct nfssvc_args *,
struct ucred *);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 handles crossing of server mount points properly. The functions nfsvno_fillattr() and nfsv4_fillattr() were modified to take the extra arguments that are the mount point, a flag to indicate that it is a file system root and the mounted on fileno. The mount point argument needs to be busy when nfsvno_fillattr() is called, since the vp argument is not locked. Reviewed by: kib MFC after: 2 weeks
2011-04-14 21:49:52 +00:00
int nfsrv_enable_crossmntpt = 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
static int nfs_commit_blks;
static int nfs_commit_miss;
extern int nfsrv_issuedelegs;
Added a SYSCTL to sys/fs/nfsserver/nfs_nfsdport.c so that the value of nfsrv_dolocallocks can be changed via sysctl. I also added some non-empty descriptor strings and reformatted some overly long lines. Approved by: kib (mentor)
2009-05-17 17:54:01 +00:00
extern int nfsrv_dolocallocks;
The new draft specification for NFSv4.0 specifies that a server should either accept owner and owner_group strings that are just the digits of the uid/gid or return NFS4ERR_BADOWNER. This patch adds a sysctl vfs.nfsd.enable_stringtouid, which can be set to enable the server w.r.t. accepting numeric string. It also ensures that NFS4ERR_BADOWNER is returned if numeric uid/gid strings are not enabled. This fixes the server for recent Linux nfs4 clients that use numeric uid/gid strings by default. Reported and tested by: craigyk@gmail.com MFC after: 2 weeks
2014-05-03 00:13:45 +00:00
extern int nfsd_enable_stringtouid;
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 nfsdevicehead nfsrv_devidhead;
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
static int nfsrv_createiovec(int, struct mbuf **, struct mbuf **,
struct iovec **);
Add optional support for ext_pgs mbufs to the NFS server's read, readlink and getxattr operations. This patch optionally enables generation of read, readlink and getxattr replies in ext_pgs mbufs. Since neither of ND_EXTPG or ND_TLS are currently ever set, there is no change in semantics at this time. It also corrects the message in a couple of panic()s that should never occur. This is another in the series of commits that add support to the NFS client and server for building RPC messages in ext_pgs mbufs with anonymous pages. This is useful so that the entire mbuf list does not need to be copied before calling sosend() when NFS over TLS is enabled. Use of ext_pgs mbufs will not be enabled until the kernel RPC is updated to handle TLS.
2020-07-31 23:35:49 +00:00
static int nfsrv_createiovec_extpgs(int, int, struct mbuf **,
struct mbuf **, struct iovec **);
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
static int nfsrv_createiovecw(int, struct mbuf *, char *, struct iovec **,
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
static void nfsrv_pnfscreate(struct vnode *, struct vattr *, struct ucred *,
NFSPROC_T *);
static void nfsrv_pnfsremovesetup(struct vnode *, NFSPROC_T *, struct vnode **,
int *, char *, fhandle_t *);
static void nfsrv_pnfsremove(struct vnode **, int, char *, fhandle_t *,
NFSPROC_T *);
Delete the unused "nd" argument for nfsrv_proxyds(). The "nd" argument for nfsrv_proxyds() is no longer used by the function. This patch deletes it. This allows a subsequent patch to delete the "nd" argument from nfsvno_getattr(), since it's only use of "nd" was to pass it 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.
2019-09-05 22:25:19 +00:00
static int nfsrv_proxyds(struct vnode *, off_t, int, struct ucred *,
struct thread *, int, struct mbuf **, char *, struct mbuf **,
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 nfsvattr *, struct acl *, off_t *, int, bool *);
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_setextattr(struct vnode *, struct nfsvattr *, NFSPROC_T *);
static int nfsrv_readdsrpc(fhandle_t *, off_t, int, struct ucred *,
NFSPROC_T *, struct nfsmount *, struct mbuf **, struct mbuf **);
static int nfsrv_writedsrpc(fhandle_t *, off_t, int, struct ucred *,
NFSPROC_T *, struct vnode *, struct nfsmount **, int, struct mbuf **,
char *, 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
static int nfsrv_allocatedsrpc(fhandle_t *, off_t, off_t, struct ucred *,
NFSPROC_T *, struct vnode *, struct nfsmount **, int, 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
static int nfsrv_setacldsrpc(fhandle_t *, struct ucred *, NFSPROC_T *,
struct vnode *, struct nfsmount **, int, struct acl *, int *);
static int nfsrv_setattrdsrpc(fhandle_t *, struct ucred *, NFSPROC_T *,
struct vnode *, struct nfsmount **, int, struct nfsvattr *, int *);
static int nfsrv_getattrdsrpc(fhandle_t *, struct ucred *, NFSPROC_T *,
struct vnode *, struct nfsmount *, struct nfsvattr *);
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
static int nfsrv_seekdsrpc(fhandle_t *, off_t *, int, bool *, struct ucred *,
NFSPROC_T *, struct nfsmount *);
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_putfhname(fhandle_t *, char *);
static int nfsrv_pnfslookupds(struct vnode *, struct vnode *,
struct pnfsdsfile *, struct vnode **, NFSPROC_T *);
Fix LORs between vn_start_write() and vn_lock() in the pNFS server. When coding the pNFS server, I added several vn_start_write() calls done while the vnode was locked, not realizing I had introduced LORs and possible deadlock when an exported file system on the MDS is suspended. This patch fixes this by removing the added vn_start_write() calls and modifying the code so that the extant vn_start_write() call before the NFS RPC/operation is done when needed by the pNFS server. Flags are changed so that LayoutCommit and LayoutReturn now get a vn_start_write() done for them. When the pNFS server is enabled, the code now also changes the flags for Getattr, so that the vn_start_write() is done for Getattr, since it may need to do a vn_set_extattr(). The nfs_writerpc flag array was made global to the NFS server and renamed nfsrv_writerpc, which is consistent naming for globals in the NFS server. Thanks go to kib@ for reporting that doing vn_start_write() while the vnode is locked results in a LOR. This patch only affects the behaviour of the pNFS server.
2018-08-17 21:12:16 +00:00
static void nfsrv_pnfssetfh(struct vnode *, struct pnfsdsfile *, char *, char *,
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 vnode *, NFSPROC_T *);
static int nfsrv_dsremove(struct vnode *, char *, struct ucred *, NFSPROC_T *);
static int nfsrv_dssetacl(struct vnode *, struct acl *, struct ucred *,
NFSPROC_T *);
Fix the pNFS server's reporting of disk space usage for the "#<path>" case. The pNFS server would report the total disk space used and free for all of the DSs, even when certain DSs are assigned to the file system via the "#<path>" suffix used in the "nfsd -p" option argument. This patch fixes this case. It only reports usage for the file system that the argument vnode resides on. This is consistent with the non-pNFS NFSv4 server. In NFSv4 it is possible to have subtrees on other file systems, but these are not included in the usage information for NFSv4. Approved by: re (gjb)
2018-10-09 01:10:50 +00:00
static int nfsrv_pnfsstatfs(struct statfs *, struct mount *);
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 nfs_pnfsio(task_fn_t *, void *);
Added a SYSCTL to sys/fs/nfsserver/nfs_nfsdport.c so that the value of nfsrv_dolocallocks can be changed via sysctl. I also added some non-empty descriptor strings and reformatted some overly long lines. Approved by: kib (mentor)
2009-05-17 17:54:01 +00:00
Mark more nodes as CTLFLAG_MPSAFE or CTLFLAG_NEEDGIANT (17 of many) r357614 added CTLFLAG_NEEDGIANT to make it easier to find nodes that are still not MPSAFE (or already are but aren’t properly marked). Use it in preparation for a general review of all nodes. This is non-functional change that adds annotations to SYSCTL_NODE and SYSCTL_PROC nodes using one of the soon-to-be-required flags. Mark all obvious cases as MPSAFE. All entries that haven't been marked as MPSAFE before are by default marked as NEEDGIANT Approved by: kib (mentor, blanket) Commented by: kib, gallatin, melifaro Differential Revision: https://reviews.freebsd.org/D23718
2020-02-26 14:26:36 +00:00
SYSCTL_NODE(_vfs, OID_AUTO, nfsd, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
"NFS server");
Change the new NFS server so that it uses vfs.nfsd naming for its sysctls instead of vfs.newnfs. This separates the names from the ones used by the client.
2011-05-08 01:01:27 +00:00
SYSCTL_INT(_vfs_nfsd, OID_AUTO, mirrormnt, CTLFLAG_RW,
Modify the experimental NFSv4 server so that it handles crossing of server mount points properly. The functions nfsvno_fillattr() and nfsv4_fillattr() were modified to take the extra arguments that are the mount point, a flag to indicate that it is a file system root and the mounted on fileno. The mount point argument needs to be busy when nfsvno_fillattr() is called, since the vp argument is not locked. Reviewed by: kib MFC after: 2 weeks
2011-04-14 21:49:52 +00:00
&nfsrv_enable_crossmntpt, 0, "Enable nfsd to cross mount points");
Change the new NFS server so that it uses vfs.nfsd naming for its sysctls instead of vfs.newnfs. This separates the names from the ones used by the client.
2011-05-08 01:01:27 +00:00
SYSCTL_INT(_vfs_nfsd, OID_AUTO, commit_blks, CTLFLAG_RW, &nfs_commit_blks,
Added a SYSCTL to sys/fs/nfsserver/nfs_nfsdport.c so that the value of nfsrv_dolocallocks can be changed via sysctl. I also added some non-empty descriptor strings and reformatted some overly long lines. Approved by: kib (mentor)
2009-05-17 17:54:01 +00:00
0, "");
Change the new NFS server so that it uses vfs.nfsd naming for its sysctls instead of vfs.newnfs. This separates the names from the ones used by the client.
2011-05-08 01:01:27 +00:00
SYSCTL_INT(_vfs_nfsd, OID_AUTO, commit_miss, CTLFLAG_RW, &nfs_commit_miss,
Added a SYSCTL to sys/fs/nfsserver/nfs_nfsdport.c so that the value of nfsrv_dolocallocks can be changed via sysctl. I also added some non-empty descriptor strings and reformatted some overly long lines. Approved by: kib (mentor)
2009-05-17 17:54:01 +00:00
0, "");
Change the new NFS server so that it uses vfs.nfsd naming for its sysctls instead of vfs.newnfs. This separates the names from the ones used by the client.
2011-05-08 01:01:27 +00:00
SYSCTL_INT(_vfs_nfsd, OID_AUTO, issue_delegations, CTLFLAG_RW,
Added a SYSCTL to sys/fs/nfsserver/nfs_nfsdport.c so that the value of nfsrv_dolocallocks can be changed via sysctl. I also added some non-empty descriptor strings and reformatted some overly long lines. Approved by: kib (mentor)
2009-05-17 17:54:01 +00:00
&nfsrv_issuedelegs, 0, "Enable nfsd to issue delegations");
Change the new NFS server so that it uses vfs.nfsd naming for its sysctls instead of vfs.newnfs. This separates the names from the ones used by the client.
2011-05-08 01:01:27 +00:00
SYSCTL_INT(_vfs_nfsd, OID_AUTO, enable_locallocks, CTLFLAG_RW,
Added a SYSCTL to sys/fs/nfsserver/nfs_nfsdport.c so that the value of nfsrv_dolocallocks can be changed via sysctl. I also added some non-empty descriptor strings and reformatted some overly long lines. Approved by: kib (mentor)
2009-05-17 17:54:01 +00:00
&nfsrv_dolocallocks, 0, "Enable nfsd to acquire local locks on files");
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
SYSCTL_INT(_vfs_nfsd, OID_AUTO, debuglevel, CTLFLAG_RW, &nfsd_debuglevel,
Replace "new NFS" with just "NFS" in some sysctl description strings. Sponsored by: The FreeBSD Foundation
2015-04-19 06:18:41 +00:00
0, "Debug level for NFS server");
The new draft specification for NFSv4.0 specifies that a server should either accept owner and owner_group strings that are just the digits of the uid/gid or return NFS4ERR_BADOWNER. This patch adds a sysctl vfs.nfsd.enable_stringtouid, which can be set to enable the server w.r.t. accepting numeric string. It also ensures that NFS4ERR_BADOWNER is returned if numeric uid/gid strings are not enabled. This fixes the server for recent Linux nfs4 clients that use numeric uid/gid strings by default. Reported and tested by: craigyk@gmail.com MFC after: 2 weeks
2014-05-03 00:13:45 +00:00
SYSCTL_INT(_vfs_nfsd, OID_AUTO, enable_stringtouid, CTLFLAG_RW,
&nfsd_enable_stringtouid, 0, "Enable nfsd to accept numeric owner_names");
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_pnfsgetdsattr = 1;
SYSCTL_INT(_vfs_nfsd, OID_AUTO, pnfsgetdsattr, CTLFLAG_RW,
&nfsrv_pnfsgetdsattr, 0, "When set getattr gets DS attributes via RPC");
/*
* nfsrv_dsdirsize can only be increased and only when the nfsd threads are
* not running.
* The dsN subdirectories for the increased values must have been created
* on all DS servers before this increase is done.
*/
u_int nfsrv_dsdirsize = 20;
static int
sysctl_dsdirsize(SYSCTL_HANDLER_ARGS)
{
int error, newdsdirsize;
newdsdirsize = nfsrv_dsdirsize;
error = sysctl_handle_int(oidp, &newdsdirsize, 0, req);
if (error != 0 || req->newptr == NULL)
return (error);
if (newdsdirsize <= nfsrv_dsdirsize || newdsdirsize > 10000 ||
newnfs_numnfsd != 0)
return (EINVAL);
nfsrv_dsdirsize = newdsdirsize;
return (0);
}
nfs: add missing CLTFLAG_MPSAFE annotations
2020-01-15 01:31:57 +00:00
SYSCTL_PROC(_vfs_nfsd, OID_AUTO, dsdirsize,
CTLTYPE_UINT | CTLFLAG_MPSAFE | CTLFLAG_RW, 0, sizeof(nfsrv_dsdirsize),
sysctl_dsdirsize, "IU", "Number of dsN subdirs on the DS servers");
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
Enhance the sequential access heuristic used to perform readahead in the NFS server and reuse it for writes as well to allow writes to the backing store to be clustered. - Use a prime number for the size of the heuristic table (1017 is not prime). - Move the logic to locate a heuristic entry from the table and compute the sequential count out of VOP_READ() and into a separate routine. - Use the logic from sequential_heuristic() in vfs_vnops.c to update the seqcount when a sequential access is performed rather than just increasing seqcount by 1. This lets the clustering count ramp up faster. - Allow for some reordering of RPCs and if it is detected leave the current seqcount as-is rather than dropping back to a seqcount of 1. Also, when out of order access is encountered, cut seqcount in half rather than dropping it all the way back to 1 to further aid with reordering. - Fix the new NFS server to properly update the next offset after a successful VOP_READ() so that the readahead actually works. Some of these changes came from an earlier patch by Bjorn Gronwall that was forwarded to me by bde@. Discussed with: bde, rmacklem, fs@ Submitted by: Bjorn Gronwall (1, 4) MFC after: 2 weeks
2011-12-01 18:46:28 +00:00
#define MAX_REORDERED_RPC 16
#define NUM_HEURISTIC 1031
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
#define NHUSE_INIT 64
#define NHUSE_INC 16
#define NHUSE_MAX 2048
static struct nfsheur {
struct vnode *nh_vp; /* vp to match (unreferenced pointer) */
Enhance the sequential access heuristic used to perform readahead in the NFS server and reuse it for writes as well to allow writes to the backing store to be clustered. - Use a prime number for the size of the heuristic table (1017 is not prime). - Move the logic to locate a heuristic entry from the table and compute the sequential count out of VOP_READ() and into a separate routine. - Use the logic from sequential_heuristic() in vfs_vnops.c to update the seqcount when a sequential access is performed rather than just increasing seqcount by 1. This lets the clustering count ramp up faster. - Allow for some reordering of RPCs and if it is detected leave the current seqcount as-is rather than dropping back to a seqcount of 1. Also, when out of order access is encountered, cut seqcount in half rather than dropping it all the way back to 1 to further aid with reordering. - Fix the new NFS server to properly update the next offset after a successful VOP_READ() so that the readahead actually works. Some of these changes came from an earlier patch by Bjorn Gronwall that was forwarded to me by bde@. Discussed with: bde, rmacklem, fs@ Submitted by: Bjorn Gronwall (1, 4) MFC after: 2 weeks
2011-12-01 18:46:28 +00:00
off_t nh_nextoff; /* next offset for sequential detection */
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 nh_use; /* use count for selection */
int nh_seqcount; /* heuristic */
} nfsheur[NUM_HEURISTIC];
Enhance the sequential access heuristic used to perform readahead in the NFS server and reuse it for writes as well to allow writes to the backing store to be clustered. - Use a prime number for the size of the heuristic table (1017 is not prime). - Move the logic to locate a heuristic entry from the table and compute the sequential count out of VOP_READ() and into a separate routine. - Use the logic from sequential_heuristic() in vfs_vnops.c to update the seqcount when a sequential access is performed rather than just increasing seqcount by 1. This lets the clustering count ramp up faster. - Allow for some reordering of RPCs and if it is detected leave the current seqcount as-is rather than dropping back to a seqcount of 1. Also, when out of order access is encountered, cut seqcount in half rather than dropping it all the way back to 1 to further aid with reordering. - Fix the new NFS server to properly update the next offset after a successful VOP_READ() so that the readahead actually works. Some of these changes came from an earlier patch by Bjorn Gronwall that was forwarded to me by bde@. Discussed with: bde, rmacklem, fs@ Submitted by: Bjorn Gronwall (1, 4) MFC after: 2 weeks
2011-12-01 18:46:28 +00:00
/*
* Heuristic to detect sequential operation.
*/
static struct nfsheur *
nfsrv_sequential_heuristic(struct uio *uio, struct vnode *vp)
{
struct nfsheur *nh;
int hi, try;
/* Locate best candidate. */
try = 32;
hi = ((int)(vm_offset_t)vp / sizeof(struct vnode)) % NUM_HEURISTIC;
nh = &nfsheur[hi];
while (try--) {
if (nfsheur[hi].nh_vp == vp) {
nh = &nfsheur[hi];
break;
}
if (nfsheur[hi].nh_use > 0)
--nfsheur[hi].nh_use;
hi = (hi + 1) % NUM_HEURISTIC;
if (nfsheur[hi].nh_use < nh->nh_use)
nh = &nfsheur[hi];
}
/* Initialize hint if this is a new file. */
if (nh->nh_vp != vp) {
nh->nh_vp = vp;
nh->nh_nextoff = uio->uio_offset;
nh->nh_use = NHUSE_INIT;
if (uio->uio_offset == 0)
nh->nh_seqcount = 4;
else
nh->nh_seqcount = 1;
}
/* Calculate heuristic. */
if ((uio->uio_offset == 0 && nh->nh_seqcount > 0) ||
uio->uio_offset == nh->nh_nextoff) {
/* See comments in vfs_vnops.c:sequential_heuristic(). */
nh->nh_seqcount += howmany(uio->uio_resid, 16384);
if (nh->nh_seqcount > IO_SEQMAX)
nh->nh_seqcount = IO_SEQMAX;
} else if (qabs(uio->uio_offset - nh->nh_nextoff) <= MAX_REORDERED_RPC *
imax(vp->v_mount->mnt_stat.f_iosize, uio->uio_resid)) {
/* Probably a reordered RPC, leave seqcount alone. */
} else if (nh->nh_seqcount > 1) {
nh->nh_seqcount /= 2;
} else {
nh->nh_seqcount = 0;
}
nh->nh_use += NHUSE_INC;
if (nh->nh_use > NHUSE_MAX)
nh->nh_use = NHUSE_MAX;
return (nh);
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 attributes into nfsvattr structure.
*/
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
nfsvno_getattr(struct vnode *vp, struct nfsvattr *nvap,
struct nfsrv_descript *nd, struct thread *p, int vpislocked,
nfsattrbit_t *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
{
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, gotattr, lockedit = 0;
struct nfsvattr na;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 an argument to nfsvno_getattr() in the experimental NFS server, so that it can avoid calling VOP_ISLOCKED() when the vnode is known to be locked. This will allow LK_SHARED to be used for these cases, which happen to be all the cases that can use LK_SHARED. This does not fix any bug, but it reduces the number of calls to VOP_ISLOCKED() and prepares the code so that it can be switched to using LK_SHARED in a future patch. Reviewed by: kib MFC after: 2 weeks
2010-12-24 21:31:18 +00:00
if (vpislocked == 0) {
/*
* When vpislocked == 0, the vnode is either exclusively
* locked by this thread or not locked by this thread.
* As such, shared lock it, if not exclusively locked.
*/
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
if (NFSVOPISLOCKED(vp) != LK_EXCLUSIVE) {
Add an argument to nfsvno_getattr() in the experimental NFS server, so that it can avoid calling VOP_ISLOCKED() when the vnode is known to be locked. This will allow LK_SHARED to be used for these cases, which happen to be all the cases that can use LK_SHARED. This does not fix any bug, but it reduces the number of calls to VOP_ISLOCKED() and prepares the code so that it can be switched to using LK_SHARED in a future patch. Reviewed by: kib MFC after: 2 weeks
2010-12-24 21:31:18 +00:00
lockedit = 1;
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
NFSVOPLOCK(vp, LK_SHARED | LK_RETRY);
Add an argument to nfsvno_getattr() in the experimental NFS server, so that it can avoid calling VOP_ISLOCKED() when the vnode is known to be locked. This will allow LK_SHARED to be used for these cases, which happen to be all the cases that can use LK_SHARED. This does not fix any bug, but it reduces the number of calls to VOP_ISLOCKED() and prepares the code so that it can be switched to using LK_SHARED in a future patch. Reviewed by: kib MFC after: 2 weeks
2010-12-24 21:31:18 +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
}
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
/*
Fix the pNFS server's reporting of SpaceUsed (va_bytes). The pNFS server currently reports SpaceUsed (va_bytes) for the metadata file. This in not correct, since the metadata file is always empty and, as such, va_bytes is just the allocation for the empty file. This patch adds va_bytes to the list of attributes acquired from the DS for a file, so that it includes the allocated data size and is updated when the file is written. For files created on a pNFS server before this patch is applied, the va_bytes value is estimated by rounding va_size up to a multiple of BLKDEV_IOSIZE. Once the file is written after this patch has been applied to the metadata server, the va_bytes returned for the file will be correct. This patch only affects a pNFS metadata server. Found during testing of the NFSv4.2 pNFS server for the Allocate operation. (Not yet in head/current.) MFC after: 2 weeks
2019-11-22 00:22:55 +00:00
* Acquire the Change, Size, TimeAccess, TimeModify and SpaceUsed
* attributes, as required.
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 needs to be done for regular files if:
* - non-NFSv4 RPCs or
* - when attrbitp == NULL or
* - an NFSv4 RPC with any of the above attributes in attrbitp.
* A return of 0 for nfsrv_proxyds() indicates that it has acquired
* these attributes. nfsrv_proxyds() will return an error if the
* server is not a pNFS one.
*/
gotattr = 0;
if (vp->v_type == VREG && nfsrv_devidcnt > 0 && (attrbitp == NULL ||
(nd->nd_flag & ND_NFSV4) == 0 ||
NFSISSET_ATTRBIT(attrbitp, NFSATTRBIT_CHANGE) ||
NFSISSET_ATTRBIT(attrbitp, NFSATTRBIT_SIZE) ||
NFSISSET_ATTRBIT(attrbitp, NFSATTRBIT_TIMEACCESS) ||
Fix the pNFS server's reporting of SpaceUsed (va_bytes). The pNFS server currently reports SpaceUsed (va_bytes) for the metadata file. This in not correct, since the metadata file is always empty and, as such, va_bytes is just the allocation for the empty file. This patch adds va_bytes to the list of attributes acquired from the DS for a file, so that it includes the allocated data size and is updated when the file is written. For files created on a pNFS server before this patch is applied, the va_bytes value is estimated by rounding va_size up to a multiple of BLKDEV_IOSIZE. Once the file is written after this patch has been applied to the metadata server, the va_bytes returned for the file will be correct. This patch only affects a pNFS metadata server. Found during testing of the NFSv4.2 pNFS server for the Allocate operation. (Not yet in head/current.) MFC after: 2 weeks
2019-11-22 00:22:55 +00:00
NFSISSET_ATTRBIT(attrbitp, NFSATTRBIT_TIMEMODIFY) ||
NFSISSET_ATTRBIT(attrbitp, NFSATTRBIT_SPACEUSED))) {
Delete the unused "nd" argument for nfsrv_proxyds(). The "nd" argument for nfsrv_proxyds() is no longer used by the function. This patch deletes it. This allows a subsequent patch to delete the "nd" argument from nfsvno_getattr(), since it's only use of "nd" was to pass it 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.
2019-09-05 22:25:19 +00:00
error = nfsrv_proxyds(vp, 0, 0, nd->nd_cred, p,
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
NFSPROC_GETATTR, NULL, NULL, NULL, &na, NULL, NULL, 0,
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
if (error == 0)
gotattr = 1;
}
error = VOP_GETATTR(vp, &nvap->na_vattr, nd->nd_cred);
Add an argument to nfsvno_getattr() in the experimental NFS server, so that it can avoid calling VOP_ISLOCKED() when the vnode is known to be locked. This will allow LK_SHARED to be used for these cases, which happen to be all the cases that can use LK_SHARED. This does not fix any bug, but it reduces the number of calls to VOP_ISLOCKED() and prepares the code so that it can be switched to using LK_SHARED in a future patch. Reviewed by: kib MFC after: 2 weeks
2010-12-24 21:31:18 +00:00
if (lockedit != 0)
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 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
/*
* If we got the Change, Size and Modify Time from the DS,
* replace them.
*/
if (gotattr != 0) {
nvap->na_atime = na.na_atime;
nvap->na_mtime = na.na_mtime;
nvap->na_filerev = na.na_filerev;
nvap->na_size = na.na_size;
Fix the pNFS server's reporting of SpaceUsed (va_bytes). The pNFS server currently reports SpaceUsed (va_bytes) for the metadata file. This in not correct, since the metadata file is always empty and, as such, va_bytes is just the allocation for the empty file. This patch adds va_bytes to the list of attributes acquired from the DS for a file, so that it includes the allocated data size and is updated when the file is written. For files created on a pNFS server before this patch is applied, the va_bytes value is estimated by rounding va_size up to a multiple of BLKDEV_IOSIZE. Once the file is written after this patch has been applied to the metadata server, the va_bytes returned for the file will be correct. This patch only affects a pNFS metadata server. Found during testing of the NFSv4.2 pNFS server for the Allocate operation. (Not yet in head/current.) MFC after: 2 weeks
2019-11-22 00:22:55 +00:00
nvap->na_bytes = na.na_bytes;
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, "nfsvno_getattr: gotattr=%d err=%d chg=%ju\n", gotattr,
error, (uintmax_t)na.na_filerev);
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 a file handle for a vnode.
*/
int
nfsvno_getfh(struct vnode *vp, fhandle_t *fhp, struct thread *p)
{
int error;
NFSBZERO((caddr_t)fhp, sizeof(fhandle_t));
fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsid;
error = VOP_VPTOFH(vp, &fhp->fh_fid);
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);
}
/*
* Perform access checking for vnodes obtained from file handles that would
* refer to files already opened by a Unix client. You cannot just use
Modify the experimental server so that it uses VOP_ACCESSX(). This is necessary in order to enable NFSv4 ACL support. The argument to nfsvno_accchk() was changed to an accmode_t and the function nfsrv_aclaccess() was no longer needed and, therefore, deleted. Reviewed by: trasz MFC after: 2 weeks
2009-12-25 20:44:19 +00:00
* vn_writechk() and VOP_ACCESSX() for two reasons.
* 1 - You must check for exported rdonly as well as MNT_RDONLY for the write
* case.
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
* 2 - The owner is to be given access irrespective of mode bits for some
* operations, so that processes that chmod after opening a file don't
* break.
*/
int
Modify the experimental server so that it uses VOP_ACCESSX(). This is necessary in order to enable NFSv4 ACL support. The argument to nfsvno_accchk() was changed to an accmode_t and the function nfsrv_aclaccess() was no longer needed and, therefore, deleted. Reviewed by: trasz MFC after: 2 weeks
2009-12-25 20:44:19 +00:00
nfsvno_accchk(struct vnode *vp, accmode_t accmode, struct ucred *cred,
struct nfsexstuff *exp, struct thread *p, int override, int vpislocked,
u_int32_t *supportedtypep)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 vattr vattr;
int error = 0, getret = 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
if (vpislocked == 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 (NFSVOPLOCK(vp, LK_SHARED) != 0) {
error = EPERM;
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 the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 (accmode & VWRITE) {
/* Just vn_writechk() changed to check rdonly */
/*
* Disallow write attempts on read-only file systems;
* unless the file is a socket or a block or character
* device resident on the file system.
*/
if (NFSVNO_EXRDONLY(exp) ||
Modify the experimental server so that it uses VOP_ACCESSX(). This is necessary in order to enable NFSv4 ACL support. The argument to nfsvno_accchk() was changed to an accmode_t and the function nfsrv_aclaccess() was no longer needed and, therefore, deleted. Reviewed by: trasz MFC after: 2 weeks
2009-12-25 20:44:19 +00:00
(vp->v_mount->mnt_flag & MNT_RDONLY)) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
switch (vp->v_type) {
case VREG:
case VDIR:
case VLNK:
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 = EROFS;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
default:
break;
}
}
/*
* If there's shared text associated with
* the inode, try to free it up once. If
* we fail, we can't allow writing.
*/
Fix the mis-handling of the VV_TEXT on the nullfs vnodes. If you have a binary on a filesystem which is also mounted over by nullfs, you could execute the binary from the lower filesystem, or from the nullfs mount. When executed from lower filesystem, the lower vnode gets VV_TEXT flag set, and the file cannot be modified while the binary is active. But, if executed as the nullfs alias, only the nullfs vnode gets VV_TEXT set, and you still can open the lower vnode for write. Add a set of VOPs for the VV_TEXT query, set and clear operations, which are correctly bypassed to lower vnode. Tested by: pho (previous version) MFC after: 2 weeks
2012-09-28 11:25:02 +00:00
if (VOP_IS_TEXT(vp) && error == 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 = ETXTBSY;
}
if (error != 0) {
if (vpislocked == 0)
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 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
}
/*
* Should the override still be applied when ACLs are enabled?
*/
Modify the experimental server so that it uses VOP_ACCESSX(). This is necessary in order to enable NFSv4 ACL support. The argument to nfsvno_accchk() was changed to an accmode_t and the function nfsrv_aclaccess() was no longer needed and, therefore, deleted. Reviewed by: trasz MFC after: 2 weeks
2009-12-25 20:44:19 +00:00
error = VOP_ACCESSX(vp, accmode, cred, p);
if (error != 0 && (accmode & (VDELETE | VDELETE_CHILD))) {
/*
* Try again with VEXPLICIT_DENY, to see if the test for
* deletion is supported.
*/
error = VOP_ACCESSX(vp, accmode | VEXPLICIT_DENY, cred, p);
if (error == 0) {
if (vp->v_type == VDIR) {
accmode &= ~(VDELETE | VDELETE_CHILD);
accmode |= VWRITE;
error = VOP_ACCESSX(vp, accmode, cred, p);
} else if (supportedtypep != NULL) {
*supportedtypep &= ~NFSACCESS_DELETE;
}
}
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* Allow certain operations for the owner (reads and writes
* on files that are already open).
*/
if (override != NFSACCCHK_NOOVERRIDE &&
(error == EPERM || error == EACCES)) {
if (cred->cr_uid == 0 && (override & NFSACCCHK_ALLOWROOT))
error = 0;
else if (override & NFSACCCHK_ALLOWOWNER) {
getret = VOP_GETATTR(vp, &vattr, cred);
if (getret == 0 && cred->cr_uid == vattr.va_uid)
error = 0;
}
}
if (vpislocked == 0)
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 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);
}
/*
* Set attribute(s) vnop.
*/
int
nfsvno_setattr(struct vnode *vp, struct nfsvattr *nvap, struct ucred *cred,
struct thread *p, struct nfsexstuff *exp)
{
Don't set a file's size for the MDS file of a pNFS service. When a pNFS service is running, the size of the files created on the MDS are normally 0, since the data is written to the data files on the DS(s). However, without this patch, if a Setattr with a non-zero size was done by a client, the MDS file was set to that size. This was thought to be benign, but it turns out that files with a non-zero size plus extended attributes can cause a "ffs_truncate3" panic in UFS. Although the exact cause of this panic() has not been isolated, this patch avoids the panic() and leaves the MDS files in a consistent state of always having a size == 0. Note that these MDS files never store data. The patch also includes an unnecessary initialization of savsize in case some compiler or static analyser complains it might not be initialized. This patch only affects the NFS server when pNFS is enabled via the "-p" command line option on nfsd.
2018-08-17 12:32:38 +00:00
u_quad_t savsize = 0;
int error, savedit;
Fix the NFS server so that it sets va_birthtime. r362490 marked that the NFSv4 attribute TimeCreate (va_birthtime) is supported, but it did not change the NFS server code to actually do it. As such, errors could occur when unrolling a tarball onto an NFSv4 mounted volume, since setting TimeCreate would fail with a NFSERR_ATTRNOTSUPP reply. This patch fixes the server so that it does TimeCreate and also makes sure that TimeCreate will not be set for a DS file for a pNFS server. A separate commit will add a check to the NFSv4 client for support of the TimeCreate attribute before attempting to set it, to avoid a problem when mounting a server that does not support the attribute. The failures will still occur for r362490 or later kernels that do not have this patch, since they indicate support for the attribute, but do not actually support the attribute.
2020-07-26 23:03:41 +00:00
time_t savbtime;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 set a file's size for the MDS file of a pNFS service. When a pNFS service is running, the size of the files created on the MDS are normally 0, since the data is written to the data files on the DS(s). However, without this patch, if a Setattr with a non-zero size was done by a client, the MDS file was set to that size. This was thought to be benign, but it turns out that files with a non-zero size plus extended attributes can cause a "ffs_truncate3" panic in UFS. Although the exact cause of this panic() has not been isolated, this patch avoids the panic() and leaves the MDS files in a consistent state of always having a size == 0. Note that these MDS files never store data. The patch also includes an unnecessary initialization of savsize in case some compiler or static analyser complains it might not be initialized. This patch only affects the NFS server when pNFS is enabled via the "-p" command line option on nfsd.
2018-08-17 12:32:38 +00:00
/*
* If this is an exported file system and a pNFS service is running,
* don't VOP_SETATTR() of size for the MDS file system.
*/
savedit = 0;
error = 0;
if (vp->v_type == VREG && (vp->v_mount->mnt_flag & MNT_EXPORTED) != 0 &&
nfsrv_devidcnt != 0 && nvap->na_vattr.va_size != VNOVAL &&
nvap->na_vattr.va_size > 0) {
savsize = nvap->na_vattr.va_size;
nvap->na_vattr.va_size = VNOVAL;
if (nvap->na_vattr.va_uid != (uid_t)VNOVAL ||
nvap->na_vattr.va_gid != (gid_t)VNOVAL ||
nvap->na_vattr.va_mode != (mode_t)VNOVAL ||
nvap->na_vattr.va_atime.tv_sec != VNOVAL ||
nvap->na_vattr.va_mtime.tv_sec != VNOVAL)
savedit = 1;
else
savedit = 2;
}
if (savedit != 2)
error = VOP_SETATTR(vp, &nvap->na_vattr, cred);
if (savedit != 0)
nvap->na_vattr.va_size = savsize;
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 && (nvap->na_vattr.va_uid != (uid_t)VNOVAL ||
nvap->na_vattr.va_gid != (gid_t)VNOVAL ||
nvap->na_vattr.va_size != VNOVAL ||
nvap->na_vattr.va_mode != (mode_t)VNOVAL ||
nvap->na_vattr.va_atime.tv_sec != VNOVAL ||
nvap->na_vattr.va_mtime.tv_sec != VNOVAL)) {
Fix the NFS server so that it sets va_birthtime. r362490 marked that the NFSv4 attribute TimeCreate (va_birthtime) is supported, but it did not change the NFS server code to actually do it. As such, errors could occur when unrolling a tarball onto an NFSv4 mounted volume, since setting TimeCreate would fail with a NFSERR_ATTRNOTSUPP reply. This patch fixes the server so that it does TimeCreate and also makes sure that TimeCreate will not be set for a DS file for a pNFS server. A separate commit will add a check to the NFSv4 client for support of the TimeCreate attribute before attempting to set it, to avoid a problem when mounting a server that does not support the attribute. The failures will still occur for r362490 or later kernels that do not have this patch, since they indicate support for the attribute, but do not actually support the attribute.
2020-07-26 23:03:41 +00:00
/* Never modify birthtime on a DS file. */
savbtime = nvap->na_vattr.va_birthtime.tv_sec;
nvap->na_vattr.va_birthtime.tv_sec = VNOVAL;
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
/* For a pNFS server, set the attributes on the DS file. */
Delete the unused "nd" argument for nfsrv_proxyds(). The "nd" argument for nfsrv_proxyds() is no longer used by the function. This patch deletes it. This allows a subsequent patch to delete the "nd" argument from nfsvno_getattr(), since it's only use of "nd" was to pass it 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.
2019-09-05 22:25:19 +00:00
error = nfsrv_proxyds(vp, 0, 0, cred, p, NFSPROC_SETATTR,
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
NULL, NULL, NULL, nvap, NULL, NULL, 0, NULL);
Fix the NFS server so that it sets va_birthtime. r362490 marked that the NFSv4 attribute TimeCreate (va_birthtime) is supported, but it did not change the NFS server code to actually do it. As such, errors could occur when unrolling a tarball onto an NFSv4 mounted volume, since setting TimeCreate would fail with a NFSERR_ATTRNOTSUPP reply. This patch fixes the server so that it does TimeCreate and also makes sure that TimeCreate will not be set for a DS file for a pNFS server. A separate commit will add a check to the NFSv4 client for support of the TimeCreate attribute before attempting to set it, to avoid a problem when mounting a server that does not support the attribute. The failures will still occur for r362490 or later kernels that do not have this patch, since they indicate support for the attribute, but do not actually support the attribute.
2020-07-26 23:03:41 +00:00
nvap->na_vattr.va_birthtime.tv_sec = savbtime;
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 == ENOENT)
error = 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
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);
}
/*
Delete a couple of out of date comments that are no longer true in the new NFS client. Requested by: bde MFC after: 1 week
2012-02-16 02:19:53 +00:00
* Set up nameidata for a lookup() call and do it.
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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
nfsvno_namei(struct nfsrv_descript *nd, struct nameidata *ndp,
struct vnode *dp, int islocked, struct nfsexstuff *exp, struct thread *p,
struct vnode **retdirp)
{
struct componentname *cnp = &ndp->ni_cnd;
int i;
struct iovec aiov;
struct uio auio;
int lockleaf = (cnp->cn_flags & LOCKLEAF) != 0, linklen;
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
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
char *cp;
*retdirp = NULL;
cnp->cn_nameptr = cnp->cn_pnbuf;
Allow some dotdot lookups in capability mode. If dotdot lookup does not escape from the file descriptor passed as the lookup root, we can allow the component traversal. Track the directories traversed, and check the result of dotdot lookup against the recorded list of the directory vnodes. Dotdot lookups are enabled by sysctl vfs.lookup_cap_dotdot, currently disabled by default until more verification of the approach is done. Disallow non-local filesystems for dotdot, since remote server might conspire with the local process to allow it to escape the namespace. This might be too cautious, provide the knob vfs.lookup_cap_dotdot_nonlocal to override as well. Idea by: rwatson Discussed with: emaste, jonathan, rwatson Reviewed by: mjg (previous version) Tested by: pho (previous version) Sponsored by: The FreeBSD Foundation MFC after: 2 week Differential revision: https://reviews.freebsd.org/D8110
2016-11-02 12:43:15 +00:00
ndp->ni_lcf = 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
/*
* Extract and set starting directory.
*/
if (dp->v_type != VDIR) {
if (islocked)
vput(dp);
else
vrele(dp);
nfsvno_relpathbuf(ndp);
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 = ENOTDIR;
goto out1;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 (islocked)
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(dp);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
VREF(dp);
*retdirp = dp;
if (NFSVNO_EXRDONLY(exp))
cnp->cn_flags |= RDONLY;
ndp->ni_segflg = UIO_SYSSPACE;
if (nd->nd_flag & ND_PUBLOOKUP) {
ndp->ni_loopcnt = 0;
if (cnp->cn_pnbuf[0] == '/') {
vrele(dp);
/*
* Check for degenerate pathnames here, since lookup()
* panics on them.
*/
for (i = 1; i < ndp->ni_pathlen; i++)
if (cnp->cn_pnbuf[i] != '/')
break;
if (i == ndp->ni_pathlen) {
error = NFSERR_ACCES;
goto out;
}
dp = rootvnode;
VREF(dp);
}
Modify the experimental NFSv4 server so that it handles crossing of server mount points properly. The functions nfsvno_fillattr() and nfsv4_fillattr() were modified to take the extra arguments that are the mount point, a flag to indicate that it is a file system root and the mounted on fileno. The mount point argument needs to be busy when nfsvno_fillattr() is called, since the vp argument is not locked. Reviewed by: kib MFC after: 2 weeks
2011-04-14 21:49:52 +00:00
} else if ((nfsrv_enable_crossmntpt == 0 && NFSVNO_EXPORTED(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
(nd->nd_flag & ND_NFSV4) == 0) {
/*
* Only cross mount points for NFSv4 when doing a
* mount while traversing the file system above
Modify the experimental NFSv4 server so that it handles crossing of server mount points properly. The functions nfsvno_fillattr() and nfsv4_fillattr() were modified to take the extra arguments that are the mount point, a flag to indicate that it is a file system root and the mounted on fileno. The mount point argument needs to be busy when nfsvno_fillattr() is called, since the vp argument is not locked. Reviewed by: kib MFC after: 2 weeks
2011-04-14 21:49:52 +00:00
* the mount point, unless nfsrv_enable_crossmntpt is set.
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
*/
cnp->cn_flags |= NOCROSSMOUNT;
}
/*
* Initialize for scan, set ni_startdir and bump ref on dp again
Fix typo in comment. Submitted by: Alex Weber <alexwebr@gmail.com> MFC after: 1 week
2013-05-15 08:38:49 +00:00
* because lookup() will dereference ni_startdir.
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
*/
cnp->cn_thread = p;
ndp->ni_startdir = dp;
ndp->ni_rootdir = rootvnode;
The name caching changes of r230394 exposed an intermittent bug in the new NFS server for NFSv4, where it would report ENOENT when the file actually existed on the server. This turned out to be caused by not initializing ni_topdir before calling lookup() and there was a rare case where the value on the stack location assigned to ni_topdir happened to be a pointer to a ".." entry, such that "dp == ndp->ni_topdir" succeeded in lookup(). This patch initializes ni_topdir to fix the problem. MFC after: 5 days
2012-03-03 16:13:20 +00:00
ndp->ni_topdir = 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 (!lockleaf)
cnp->cn_flags |= LOCKLEAF;
for (;;) {
cnp->cn_nameptr = cnp->cn_pnbuf;
/*
* Call lookup() to do the real work. If an error occurs,
* ndp->ni_vp and ni_dvp are left uninitialized or NULL and
* we do not have to dereference anything before returning.
* In either case ni_startdir will be dereferenced and NULLed
* out.
*/
error = lookup(ndp);
if (error)
break;
/*
* Check for encountering a symbolic link. Trivial
* termination occurs if no symlink encountered.
*/
if ((cnp->cn_flags & ISSYMLINK) == 0) {
if ((cnp->cn_flags & (SAVENAME | SAVESTART)) == 0)
nfsvno_relpathbuf(ndp);
if (ndp->ni_vp && !lockleaf)
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(ndp->ni_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
break;
}
/*
* Validate symlink
*/
if ((cnp->cn_flags & LOCKPARENT) && ndp->ni_pathlen == 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(ndp->ni_dvp);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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_PUBLOOKUP)) {
error = EINVAL;
goto badlink2;
}
if (ndp->ni_loopcnt++ >= MAXSYMLINKS) {
error = ELOOP;
goto badlink2;
}
if (ndp->ni_pathlen > 1)
cp = uma_zalloc(namei_zone, M_WAITOK);
else
cp = cnp->cn_pnbuf;
aiov.iov_base = cp;
aiov.iov_len = MAXPATHLEN;
auio.uio_iov = &aiov;
auio.uio_iovcnt = 1;
auio.uio_offset = 0;
auio.uio_rw = UIO_READ;
auio.uio_segflg = UIO_SYSSPACE;
auio.uio_td = NULL;
auio.uio_resid = MAXPATHLEN;
error = VOP_READLINK(ndp->ni_vp, &auio, cnp->cn_cred);
if (error) {
badlink1:
if (ndp->ni_pathlen > 1)
uma_zfree(namei_zone, cp);
badlink2:
vrele(ndp->ni_dvp);
vput(ndp->ni_vp);
break;
}
linklen = MAXPATHLEN - auio.uio_resid;
if (linklen == 0) {
error = ENOENT;
goto badlink1;
}
if (linklen + ndp->ni_pathlen >= MAXPATHLEN) {
error = ENAMETOOLONG;
goto badlink1;
}
/*
* Adjust or replace path
*/
if (ndp->ni_pathlen > 1) {
NFSBCOPY(ndp->ni_next, cp + linklen, ndp->ni_pathlen);
uma_zfree(namei_zone, cnp->cn_pnbuf);
cnp->cn_pnbuf = cp;
} else
cnp->cn_pnbuf[linklen] = '\0';
ndp->ni_pathlen += linklen;
/*
* Cleanup refs for next loop and check if root directory
* should replace current directory. Normally ni_dvp
* becomes the new base directory and is cleaned up when
* we loop. Explicitly null pointers after invalidation
* to clarify operation.
*/
vput(ndp->ni_vp);
ndp->ni_vp = NULL;
if (cnp->cn_pnbuf[0] == '/') {
vrele(ndp->ni_dvp);
ndp->ni_dvp = ndp->ni_rootdir;
VREF(ndp->ni_dvp);
}
ndp->ni_startdir = ndp->ni_dvp;
ndp->ni_dvp = NULL;
}
if (!lockleaf)
cnp->cn_flags &= ~LOCKLEAF;
out:
if (error) {
Use the common api helper routine instead of freeing the namei buffer directly. Approved by: rmacklem (mentor) MFC after: 1 month
2012-05-08 03:39:44 +00:00
nfsvno_relpathbuf(ndp);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
ndp->ni_vp = NULL;
ndp->ni_dvp = NULL;
ndp->ni_startdir = NULL;
} else if ((ndp->ni_cnd.cn_flags & (WANTPARENT|LOCKPARENT)) == 0) {
ndp->ni_dvp = 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
out1:
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);
}
/*
* Set up a pathname buffer and return a pointer to it and, optionally
* set a hash pointer.
*/
void
nfsvno_setpathbuf(struct nameidata *ndp, char **bufpp, u_long **hashpp)
{
struct componentname *cnp = &ndp->ni_cnd;
cnp->cn_flags |= (NOMACCHECK | HASBUF);
cnp->cn_pnbuf = uma_zalloc(namei_zone, M_WAITOK);
if (hashpp != NULL)
*hashpp = NULL;
*bufpp = cnp->cn_pnbuf;
}
/*
* Release the above path buffer, if not released by nfsvno_namei().
*/
void
nfsvno_relpathbuf(struct nameidata *ndp)
{
if ((ndp->ni_cnd.cn_flags & HASBUF) == 0)
panic("nfsrelpath");
uma_zfree(namei_zone, ndp->ni_cnd.cn_pnbuf);
ndp->ni_cnd.cn_flags &= ~HASBUF;
}
/*
* Readlink vnode op into an mbuf list.
*/
int
Add optional support for ext_pgs mbufs to the NFS server's read, readlink and getxattr operations. This patch optionally enables generation of read, readlink and getxattr replies in ext_pgs mbufs. Since neither of ND_EXTPG or ND_TLS are currently ever set, there is no change in semantics at this time. It also corrects the message in a couple of panic()s that should never occur. This is another in the series of commits that add support to the NFS client and server for building RPC messages in ext_pgs mbufs with anonymous pages. This is useful so that the entire mbuf list does not need to be copied before calling sosend() when NFS over TLS is enabled. Use of ext_pgs mbufs will not be enabled until the kernel RPC is updated to handle TLS.
2020-07-31 23:35:49 +00:00
nfsvno_readlink(struct vnode *vp, struct ucred *cred, int maxextsiz,
struct thread *p, struct mbuf **mpp, struct mbuf **mpendp, int *lenp)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 iovec *iv;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 uio io, *uiop = &io;
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 mbuf *mp, *mp3;
int len, tlen, 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
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
len = NFS_MAXPATHLEN;
Add optional support for ext_pgs mbufs to the NFS server's read, readlink and getxattr operations. This patch optionally enables generation of read, readlink and getxattr replies in ext_pgs mbufs. Since neither of ND_EXTPG or ND_TLS are currently ever set, there is no change in semantics at this time. It also corrects the message in a couple of panic()s that should never occur. This is another in the series of commits that add support to the NFS client and server for building RPC messages in ext_pgs mbufs with anonymous pages. This is useful so that the entire mbuf list does not need to be copied before calling sosend() when NFS over TLS is enabled. Use of ext_pgs mbufs will not be enabled until the kernel RPC is updated to handle TLS.
2020-07-31 23:35:49 +00:00
if (maxextsiz > 0)
uiop->uio_iovcnt = nfsrv_createiovec_extpgs(len, maxextsiz,
&mp3, &mp, &iv);
else
uiop->uio_iovcnt = nfsrv_createiovec(len, &mp3, &mp, &iv);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
uiop->uio_iov = iv;
uiop->uio_offset = 0;
uiop->uio_resid = len;
uiop->uio_rw = UIO_READ;
uiop->uio_segflg = UIO_SYSSPACE;
uiop->uio_td = NULL;
error = VOP_READLINK(vp, uiop, cred);
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
free(iv, 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) {
m_freem(mp3);
*lenp = 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
}
if (uiop->uio_resid > 0) {
len -= uiop->uio_resid;
tlen = NFSM_RNDUP(len);
Add support for ext_pgs mbufs to nfsrv_adj(). This patch uses a slightly different algorithm for nfsrv_adj() since ext_pgs mbuf lists are not permitted to have m_len == 0 mbufs. As such, the code now frees mbufs after the adjustment in the list instead of setting their m_len field to 0. Since mbuf(s) may be trimmed off the tail of the list, the function now returns a pointer to the last mbuf in the list. This saves the caller from needing to use m_last() to find the last mbuf. It also implies that it might return a nul list, which required a check for that in nfsrvd_readlink(). This is another in the series of commits that add support to the NFS client and server for building RPC messages in ext_pgs mbufs with anonymous pages. This is useful so that the entire mbuf list does not need to be copied before calling sosend() when NFS over TLS is enabled. Use of ext_pgs mbufs will not be enabled until the kernel RPC is updated to handle TLS.
2020-07-26 02:42:09 +00:00
if (tlen == 0) {
m_freem(mp3);
mp3 = mp = NULL;
} else if (tlen != NFS_MAXPATHLEN || tlen != len)
mp = nfsrv_adj(mp3, NFS_MAXPATHLEN - tlen,
tlen - 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
}
*lenp = len;
*mpp = mp3;
*mpendp = mp;
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
}
/*
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
* Create an mbuf chain and an associated iovec that can be used to Read
* or Getextattr of data.
* Upon success, return pointers to the first and last mbufs in the chain
* plus the malloc'd iovec and its iovlen.
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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
static int
nfsrv_createiovec(int len, struct mbuf **mpp, struct mbuf **mpendp,
struct iovec **ivp)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 mbuf *m, *m2 = NULL, *m3;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 iovec *iv;
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
int i, left, siz;
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 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
left = 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
m3 = NULL;
/*
* Generate the mbuf list with the uio_iov ref. to it.
*/
i = 0;
while (left > 0) {
NFSMGET(m);
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
MCLGET(m, 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
m->m_len = 0;
siz = min(M_TRAILINGSPACE(m), left);
left -= siz;
i++;
if (m3)
m2->m_next = m;
else
m3 = m;
m2 = m;
}
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
*ivp = iv = malloc(i * sizeof (struct iovec), 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
m = m3;
left = len;
i = 0;
while (left > 0) {
if (m == NULL)
Add optional support for ext_pgs mbufs to the NFS server's read, readlink and getxattr operations. This patch optionally enables generation of read, readlink and getxattr replies in ext_pgs mbufs. Since neither of ND_EXTPG or ND_TLS are currently ever set, there is no change in semantics at this time. It also corrects the message in a couple of panic()s that should never occur. This is another in the series of commits that add support to the NFS client and server for building RPC messages in ext_pgs mbufs with anonymous pages. This is useful so that the entire mbuf list does not need to be copied before calling sosend() when NFS over TLS is enabled. Use of ext_pgs mbufs will not be enabled until the kernel RPC is updated to handle TLS.
2020-07-31 23:35:49 +00:00
panic("nfsrv_createiovec iov");
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
siz = min(M_TRAILINGSPACE(m), left);
if (siz > 0) {
iv->iov_base = mtod(m, caddr_t) + m->m_len;
iv->iov_len = siz;
m->m_len += siz;
left -= siz;
iv++;
i++;
}
m = m->m_next;
}
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
*mpp = m3;
*mpendp = m2;
return (i);
}
Add optional support for ext_pgs mbufs to the NFS server's read, readlink and getxattr operations. This patch optionally enables generation of read, readlink and getxattr replies in ext_pgs mbufs. Since neither of ND_EXTPG or ND_TLS are currently ever set, there is no change in semantics at this time. It also corrects the message in a couple of panic()s that should never occur. This is another in the series of commits that add support to the NFS client and server for building RPC messages in ext_pgs mbufs with anonymous pages. This is useful so that the entire mbuf list does not need to be copied before calling sosend() when NFS over TLS is enabled. Use of ext_pgs mbufs will not be enabled until the kernel RPC is updated to handle TLS.
2020-07-31 23:35:49 +00:00
/*
* Create an mbuf chain and an associated iovec that can be used to Read
* or Getextattr of data.
* Upon success, return pointers to the first and last mbufs in the chain
* plus the malloc'd iovec and its iovlen.
* Same as above, but creates ext_pgs mbuf(s).
*/
static int
nfsrv_createiovec_extpgs(int len, int maxextsiz, struct mbuf **mpp,
struct mbuf **mpendp, struct iovec **ivp)
{
struct mbuf *m, *m2 = NULL, *m3;
struct iovec *iv;
int i, left, pgno, siz;
left = len;
m3 = NULL;
/*
* Generate the mbuf list with the uio_iov ref. to it.
*/
i = 0;
while (left > 0) {
siz = min(left, maxextsiz);
m = mb_alloc_ext_plus_pages(siz, M_WAITOK);
left -= siz;
i += m->m_epg_npgs;
if (m3 != NULL)
m2->m_next = m;
else
m3 = m;
m2 = m;
}
*ivp = iv = malloc(i * sizeof (struct iovec), M_TEMP, M_WAITOK);
m = m3;
left = len;
i = 0;
pgno = 0;
while (left > 0) {
if (m == NULL)
panic("nfsvno_createiovec_extpgs iov");
siz = min(PAGE_SIZE, left);
if (siz > 0) {
iv->iov_base = (void *)PHYS_TO_DMAP(m->m_epg_pa[pgno]);
iv->iov_len = siz;
m->m_len += siz;
if (pgno == m->m_epg_npgs - 1)
m->m_epg_last_len = siz;
left -= siz;
iv++;
i++;
pgno++;
}
if (pgno == m->m_epg_npgs && left > 0) {
m = m->m_next;
if (m == NULL)
panic("nfsvno_createiovec_extpgs iov");
pgno = 0;
}
}
*mpp = m3;
*mpendp = m2;
return (i);
}
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
/*
* Read vnode op call into mbuf list.
*/
int
nfsvno_read(struct vnode *vp, off_t off, int cnt, struct ucred *cred,
Add optional support for ext_pgs mbufs to the NFS server's read, readlink and getxattr operations. This patch optionally enables generation of read, readlink and getxattr replies in ext_pgs mbufs. Since neither of ND_EXTPG or ND_TLS are currently ever set, there is no change in semantics at this time. It also corrects the message in a couple of panic()s that should never occur. This is another in the series of commits that add support to the NFS client and server for building RPC messages in ext_pgs mbufs with anonymous pages. This is useful so that the entire mbuf list does not need to be copied before calling sosend() when NFS over TLS is enabled. Use of ext_pgs mbufs will not be enabled until the kernel RPC is updated to handle TLS.
2020-07-31 23:35:49 +00:00
int maxextsiz, struct thread *p, struct mbuf **mpp,
struct mbuf **mpendp)
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 mbuf *m;
struct iovec *iv;
int error = 0, len, tlen, ioflag = 0;
struct mbuf *m3;
struct uio io, *uiop = &io;
struct nfsheur *nh;
/*
* Attempt to read from a DS file. A return of ENOENT implies
* there is no DS file to read.
*/
error = nfsrv_proxyds(vp, off, cnt, cred, p, NFSPROC_READDS, mpp,
NULL, mpendp, NULL, NULL, NULL, 0, NULL);
if (error != ENOENT)
return (error);
len = NFSM_RNDUP(cnt);
Add optional support for ext_pgs mbufs to the NFS server's read, readlink and getxattr operations. This patch optionally enables generation of read, readlink and getxattr replies in ext_pgs mbufs. Since neither of ND_EXTPG or ND_TLS are currently ever set, there is no change in semantics at this time. It also corrects the message in a couple of panic()s that should never occur. This is another in the series of commits that add support to the NFS client and server for building RPC messages in ext_pgs mbufs with anonymous pages. This is useful so that the entire mbuf list does not need to be copied before calling sosend() when NFS over TLS is enabled. Use of ext_pgs mbufs will not be enabled until the kernel RPC is updated to handle TLS.
2020-07-31 23:35:49 +00:00
if (maxextsiz > 0)
uiop->uio_iovcnt = nfsrv_createiovec_extpgs(len, maxextsiz,
&m3, &m, &iv);
else
uiop->uio_iovcnt = nfsrv_createiovec(len, &m3, &m, &iv);
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
uiop->uio_iov = iv;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
uiop->uio_offset = off;
uiop->uio_resid = len;
uiop->uio_rw = UIO_READ;
uiop->uio_segflg = UIO_SYSSPACE;
The PR reported that the old NFS server did not set uio_td == NULL for the VOP_READ() call. This patch fixes both the old and new server for this case. PR: 185232 Submitted by: PR had patch for old server Reviewed by: kib MFC after: 2 weeks
2014-04-24 20:47:58 +00:00
uiop->uio_td = NULL;
Enhance the sequential access heuristic used to perform readahead in the NFS server and reuse it for writes as well to allow writes to the backing store to be clustered. - Use a prime number for the size of the heuristic table (1017 is not prime). - Move the logic to locate a heuristic entry from the table and compute the sequential count out of VOP_READ() and into a separate routine. - Use the logic from sequential_heuristic() in vfs_vnops.c to update the seqcount when a sequential access is performed rather than just increasing seqcount by 1. This lets the clustering count ramp up faster. - Allow for some reordering of RPCs and if it is detected leave the current seqcount as-is rather than dropping back to a seqcount of 1. Also, when out of order access is encountered, cut seqcount in half rather than dropping it all the way back to 1 to further aid with reordering. - Fix the new NFS server to properly update the next offset after a successful VOP_READ() so that the readahead actually works. Some of these changes came from an earlier patch by Bjorn Gronwall that was forwarded to me by bde@. Discussed with: bde, rmacklem, fs@ Submitted by: Bjorn Gronwall (1, 4) MFC after: 2 weeks
2011-12-01 18:46:28 +00:00
nh = nfsrv_sequential_heuristic(uiop, vp);
ioflag |= nh->nh_seqcount << IO_SEQSHIFT;
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
/* XXX KDM make this more systematic? */
nfsstatsv1.srvbytes[NFSV4OP_READ] += uiop->uio_resid;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 = VOP_READ(vp, uiop, IO_NODELOCKED | ioflag, cred);
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
free(iv, 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) {
m_freem(m3);
*mpp = 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
}
Enhance the sequential access heuristic used to perform readahead in the NFS server and reuse it for writes as well to allow writes to the backing store to be clustered. - Use a prime number for the size of the heuristic table (1017 is not prime). - Move the logic to locate a heuristic entry from the table and compute the sequential count out of VOP_READ() and into a separate routine. - Use the logic from sequential_heuristic() in vfs_vnops.c to update the seqcount when a sequential access is performed rather than just increasing seqcount by 1. This lets the clustering count ramp up faster. - Allow for some reordering of RPCs and if it is detected leave the current seqcount as-is rather than dropping back to a seqcount of 1. Also, when out of order access is encountered, cut seqcount in half rather than dropping it all the way back to 1 to further aid with reordering. - Fix the new NFS server to properly update the next offset after a successful VOP_READ() so that the readahead actually works. Some of these changes came from an earlier patch by Bjorn Gronwall that was forwarded to me by bde@. Discussed with: bde, rmacklem, fs@ Submitted by: Bjorn Gronwall (1, 4) MFC after: 2 weeks
2011-12-01 18:46:28 +00:00
nh->nh_nextoff = uiop->uio_offset;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
tlen = len - uiop->uio_resid;
cnt = cnt < tlen ? cnt : tlen;
tlen = NFSM_RNDUP(cnt);
if (tlen == 0) {
m_freem(m3);
Add support for ext_pgs mbufs to nfsrv_adj(). This patch uses a slightly different algorithm for nfsrv_adj() since ext_pgs mbuf lists are not permitted to have m_len == 0 mbufs. As such, the code now frees mbufs after the adjustment in the list instead of setting their m_len field to 0. Since mbuf(s) may be trimmed off the tail of the list, the function now returns a pointer to the last mbuf in the list. This saves the caller from needing to use m_last() to find the last mbuf. It also implies that it might return a nul list, which required a check for that in nfsrvd_readlink(). This is another in the series of commits that add support to the NFS client and server for building RPC messages in ext_pgs mbufs with anonymous pages. This is useful so that the entire mbuf list does not need to be copied before calling sosend() when NFS over TLS is enabled. Use of ext_pgs mbufs will not be enabled until the kernel RPC is updated to handle TLS.
2020-07-26 02:42:09 +00:00
m3 = m = 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
} else if (len != tlen || tlen != cnt)
Add support for ext_pgs mbufs to nfsrv_adj(). This patch uses a slightly different algorithm for nfsrv_adj() since ext_pgs mbuf lists are not permitted to have m_len == 0 mbufs. As such, the code now frees mbufs after the adjustment in the list instead of setting their m_len field to 0. Since mbuf(s) may be trimmed off the tail of the list, the function now returns a pointer to the last mbuf in the list. This saves the caller from needing to use m_last() to find the last mbuf. It also implies that it might return a nul list, which required a check for that in nfsrvd_readlink(). This is another in the series of commits that add support to the NFS client and server for building RPC messages in ext_pgs mbufs with anonymous pages. This is useful so that the entire mbuf list does not need to be copied before calling sosend() when NFS over TLS is enabled. Use of ext_pgs mbufs will not be enabled until the kernel RPC is updated to handle TLS.
2020-07-26 02:42:09 +00:00
m = nfsrv_adj(m3, len - tlen, tlen - cnt);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
*mpp = m3;
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
*mpendp = m;
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
}
/*
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
* Create the iovec for the mbuf chain passed in as an argument.
* The "cp" argument is where the data starts within the first mbuf in
* the chain. It returns the iovec and the iovcnt.
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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
static int
nfsrv_createiovecw(int retlen, struct mbuf *m, char *cp, struct iovec **ivpp,
int *iovcntp)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 mbuf *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
struct iovec *ivp;
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
int cnt, i, 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
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 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
* Loop through the mbuf chain, counting how many mbufs are a
* part of this write operation, so the iovec size is known.
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 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
cnt = 0;
len = retlen;
mp = 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
i = mtod(mp, caddr_t) + mp->m_len - cp;
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
while (len > 0) {
if (i > 0) {
len -= i;
cnt++;
}
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
mp = mp->m_next;
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 (!mp) {
if (len > 0)
return (EBADRPC);
} else
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
i = mp->m_len;
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 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
/* Now, create the iovec. */
mp = m;
*ivpp = ivp = malloc(cnt * sizeof (struct iovec), 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
M_WAITOK);
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
*iovcntp = cnt;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
i = mtod(mp, caddr_t) + mp->m_len - cp;
len = retlen;
while (len > 0) {
if (mp == NULL)
Add optional support for ext_pgs mbufs to the NFS server's read, readlink and getxattr operations. This patch optionally enables generation of read, readlink and getxattr replies in ext_pgs mbufs. Since neither of ND_EXTPG or ND_TLS are currently ever set, there is no change in semantics at this time. It also corrects the message in a couple of panic()s that should never occur. This is another in the series of commits that add support to the NFS client and server for building RPC messages in ext_pgs mbufs with anonymous pages. This is useful so that the entire mbuf list does not need to be copied before calling sosend() when NFS over TLS is enabled. Use of ext_pgs mbufs will not be enabled until the kernel RPC is updated to handle TLS.
2020-07-31 23:35:49 +00:00
panic("nfsrv_createiovecw");
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 (i > 0) {
i = min(i, len);
ivp->iov_base = cp;
ivp->iov_len = i;
ivp++;
len -= i;
}
mp = mp->m_next;
if (mp) {
i = mp->m_len;
cp = mtod(mp, caddr_t);
}
}
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
return (0);
}
/*
* Write vnode op from an mbuf list.
*/
int
nfsvno_write(struct vnode *vp, off_t off, int retlen, int *stable,
struct mbuf *mp, char *cp, struct ucred *cred, struct thread *p)
{
struct iovec *iv;
int cnt, ioflags, error;
struct uio io, *uiop = &io;
struct nfsheur *nh;
/*
* Attempt to write to a DS file. A return of ENOENT implies
* there is no DS file to write.
*/
error = nfsrv_proxyds(vp, off, retlen, cred, p, NFSPROC_WRITEDS,
&mp, cp, NULL, NULL, NULL, NULL, 0, NULL);
if (error != ENOENT) {
*stable = NFSWRITE_FILESYNC;
return (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 (*stable == NFSWRITE_UNSTABLE)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
ioflags = IO_NODELOCKED;
else
ioflags = (IO_SYNC | IO_NODELOCKED);
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_createiovecw(retlen, mp, cp, &iv, &cnt);
if (error != 0)
return (error);
uiop->uio_iov = iv;
uiop->uio_iovcnt = cnt;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
uiop->uio_resid = retlen;
uiop->uio_rw = UIO_WRITE;
uiop->uio_segflg = UIO_SYSSPACE;
NFSUIOPROC(uiop, p);
uiop->uio_offset = off;
Enhance the sequential access heuristic used to perform readahead in the NFS server and reuse it for writes as well to allow writes to the backing store to be clustered. - Use a prime number for the size of the heuristic table (1017 is not prime). - Move the logic to locate a heuristic entry from the table and compute the sequential count out of VOP_READ() and into a separate routine. - Use the logic from sequential_heuristic() in vfs_vnops.c to update the seqcount when a sequential access is performed rather than just increasing seqcount by 1. This lets the clustering count ramp up faster. - Allow for some reordering of RPCs and if it is detected leave the current seqcount as-is rather than dropping back to a seqcount of 1. Also, when out of order access is encountered, cut seqcount in half rather than dropping it all the way back to 1 to further aid with reordering. - Fix the new NFS server to properly update the next offset after a successful VOP_READ() so that the readahead actually works. Some of these changes came from an earlier patch by Bjorn Gronwall that was forwarded to me by bde@. Discussed with: bde, rmacklem, fs@ Submitted by: Bjorn Gronwall (1, 4) MFC after: 2 weeks
2011-12-01 18:46:28 +00:00
nh = nfsrv_sequential_heuristic(uiop, vp);
ioflags |= nh->nh_seqcount << IO_SEQSHIFT;
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
/* XXX KDM make this more systematic? */
nfsstatsv1.srvbytes[NFSV4OP_WRITE] += uiop->uio_resid;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 = VOP_WRITE(vp, uiop, ioflags, cred);
Enhance the sequential access heuristic used to perform readahead in the NFS server and reuse it for writes as well to allow writes to the backing store to be clustered. - Use a prime number for the size of the heuristic table (1017 is not prime). - Move the logic to locate a heuristic entry from the table and compute the sequential count out of VOP_READ() and into a separate routine. - Use the logic from sequential_heuristic() in vfs_vnops.c to update the seqcount when a sequential access is performed rather than just increasing seqcount by 1. This lets the clustering count ramp up faster. - Allow for some reordering of RPCs and if it is detected leave the current seqcount as-is rather than dropping back to a seqcount of 1. Also, when out of order access is encountered, cut seqcount in half rather than dropping it all the way back to 1 to further aid with reordering. - Fix the new NFS server to properly update the next offset after a successful VOP_READ() so that the readahead actually works. Some of these changes came from an earlier patch by Bjorn Gronwall that was forwarded to me by bde@. Discussed with: bde, rmacklem, fs@ Submitted by: Bjorn Gronwall (1, 4) MFC after: 2 weeks
2011-12-01 18:46:28 +00:00
if (error == 0)
nh->nh_nextoff = uiop->uio_offset;
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(iv, 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);
}
/*
* Common code for creating a regular file (plus special files for V2).
*/
int
nfsvno_createsub(struct nfsrv_descript *nd, struct nameidata *ndp,
struct vnode **vpp, struct nfsvattr *nvap, int *exclusive_flagp,
Don't pass td to nfsvno_createsub(). MFC after: 2 weeks Sponsored by: DARPA, AFRL
2019-03-04 14:30:53 +00:00
int32_t *cverf, NFSDEV_T rdev, 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
{
u_quad_t tempsize;
int error;
Don't pass td to nfsvno_createsub(). MFC after: 2 weeks Sponsored by: DARPA, AFRL
2019-03-04 14:30:53 +00:00
struct thread *p = curthread;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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;
if (!error && ndp->ni_vp == NULL) {
if (nvap->na_type == VREG || nvap->na_type == VSOCK) {
vrele(ndp->ni_startdir);
error = VOP_CREATE(ndp->ni_dvp,
&ndp->ni_vp, &ndp->ni_cnd, &nvap->na_vattr);
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
/* For a pNFS server, create the data file on a DS. */
if (error == 0 && nvap->na_type == VREG) {
/*
* Create a data file on a DS for a pNFS server.
* This function just returns if not
* running a pNFS DS or the creation fails.
*/
nfsrv_pnfscreate(ndp->ni_vp, &nvap->na_vattr,
nd->nd_cred, p);
}
nfsserver: use VOP_VPUT_PAIR(). Apply VOP_VPUT_PAIR() to the end of vnode operations after the VOP_MKNOD(), VOP_MKDIR(), VOP_LINK(), VOP_SYMLINK(), VOP_CREATE(). Reviewed by: chs, mckusick Tested by: pho MFC after: 2 weeks Sponsored by: The FreeBSD Foundation
2021-01-24 13:02:27 +00:00
VOP_VPUT_PAIR(ndp->ni_dvp, error == 0 ? &ndp->ni_vp :
NULL, false);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsvno_relpathbuf(ndp);
if (!error) {
if (*exclusive_flagp) {
*exclusive_flagp = 0;
NFSVNO_ATTRINIT(nvap);
Patch the experimental NFS server is a manner analagous to r197525, so that the creation verifier is handled correctly in va_atime for 64bit architectures. There were two problems. One was that the code incorrectly assumed that sizeof (struct timespec) == 8 and the other was that the tv_sec field needs to be assigned from a signed 32bit integer, so that sign extension occurs on 64bit architectures. This is required for correct operation when exporting ZFS volumes. Reviewed by: pjd MFC after: 2 weeks
2009-11-20 21:21:13 +00:00
nvap->na_atime.tv_sec = cverf[0];
nvap->na_atime.tv_nsec = cverf[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
error = VOP_SETATTR(ndp->ni_vp,
&nvap->na_vattr, nd->nd_cred);
If the VOP_SETATTR() call that saves the exclusive create verifier failed, the NFS server would leave the newly created vnode locked. This could result in a file system that would not unmount and processes wedged, waiting for the file to be unlocked. Since this VOP_SETATTR() never fails for most file systems, this bug doesn't normally manifest itself. I found it during testing of an exported GlusterFS file system, which can fail. This patch adds the vput() and changes the error to the correct NFS one. MFC after: 2 weeks
2016-04-12 20:23:09 +00:00
if (error != 0) {
vput(ndp->ni_vp);
ndp->ni_vp = NULL;
error = NFSERR_NOTSUPP;
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 V2 Only. nfsrvd_mknod() does this for V3.
* (This implies, just get out on an error.)
*/
} else if (nvap->na_type == VCHR || nvap->na_type == VBLK ||
nvap->na_type == VFIFO) {
if (nvap->na_type == VCHR && rdev == 0xffffffff)
nvap->na_type = VFIFO;
if (nvap->na_type != VFIFO &&
Remove unused argument to priv_check_cred. Patch mostly generated with cocinnelle: @@ expression E1,E2; @@ - priv_check_cred(E1,E2,0) + priv_check_cred(E1,E2) Sponsored by: The FreeBSD Foundation
2018-12-11 19:32:16 +00:00
(error = priv_check_cred(nd->nd_cred, PRIV_VFS_MKNOD_DEV))) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
vrele(ndp->ni_startdir);
nfsvno_relpathbuf(ndp);
vput(ndp->ni_dvp);
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
}
nvap->na_rdev = rdev;
error = VOP_MKNOD(ndp->ni_dvp, &ndp->ni_vp,
&ndp->ni_cnd, &nvap->na_vattr);
nfsserver: use VOP_VPUT_PAIR(). Apply VOP_VPUT_PAIR() to the end of vnode operations after the VOP_MKNOD(), VOP_MKDIR(), VOP_LINK(), VOP_SYMLINK(), VOP_CREATE(). Reviewed by: chs, mckusick Tested by: pho MFC after: 2 weeks Sponsored by: The FreeBSD Foundation
2021-01-24 13:02:27 +00:00
VOP_VPUT_PAIR(ndp->ni_dvp, error == 0 ? &ndp->ni_vp :
NULL, false);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsvno_relpathbuf(ndp);
Vrele ni_startdir in the experimental NFS server for the case of NFSv2 getting an error return from VOP_MKNOD(). Without this patch, the server file system remains busy after an NFSv2 VOP_MKNOD() fails. MFC after: 2 weeks
2011-04-11 20:54:30 +00:00
vrele(ndp->ni_startdir);
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
} else {
vrele(ndp->ni_startdir);
nfsvno_relpathbuf(ndp);
vput(ndp->ni_dvp);
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 = ENXIO;
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
}
*vpp = ndp->ni_vp;
} else {
/*
* Handle cases where error is already set and/or
* the file exists.
* 1 - clean up the lookup
* 2 - iff !error and na_size set, truncate it
*/
vrele(ndp->ni_startdir);
nfsvno_relpathbuf(ndp);
*vpp = ndp->ni_vp;
if (ndp->ni_dvp == *vpp)
vrele(ndp->ni_dvp);
else
vput(ndp->ni_dvp);
if (!error && nvap->na_size != VNOVAL) {
Modify the experimental server so that it uses VOP_ACCESSX(). This is necessary in order to enable NFSv4 ACL support. The argument to nfsvno_accchk() was changed to an accmode_t and the function nfsrv_aclaccess() was no longer needed and, therefore, deleted. Reviewed by: trasz MFC after: 2 weeks
2009-12-25 20:44:19 +00:00
error = nfsvno_accchk(*vpp, VWRITE,
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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_cred, exp, p, NFSACCCHK_NOOVERRIDE,
Modify the experimental server so that it uses VOP_ACCESSX(). This is necessary in order to enable NFSv4 ACL support. The argument to nfsvno_accchk() was changed to an accmode_t and the function nfsrv_aclaccess() was no longer needed and, therefore, deleted. Reviewed by: trasz MFC after: 2 weeks
2009-12-25 20:44:19 +00:00
NFSACCCHK_VPISLOCKED, 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 (!error) {
tempsize = nvap->na_size;
NFSVNO_ATTRINIT(nvap);
nvap->na_size = tempsize;
error = VOP_SETATTR(*vpp,
&nvap->na_vattr, nd->nd_cred);
}
}
if (error)
vput(*vpp);
}
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);
}
/*
* Do a mknod vnode op.
*/
int
nfsvno_mknod(struct nameidata *ndp, struct nfsvattr *nvap, struct ucred *cred,
struct thread *p)
{
int error = 0;
enum vtype vtyp;
vtyp = nvap->na_type;
/*
* Iff doesn't exist, create it.
*/
if (ndp->ni_vp) {
vrele(ndp->ni_startdir);
nfsvno_relpathbuf(ndp);
vput(ndp->ni_dvp);
vrele(ndp->ni_vp);
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 = EEXIST;
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 (vtyp != VCHR && vtyp != VBLK && vtyp != VSOCK && vtyp != VFIFO) {
vrele(ndp->ni_startdir);
nfsvno_relpathbuf(ndp);
vput(ndp->ni_dvp);
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_BADTYPE;
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 (vtyp == VSOCK) {
vrele(ndp->ni_startdir);
error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp,
&ndp->ni_cnd, &nvap->na_vattr);
nfsserver: use VOP_VPUT_PAIR(). Apply VOP_VPUT_PAIR() to the end of vnode operations after the VOP_MKNOD(), VOP_MKDIR(), VOP_LINK(), VOP_SYMLINK(), VOP_CREATE(). Reviewed by: chs, mckusick Tested by: pho MFC after: 2 weeks Sponsored by: The FreeBSD Foundation
2021-01-24 13:02:27 +00:00
VOP_VPUT_PAIR(ndp->ni_dvp, error == 0 ? &ndp->ni_vp : NULL,
false);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsvno_relpathbuf(ndp);
} else {
if (nvap->na_type != VFIFO &&
Remove unused argument to priv_check_cred. Patch mostly generated with cocinnelle: @@ expression E1,E2; @@ - priv_check_cred(E1,E2,0) + priv_check_cred(E1,E2) Sponsored by: The FreeBSD Foundation
2018-12-11 19:32:16 +00:00
(error = priv_check_cred(cred, PRIV_VFS_MKNOD_DEV))) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
vrele(ndp->ni_startdir);
nfsvno_relpathbuf(ndp);
vput(ndp->ni_dvp);
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
}
error = VOP_MKNOD(ndp->ni_dvp, &ndp->ni_vp,
&ndp->ni_cnd, &nvap->na_vattr);
nfsserver: use VOP_VPUT_PAIR(). Apply VOP_VPUT_PAIR() to the end of vnode operations after the VOP_MKNOD(), VOP_MKDIR(), VOP_LINK(), VOP_SYMLINK(), VOP_CREATE(). Reviewed by: chs, mckusick Tested by: pho MFC after: 2 weeks Sponsored by: The FreeBSD Foundation
2021-01-24 13:02:27 +00:00
VOP_VPUT_PAIR(ndp->ni_dvp, error == 0 ? &ndp->ni_vp : NULL,
false);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsvno_relpathbuf(ndp);
Fix the experimental NFS server so that it doesn't leak a reference count on the directory when creating device special files. MFC after: 2 weeks
2011-01-03 00:40:13 +00:00
vrele(ndp->ni_startdir);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 VOP_MKNOD returns the ni_vp, I can't
* see any reason to do the lookup.
*/
}
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);
}
/*
* Mkdir vnode op.
*/
int
nfsvno_mkdir(struct nameidata *ndp, struct nfsvattr *nvap, uid_t saved_uid,
struct ucred *cred, struct thread *p, struct nfsexstuff *exp)
{
int error = 0;
if (ndp->ni_vp != NULL) {
if (ndp->ni_dvp == ndp->ni_vp)
vrele(ndp->ni_dvp);
else
vput(ndp->ni_dvp);
vrele(ndp->ni_vp);
For the experimental NFS server, add a call to free the lookup path buffer for one case where it was missing when doing mkdir. This could have conceivably resulted in a leak of a buffer, but a leak was never observed during testing, so I suspect it would have occurred rarely, if ever, in practice. MFC after: 2 weeks
2010-04-02 02:19:28 +00:00
nfsvno_relpathbuf(ndp);
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 = EEXIST;
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
}
error = VOP_MKDIR(ndp->ni_dvp, &ndp->ni_vp, &ndp->ni_cnd,
&nvap->na_vattr);
nfsserver: use VOP_VPUT_PAIR(). Apply VOP_VPUT_PAIR() to the end of vnode operations after the VOP_MKNOD(), VOP_MKDIR(), VOP_LINK(), VOP_SYMLINK(), VOP_CREATE(). Reviewed by: chs, mckusick Tested by: pho MFC after: 2 weeks Sponsored by: The FreeBSD Foundation
2021-01-24 13:02:27 +00:00
VOP_VPUT_PAIR(ndp->ni_dvp, error == 0 ? &ndp->ni_vp : NULL, false);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsvno_relpathbuf(ndp);
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);
}
/*
* symlink vnode op.
*/
int
nfsvno_symlink(struct nameidata *ndp, struct nfsvattr *nvap, char *pathcp,
int pathlen, int not_v2, uid_t saved_uid, struct ucred *cred, struct thread *p,
struct nfsexstuff *exp)
{
int error = 0;
if (ndp->ni_vp) {
vrele(ndp->ni_startdir);
nfsvno_relpathbuf(ndp);
if (ndp->ni_dvp == ndp->ni_vp)
vrele(ndp->ni_dvp);
else
vput(ndp->ni_dvp);
vrele(ndp->ni_vp);
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 = EEXIST;
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
}
error = VOP_SYMLINK(ndp->ni_dvp, &ndp->ni_vp, &ndp->ni_cnd,
&nvap->na_vattr, pathcp);
/*
* Although FreeBSD still had the lookup code in
* it for 7/current, there doesn't seem to be any
* point, since VOP_SYMLINK() returns the ni_vp.
* Just vput it for v2.
*/
nfsserver: use VOP_VPUT_PAIR(). Apply VOP_VPUT_PAIR() to the end of vnode operations after the VOP_MKNOD(), VOP_MKDIR(), VOP_LINK(), VOP_SYMLINK(), VOP_CREATE(). Reviewed by: chs, mckusick Tested by: pho MFC after: 2 weeks Sponsored by: The FreeBSD Foundation
2021-01-24 13:02:27 +00:00
VOP_VPUT_PAIR(ndp->ni_dvp, &ndp->ni_vp, !not_v2 && error == 0);
vrele(ndp->ni_startdir);
nfsvno_relpathbuf(ndp);
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);
}
/*
* Parse symbolic link arguments.
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
* This function has an ugly side effect. It will malloc() an area 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
* the symlink and set iov_base to point to it, only if it succeeds.
* So, if it returns with uiop->uio_iov->iov_base != NULL, that must
* be FREE'd later.
*/
int
nfsvno_getsymlink(struct nfsrv_descript *nd, struct nfsvattr *nvap,
struct thread *p, char **pathcpp, int *lenp)
{
u_int32_t *tl;
char *pathcp = NULL;
int error = 0, len;
struct nfsv2_sattr *sp;
*pathcpp = NULL;
*lenp = 0;
if ((nd->nd_flag & ND_NFSV3) &&
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
(error = nfsrv_sattr(nd, NULL, nvap, NULL, 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
goto nfsmout;
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
len = fxdr_unsigned(int, *tl);
if (len > NFS_MAXPATHLEN || len <= 0) {
error = EBADRPC;
goto nfsmout;
}
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
pathcp = malloc(len + 1, 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
error = nfsrv_mtostr(nd, pathcp, len);
if (error)
goto nfsmout;
if (nd->nd_flag & ND_NFSV2) {
NFSM_DISSECT(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
nvap->na_mode = fxdr_unsigned(u_int16_t, sp->sa_mode);
}
*pathcpp = pathcp;
*lenp = len;
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(0, 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 (0);
nfsmout:
if (pathcp)
free(pathcp, 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);
}
/*
* Remove a non-directory object.
*/
int
nfsvno_removesub(struct nameidata *ndp, int is_v4, struct ucred *cred,
struct thread *p, struct nfsexstuff *exp)
{
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 vnode *vp, *dsdvp[NFSDEV_MAXMIRRORS];
int error = 0, mirrorcnt;
char fname[PNFS_FILENAME_LEN + 1];
fhandle_t fh;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
vp = ndp->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
dsdvp[0] = 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 (vp->v_type == VDIR)
error = NFSERR_ISDIR;
else if (is_v4)
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, 1, NULL, (nfsquad_t)((u_quad_t)0),
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 (error == 0)
nfsrv_pnfsremovesetup(vp, p, dsdvp, &mirrorcnt, fname, &fh);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 = VOP_REMOVE(ndp->ni_dvp, vp, &ndp->ni_cnd);
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 && dsdvp[0] != NULL)
nfsrv_pnfsremove(dsdvp, mirrorcnt, fname, &fh, 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 (ndp->ni_dvp == vp)
vrele(ndp->ni_dvp);
else
vput(ndp->ni_dvp);
vput(vp);
Fix a leak of namei lookup path buffers that occurs when a ZFS volume is exported via the new NFS server. The leak occurred because the new NFS server code didn't handle the case where a file system sets the SAVENAME flag in its VOP_LOOKUP() and ZFS does this for the DELETE case. Tested by: Oliver Brandmueller (ob at gruft.de), hrs PR: kern/167266 MFC after: 1 month
2012-04-27 20:23:24 +00:00
if ((ndp->ni_cnd.cn_flags & SAVENAME) != 0)
nfsvno_relpathbuf(ndp);
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);
}
/*
* Remove a directory.
*/
int
nfsvno_rmdirsub(struct nameidata *ndp, int is_v4, struct ucred *cred,
struct thread *p, struct nfsexstuff *exp)
{
struct vnode *vp;
int error = 0;
vp = ndp->ni_vp;
if (vp->v_type != VDIR) {
error = ENOTDIR;
goto out;
}
/*
* No rmdir "." please.
*/
if (ndp->ni_dvp == vp) {
error = EINVAL;
goto out;
}
/*
* The root of a mounted filesystem cannot be deleted.
*/
if (vp->v_vflag & VV_ROOT)
error = EBUSY;
out:
if (!error)
error = VOP_RMDIR(ndp->ni_dvp, vp, &ndp->ni_cnd);
if (ndp->ni_dvp == vp)
vrele(ndp->ni_dvp);
else
vput(ndp->ni_dvp);
vput(vp);
Fix a leak of namei lookup path buffers that occurs when a ZFS volume is exported via the new NFS server. The leak occurred because the new NFS server code didn't handle the case where a file system sets the SAVENAME flag in its VOP_LOOKUP() and ZFS does this for the DELETE case. Tested by: Oliver Brandmueller (ob at gruft.de), hrs PR: kern/167266 MFC after: 1 month
2012-04-27 20:23:24 +00:00
if ((ndp->ni_cnd.cn_flags & SAVENAME) != 0)
nfsvno_relpathbuf(ndp);
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);
}
/*
* Rename vnode op.
*/
int
nfsvno_rename(struct nameidata *fromndp, struct nameidata *tondp,
u_int32_t ndstat, u_int32_t ndflag, struct ucred *cred, struct thread *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 vnode *fvp, *tvp, *tdvp, *dsdvp[NFSDEV_MAXMIRRORS];
int error = 0, mirrorcnt;
char fname[PNFS_FILENAME_LEN + 1];
fhandle_t fh;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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
dsdvp[0] = 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
fvp = fromndp->ni_vp;
if (ndstat) {
vrele(fromndp->ni_dvp);
vrele(fvp);
error = ndstat;
goto out1;
}
tdvp = tondp->ni_dvp;
tvp = tondp->ni_vp;
if (tvp != NULL) {
if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
error = (ndflag & ND_NFSV2) ? EISDIR : EEXIST;
goto out;
} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
error = (ndflag & ND_NFSV2) ? ENOTDIR : EEXIST;
goto out;
}
if (tvp->v_type == VDIR && tvp->v_mountedhere) {
error = (ndflag & ND_NFSV2) ? ENOTEMPTY : EXDEV;
goto out;
}
/*
* A rename to '.' or '..' results in a prematurely
* unlocked vnode on FreeBSD5, so I'm just going to fail that
* here.
*/
if ((tondp->ni_cnd.cn_namelen == 1 &&
tondp->ni_cnd.cn_nameptr[0] == '.') ||
(tondp->ni_cnd.cn_namelen == 2 &&
tondp->ni_cnd.cn_nameptr[0] == '.' &&
tondp->ni_cnd.cn_nameptr[1] == '.')) {
error = EINVAL;
goto out;
}
}
if (fvp->v_type == VDIR && fvp->v_mountedhere) {
error = (ndflag & ND_NFSV2) ? ENOTEMPTY : EXDEV;
goto out;
}
if (fvp->v_mount != tdvp->v_mount) {
error = (ndflag & ND_NFSV2) ? ENOTEMPTY : EXDEV;
goto out;
}
if (fvp == tdvp) {
error = (ndflag & ND_NFSV2) ? ENOTEMPTY : EINVAL;
goto out;
}
if (fvp == tvp) {
/*
* If source and destination are the same, there is nothing to
* do. Set error to -1 to indicate this.
*/
error = -1;
goto out;
}
if (ndflag & ND_NFSV4) {
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(fvp, 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(fvp, 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(fvp);
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;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 (tvp && !error)
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(tvp, 1, NULL,
(nfsquad_t)((u_quad_t)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 {
/*
* For NFSv2 and NFSv3, try to get rid of the delegation, so
* that the NFSv4 client won't be confused by the rename.
* Since nfsd_recalldelegation() can only be called on an
* unlocked vnode at this point and fvp is the file that will
* still exist after the rename, just do fvp.
*/
nfsd_recalldelegation(fvp, 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 (error == 0 && tvp != NULL) {
nfsrv_pnfsremovesetup(tvp, p, dsdvp, &mirrorcnt, fname, &fh);
NFSD_DEBUG(4, "nfsvno_rename: pnfsremovesetup"
" dsdvp=%p\n", dsdvp[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
out:
if (!error) {
error = VOP_RENAME(fromndp->ni_dvp, fromndp->ni_vp,
&fromndp->ni_cnd, tondp->ni_dvp, tondp->ni_vp,
&tondp->ni_cnd);
} else {
if (tdvp == tvp)
vrele(tdvp);
else
vput(tdvp);
if (tvp)
vput(tvp);
vrele(fromndp->ni_dvp);
vrele(fvp);
if (error == -1)
error = 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
/*
* If dsdvp[0] != NULL, it was set up by nfsrv_pnfsremovesetup() and
* if the rename succeeded, the DS file for the tvp needs to be
* removed.
*/
if (error == 0 && dsdvp[0] != NULL) {
nfsrv_pnfsremove(dsdvp, mirrorcnt, fname, &fh, p);
NFSD_DEBUG(4, "nfsvno_rename: pnfsremove\n");
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
vrele(tondp->ni_startdir);
nfsvno_relpathbuf(tondp);
out1:
vrele(fromndp->ni_startdir);
nfsvno_relpathbuf(fromndp);
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);
}
/*
* Link vnode op.
*/
int
nfsvno_link(struct nameidata *ndp, struct vnode *vp, struct ucred *cred,
struct thread *p, struct nfsexstuff *exp)
{
struct vnode *xp;
int error = 0;
xp = ndp->ni_vp;
if (xp != NULL) {
error = EEXIST;
} else {
xp = ndp->ni_dvp;
if (vp->v_mount != xp->v_mount)
error = EXDEV;
}
if (!error) {
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
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 (!VN_IS_DOOMED(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
error = VOP_LINK(ndp->ni_dvp, vp, &ndp->ni_cnd);
else
error = EPERM;
nfsserver: use VOP_VPUT_PAIR(). Apply VOP_VPUT_PAIR() to the end of vnode operations after the VOP_MKNOD(), VOP_MKDIR(), VOP_LINK(), VOP_SYMLINK(), VOP_CREATE(). Reviewed by: chs, mckusick Tested by: pho MFC after: 2 weeks Sponsored by: The FreeBSD Foundation
2021-01-24 13:02:27 +00:00
if (ndp->ni_dvp == 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
vrele(ndp->ni_dvp);
nfsserver: use VOP_VPUT_PAIR(). Apply VOP_VPUT_PAIR() to the end of vnode operations after the VOP_MKNOD(), VOP_MKDIR(), VOP_LINK(), VOP_SYMLINK(), VOP_CREATE(). Reviewed by: chs, mckusick Tested by: pho MFC after: 2 weeks Sponsored by: The FreeBSD Foundation
2021-01-24 13:02:27 +00:00
NFSVOPUNLOCK(vp);
} else {
vref(vp);
VOP_VPUT_PAIR(ndp->ni_dvp, &vp, 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
} else {
if (ndp->ni_dvp == ndp->ni_vp)
vrele(ndp->ni_dvp);
else
vput(ndp->ni_dvp);
if (ndp->ni_vp)
vrele(ndp->ni_vp);
}
nfsvno_relpathbuf(ndp);
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);
}
/*
* Do the fsync() appropriate for the commit.
*/
int
nfsvno_fsync(struct vnode *vp, u_int64_t off, int cnt, struct ucred *cred,
struct thread *td)
{
int error = 0;
Patch the new NFS server in a manner analagous to r228520 for the old NFS server, so that it correctly handles a count == 0 argument for Commit. PR: kern/118126 MFC after: 2 weeks
2011-12-16 00:58:41 +00:00
/*
* RFC 1813 3.3.21: if count is 0, a flush from offset to the end of
* file is done. At this time VOP_FSYNC does not accept offset and
* byte count parameters so call VOP_FSYNC the whole file for now.
* The same is true for NFSv4: RFC 3530 Sec. 14.2.3.
File systems that do not use the buffer cache (such as ZFS) must use VOP_FSYNC() to perform the NFS server's Commit operation. This patch adds a mnt_kern_flag called MNTK_USES_BCACHE which is set by file systems that use the buffer cache. If this flag is not set, the NFS server always does a VOP_FSYNC(). This should be ok for old file system modules that do not set MNTK_USES_BCACHE, since calling VOP_FSYNC() is correct, although it might not be optimal for file systems that use the buffer cache. Reviewed by: kib MFC after: 2 weeks
2015-04-15 20:16:31 +00:00
* File systems that do not use the buffer cache (as indicated
* by MNTK_USES_BCACHE not being set) must use VOP_FSYNC().
Patch the new NFS server in a manner analagous to r228520 for the old NFS server, so that it correctly handles a count == 0 argument for Commit. PR: kern/118126 MFC after: 2 weeks
2011-12-16 00:58:41 +00:00
*/
File systems that do not use the buffer cache (such as ZFS) must use VOP_FSYNC() to perform the NFS server's Commit operation. This patch adds a mnt_kern_flag called MNTK_USES_BCACHE which is set by file systems that use the buffer cache. If this flag is not set, the NFS server always does a VOP_FSYNC(). This should be ok for old file system modules that do not set MNTK_USES_BCACHE, since calling VOP_FSYNC() is correct, although it might not be optimal for file systems that use the buffer cache. Reviewed by: kib MFC after: 2 weeks
2015-04-15 20:16:31 +00:00
if (cnt == 0 || cnt > MAX_COMMIT_COUNT ||
(vp->v_mount->mnt_kern_flag & MNTK_USES_BCACHE) == 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
/*
* Give up and do the whole thing
*/
Replace OBJ_MIGHTBEDIRTY with a system using atomics. Remove the TMPFS_DIRTY flag and use the same system. This enables further fault locking improvements by allowing more faults to proceed with a shared lock. Reviewed by: kib Tested by: pho Differential Revision: https://reviews.freebsd.org/D22116
2019-10-29 21:06:34 +00:00
if (vp->v_object && vm_object_mightbedirty(vp->v_object)) {
Switch the vm_object mutex to be a rwlock. This will enable in the future further optimizations where the vm_object lock will be held in read mode most of the time the page cache resident pool of pages are accessed for reading purposes. The change is mostly mechanical but few notes are reported: * The KPI changes as follow: - VM_OBJECT_LOCK() -> VM_OBJECT_WLOCK() - VM_OBJECT_TRYLOCK() -> VM_OBJECT_TRYWLOCK() - VM_OBJECT_UNLOCK() -> VM_OBJECT_WUNLOCK() - VM_OBJECT_LOCK_ASSERT(MA_OWNED) -> VM_OBJECT_ASSERT_WLOCKED() (in order to avoid visibility of implementation details) - The read-mode operations are added: VM_OBJECT_RLOCK(), VM_OBJECT_TRYRLOCK(), VM_OBJECT_RUNLOCK(), VM_OBJECT_ASSERT_RLOCKED(), VM_OBJECT_ASSERT_LOCKED() * The vm/vm_pager.h namespace pollution avoidance (forcing requiring sys/mutex.h in consumers directly to cater its inlining functions using VM_OBJECT_LOCK()) imposes that all the vm/vm_pager.h consumers now must include also sys/rwlock.h. * zfs requires a quite convoluted fix to include FreeBSD rwlocks into the compat layer because the name clash between FreeBSD and solaris versions must be avoided. At this purpose zfs redefines the vm_object locking functions directly, isolating the FreeBSD components in specific compat stubs. The KPI results heavilly broken by this commit. Thirdy part ports must be updated accordingly (I can think off-hand of VirtualBox, for example). Sponsored by: EMC / Isilon storage division Reviewed by: jeff Reviewed by: pjd (ZFS specific review) Discussed with: alc Tested by: pho
2013-03-09 02:32:23 +00:00
VM_OBJECT_WLOCK(vp->v_object);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
vm_object_page_clean(vp->v_object, 0, 0, OBJPC_SYNC);
Switch the vm_object mutex to be a rwlock. This will enable in the future further optimizations where the vm_object lock will be held in read mode most of the time the page cache resident pool of pages are accessed for reading purposes. The change is mostly mechanical but few notes are reported: * The KPI changes as follow: - VM_OBJECT_LOCK() -> VM_OBJECT_WLOCK() - VM_OBJECT_TRYLOCK() -> VM_OBJECT_TRYWLOCK() - VM_OBJECT_UNLOCK() -> VM_OBJECT_WUNLOCK() - VM_OBJECT_LOCK_ASSERT(MA_OWNED) -> VM_OBJECT_ASSERT_WLOCKED() (in order to avoid visibility of implementation details) - The read-mode operations are added: VM_OBJECT_RLOCK(), VM_OBJECT_TRYRLOCK(), VM_OBJECT_RUNLOCK(), VM_OBJECT_ASSERT_RLOCKED(), VM_OBJECT_ASSERT_LOCKED() * The vm/vm_pager.h namespace pollution avoidance (forcing requiring sys/mutex.h in consumers directly to cater its inlining functions using VM_OBJECT_LOCK()) imposes that all the vm/vm_pager.h consumers now must include also sys/rwlock.h. * zfs requires a quite convoluted fix to include FreeBSD rwlocks into the compat layer because the name clash between FreeBSD and solaris versions must be avoided. At this purpose zfs redefines the vm_object locking functions directly, isolating the FreeBSD components in specific compat stubs. The KPI results heavilly broken by this commit. Thirdy part ports must be updated accordingly (I can think off-hand of VirtualBox, for example). Sponsored by: EMC / Isilon storage division Reviewed by: jeff Reviewed by: pjd (ZFS specific review) Discussed with: alc Tested by: pho
2013-03-09 02:32:23 +00:00
VM_OBJECT_WUNLOCK(vp->v_object);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 = VOP_FSYNC(vp, MNT_WAIT, td);
} else {
/*
* Locate and synchronously write any buffers that fall
* into the requested range. Note: we are assuming that
* f_iosize is a power of 2.
*/
int iosize = vp->v_mount->mnt_stat.f_iosize;
int iomask = iosize - 1;
struct bufobj *bo;
daddr_t lblkno;
/*
NFS: spelling fixes on comments. No funcional change.
2016-04-29 16:07:25 +00:00
* Align to iosize boundary, super-align to page boundary.
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 (off & iomask) {
cnt += off & iomask;
off &= ~(u_quad_t)iomask;
}
if (off & PAGE_MASK) {
cnt += off & PAGE_MASK;
off &= ~(u_quad_t)PAGE_MASK;
}
lblkno = off / iosize;
Replace OBJ_MIGHTBEDIRTY with a system using atomics. Remove the TMPFS_DIRTY flag and use the same system. This enables further fault locking improvements by allowing more faults to proceed with a shared lock. Reviewed by: kib Tested by: pho Differential Revision: https://reviews.freebsd.org/D22116
2019-10-29 21:06:34 +00:00
if (vp->v_object && vm_object_mightbedirty(vp->v_object)) {
Switch the vm_object mutex to be a rwlock. This will enable in the future further optimizations where the vm_object lock will be held in read mode most of the time the page cache resident pool of pages are accessed for reading purposes. The change is mostly mechanical but few notes are reported: * The KPI changes as follow: - VM_OBJECT_LOCK() -> VM_OBJECT_WLOCK() - VM_OBJECT_TRYLOCK() -> VM_OBJECT_TRYWLOCK() - VM_OBJECT_UNLOCK() -> VM_OBJECT_WUNLOCK() - VM_OBJECT_LOCK_ASSERT(MA_OWNED) -> VM_OBJECT_ASSERT_WLOCKED() (in order to avoid visibility of implementation details) - The read-mode operations are added: VM_OBJECT_RLOCK(), VM_OBJECT_TRYRLOCK(), VM_OBJECT_RUNLOCK(), VM_OBJECT_ASSERT_RLOCKED(), VM_OBJECT_ASSERT_LOCKED() * The vm/vm_pager.h namespace pollution avoidance (forcing requiring sys/mutex.h in consumers directly to cater its inlining functions using VM_OBJECT_LOCK()) imposes that all the vm/vm_pager.h consumers now must include also sys/rwlock.h. * zfs requires a quite convoluted fix to include FreeBSD rwlocks into the compat layer because the name clash between FreeBSD and solaris versions must be avoided. At this purpose zfs redefines the vm_object locking functions directly, isolating the FreeBSD components in specific compat stubs. The KPI results heavilly broken by this commit. Thirdy part ports must be updated accordingly (I can think off-hand of VirtualBox, for example). Sponsored by: EMC / Isilon storage division Reviewed by: jeff Reviewed by: pjd (ZFS specific review) Discussed with: alc Tested by: pho
2013-03-09 02:32:23 +00:00
VM_OBJECT_WLOCK(vp->v_object);
Unless "cnt" exceeds MAX_COMMIT_COUNT, nfsrv_commit() and nfsvno_fsync() are incorrectly calling vm_object_page_clean(). They are passing the length of the range rather than the ending offset of the range. Perform the OFF_TO_IDX() conversion in vm_object_page_clean() rather than the callers. Reviewed by: kib MFC after: 3 weeks
2011-02-05 21:21:27 +00:00
vm_object_page_clean(vp->v_object, off, off + cnt,
OBJPC_SYNC);
Switch the vm_object mutex to be a rwlock. This will enable in the future further optimizations where the vm_object lock will be held in read mode most of the time the page cache resident pool of pages are accessed for reading purposes. The change is mostly mechanical but few notes are reported: * The KPI changes as follow: - VM_OBJECT_LOCK() -> VM_OBJECT_WLOCK() - VM_OBJECT_TRYLOCK() -> VM_OBJECT_TRYWLOCK() - VM_OBJECT_UNLOCK() -> VM_OBJECT_WUNLOCK() - VM_OBJECT_LOCK_ASSERT(MA_OWNED) -> VM_OBJECT_ASSERT_WLOCKED() (in order to avoid visibility of implementation details) - The read-mode operations are added: VM_OBJECT_RLOCK(), VM_OBJECT_TRYRLOCK(), VM_OBJECT_RUNLOCK(), VM_OBJECT_ASSERT_RLOCKED(), VM_OBJECT_ASSERT_LOCKED() * The vm/vm_pager.h namespace pollution avoidance (forcing requiring sys/mutex.h in consumers directly to cater its inlining functions using VM_OBJECT_LOCK()) imposes that all the vm/vm_pager.h consumers now must include also sys/rwlock.h. * zfs requires a quite convoluted fix to include FreeBSD rwlocks into the compat layer because the name clash between FreeBSD and solaris versions must be avoided. At this purpose zfs redefines the vm_object locking functions directly, isolating the FreeBSD components in specific compat stubs. The KPI results heavilly broken by this commit. Thirdy part ports must be updated accordingly (I can think off-hand of VirtualBox, for example). Sponsored by: EMC / Isilon storage division Reviewed by: jeff Reviewed by: pjd (ZFS specific review) Discussed with: alc Tested by: pho
2013-03-09 02:32:23 +00:00
VM_OBJECT_WUNLOCK(vp->v_object);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
bo = &vp->v_bufobj;
BO_LOCK(bo);
while (cnt > 0) {
struct buf *bp;
/*
* If we have a buffer and it is marked B_DELWRI we
* have to lock and write it. Otherwise the prior
* write is assumed to have already been committed.
*
* gbincore() can return invalid buffers now so we
* have to check that bit as well (though B_DELWRI
* should not be set if B_INVAL is set there could be
* a race here since we haven't locked the buffer).
*/
if ((bp = gbincore(&vp->v_bufobj, lblkno)) != NULL) {
if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_SLEEPFAIL |
- Convert the bufobj lock to rwlock. - Use a shared bufobj lock in getblk() and inmem(). - Convert softdep's lk to rwlock to match the bufobj lock. - Move INFREECNT to b_flags and protect it with the buf lock. - Remove unnecessary locking around bremfree() and BKGRDINPROG. Sponsored by: EMC / Isilon Storage Division Discussed with: mckusick, kib, mdf
2013-05-31 00:43:41 +00:00
LK_INTERLOCK, BO_LOCKPTR(bo)) == ENOLCK) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
BO_LOCK(bo);
continue; /* retry */
}
if ((bp->b_flags & (B_DELWRI|B_INVAL)) ==
B_DELWRI) {
bremfree(bp);
bp->b_flags &= ~B_ASYNC;
bwrite(bp);
++nfs_commit_miss;
} else
BUF_UNLOCK(bp);
BO_LOCK(bo);
}
++nfs_commit_blks;
if (cnt < iosize)
break;
cnt -= iosize;
++lblkno;
}
BO_UNLOCK(bo);
}
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);
}
/*
* Statfs vnode op.
*/
int
Remove the thread argument from the FSD (File-System Dependent) parts of the VFS. Now all the VFS_* functions and relating parts don't want the context as long as it always refers to curthread. In some points, in particular when dealing with VOPs and functions living in the same namespace (eg. vflush) which still need to be converted, pass curthread explicitly in order to retain the old behaviour. Such loose ends will be fixed ASAP. While here fix a bug: now, UFS_EXTATTR can be compiled alone without the UFS_EXTATTR_AUTOSTART option. VFS KPI is heavilly changed by this commit so thirdy parts modules needs to be recompiled. Bump __FreeBSD_version in order to signal such situation.
2009-05-11 15:33:26 +00:00
nfsvno_statfs(struct vnode *vp, struct statfs *sf)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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
struct statfs *tsf;
Change the new NFS server so that it returns 0 when the f_bavail or f_ffree fields of "struct statfs" are negative, since the values that go on the wire are unsigned and will appear to be very large positive values otherwise. This makes the handling of a negative f_bavail compatible with the old/regular NFS server. MFC after: 2 weeks
2011-05-06 01:29:14 +00:00
int 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
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
tsf = NULL;
if (nfsrv_devidcnt > 0) {
/* For a pNFS service, get the DS numbers. */
tsf = malloc(sizeof(*tsf), M_TEMP, M_WAITOK | M_ZERO);
Fix the pNFS server's reporting of disk space usage for the "#<path>" case. The pNFS server would report the total disk space used and free for all of the DSs, even when certain DSs are assigned to the file system via the "#<path>" suffix used in the "nfsd -p" option argument. This patch fixes this case. It only reports usage for the file system that the argument vnode resides on. This is consistent with the non-pNFS NFSv4 server. In NFSv4 it is possible to have subtrees on other file systems, but these are not included in the usage information for NFSv4. Approved by: re (gjb)
2018-10-09 01:10:50 +00:00
error = nfsrv_pnfsstatfs(tsf, vp->v_mount);
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(tsf, M_TEMP);
tsf = NULL;
}
}
Change the new NFS server so that it returns 0 when the f_bavail or f_ffree fields of "struct statfs" are negative, since the values that go on the wire are unsigned and will appear to be very large positive values otherwise. This makes the handling of a negative f_bavail compatible with the old/regular NFS server. MFC after: 2 weeks
2011-05-06 01:29:14 +00:00
error = VFS_STATFS(vp->v_mount, sf);
if (error == 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
if (tsf != NULL) {
sf->f_blocks = tsf->f_blocks;
sf->f_bavail = tsf->f_bavail;
sf->f_bfree = tsf->f_bfree;
sf->f_bsize = tsf->f_bsize;
}
Change the new NFS server so that it returns 0 when the f_bavail or f_ffree fields of "struct statfs" are negative, since the values that go on the wire are unsigned and will appear to be very large positive values otherwise. This makes the handling of a negative f_bavail compatible with the old/regular NFS server. MFC after: 2 weeks
2011-05-06 01:29:14 +00:00
/*
* Since NFS handles these values as unsigned on the
* wire, there is no way to represent negative values,
* so set them to 0. Without this, they will appear
* to be very large positive values for clients like
* Solaris10.
*/
if (sf->f_bavail < 0)
sf->f_bavail = 0;
if (sf->f_ffree < 0)
sf->f_ffree = 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
free(tsf, 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);
Change the new NFS server so that it returns 0 when the f_bavail or f_ffree fields of "struct statfs" are negative, since the values that go on the wire are unsigned and will appear to be very large positive values otherwise. This makes the handling of a negative f_bavail compatible with the old/regular NFS server. MFC after: 2 weeks
2011-05-06 01:29:14 +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
}
/*
* Do the vnode op stuff for Open. Similar to nfsvno_createsub(), but
* must handle nfsrv_opencheck() calls after any other access checks.
*/
void
nfsvno_open(struct nfsrv_descript *nd, struct nameidata *ndp,
nfsquad_t clientid, nfsv4stateid_t *stateidp, struct nfsstate *stp,
Patch the experimental NFS server is a manner analagous to r197525, so that the creation verifier is handled correctly in va_atime for 64bit architectures. There were two problems. One was that the code incorrectly assumed that sizeof (struct timespec) == 8 and the other was that the tv_sec field needs to be assigned from a signed 32bit integer, so that sign extension occurs on 64bit architectures. This is required for correct operation when exporting ZFS volumes. Reviewed by: pjd MFC after: 2 weeks
2009-11-20 21:21:13 +00:00
int *exclusive_flagp, struct nfsvattr *nvap, int32_t *cverf, int create,
Don't pass td to nfsvno_open(). MFC after: 2 weeks Sponsored by: DARPA, AFRL
2019-03-04 14:50:00 +00:00
NFSACL_T *aclp, nfsattrbit_t *attrbitp, struct ucred *cred,
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 nfsexstuff *exp, struct vnode **vpp)
{
struct vnode *vp = NULL;
u_quad_t tempsize;
struct nfsexstuff nes;
Don't pass td to nfsvno_open(). MFC after: 2 weeks Sponsored by: DARPA, AFRL
2019-03-04 14:50:00 +00:00
struct thread *p = curthread;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 (ndp->ni_vp == NULL)
nd->nd_repstat = nfsrv_opencheck(clientid,
stateidp, stp, NULL, nd, p, nd->nd_repstat);
if (!nd->nd_repstat) {
if (ndp->ni_vp == NULL) {
vrele(ndp->ni_startdir);
nd->nd_repstat = VOP_CREATE(ndp->ni_dvp,
&ndp->ni_vp, &ndp->ni_cnd, &nvap->na_vattr);
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
/* For a pNFS server, create the data file on a DS. */
if (nd->nd_repstat == 0) {
/*
* Create a data file on a DS for a pNFS server.
* This function just returns if not
* running a pNFS DS or the creation fails.
*/
nfsrv_pnfscreate(ndp->ni_vp, &nvap->na_vattr,
cred, p);
}
nfsserver: use VOP_VPUT_PAIR(). Apply VOP_VPUT_PAIR() to the end of vnode operations after the VOP_MKNOD(), VOP_MKDIR(), VOP_LINK(), VOP_SYMLINK(), VOP_CREATE(). Reviewed by: chs, mckusick Tested by: pho MFC after: 2 weeks Sponsored by: The FreeBSD Foundation
2021-01-24 13:02:27 +00:00
VOP_VPUT_PAIR(ndp->ni_dvp, nd->nd_repstat == 0 ?
&ndp->ni_vp : NULL, false);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsvno_relpathbuf(ndp);
if (!nd->nd_repstat) {
if (*exclusive_flagp) {
*exclusive_flagp = 0;
NFSVNO_ATTRINIT(nvap);
Patch the experimental NFS server is a manner analagous to r197525, so that the creation verifier is handled correctly in va_atime for 64bit architectures. There were two problems. One was that the code incorrectly assumed that sizeof (struct timespec) == 8 and the other was that the tv_sec field needs to be assigned from a signed 32bit integer, so that sign extension occurs on 64bit architectures. This is required for correct operation when exporting ZFS volumes. Reviewed by: pjd MFC after: 2 weeks
2009-11-20 21:21:13 +00:00
nvap->na_atime.tv_sec = cverf[0];
nvap->na_atime.tv_nsec = cverf[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
nd->nd_repstat = VOP_SETATTR(ndp->ni_vp,
&nvap->na_vattr, cred);
If the VOP_SETATTR() call that saves the exclusive create verifier failed, the NFS server would leave the newly created vnode locked. This could result in a file system that would not unmount and processes wedged, waiting for the file to be unlocked. Since this VOP_SETATTR() never fails for most file systems, this bug doesn't normally manifest itself. I found it during testing of an exported GlusterFS file system, which can fail. This patch adds the vput() and changes the error to the correct NFS one. MFC after: 2 weeks
2016-04-12 20:23:09 +00:00
if (nd->nd_repstat != 0) {
vput(ndp->ni_vp);
ndp->ni_vp = NULL;
nd->nd_repstat = NFSERR_NOTSUPP;
Fix the setting of atime for Linux client NFSv4 mounts. The FreeBSD NFSv4 server did not set the attribute bit for TimeAccess in the reply to an Open with exclusive_create, as required by the RFCs. (This is required since the FreeBSD NFS server stores the create_verifier in the va_atime attribute.) As such, the Linux NFSv4 client did not set the TimeAccess (atime) in the Setattr done in an RPC after the one with the Open/exclusive_create. This patch fixes the server to set the TimeAccess bit in the reply. I believe that storing the create_verifier in an extended attribute for file systems that support extended attributes might be a good idea, but I will wait for a discussion of this on the freebsd-fs@ email list before considering committing a patch to do this. Reported by: jim@ks.uiuc.edu Suggested by: dfr MFC after: 2 weeks
2017-04-21 00:17:47 +00:00
} else
NFSSETBIT_ATTRBIT(attrbitp,
NFSATTRBIT_TIMEACCESS);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 {
nfsrv_fixattr(nd, ndp->ni_vp, nvap,
aclp, p, attrbitp, exp);
}
}
vp = ndp->ni_vp;
} else {
if (ndp->ni_startdir)
vrele(ndp->ni_startdir);
nfsvno_relpathbuf(ndp);
vp = ndp->ni_vp;
if (create == NFSV4OPEN_CREATE) {
if (ndp->ni_dvp == vp)
vrele(ndp->ni_dvp);
else
vput(ndp->ni_dvp);
}
if (NFSVNO_ISSETSIZE(nvap) && vp->v_type == VREG) {
if (ndp->ni_cnd.cn_flags & RDONLY)
NFSVNO_SETEXRDONLY(&nes);
else
NFSVNO_EXINIT(&nes);
nd->nd_repstat = nfsvno_accchk(vp,
Modify the experimental server so that it uses VOP_ACCESSX(). This is necessary in order to enable NFSv4 ACL support. The argument to nfsvno_accchk() was changed to an accmode_t and the function nfsrv_aclaccess() was no longer needed and, therefore, deleted. Reviewed by: trasz MFC after: 2 weeks
2009-12-25 20:44:19 +00:00
VWRITE, cred, &nes, p,
NFSACCCHK_NOOVERRIDE,
NFSACCCHK_VPISLOCKED, 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
nd->nd_repstat = nfsrv_opencheck(clientid,
stateidp, stp, vp, nd, p, nd->nd_repstat);
if (!nd->nd_repstat) {
tempsize = nvap->na_size;
NFSVNO_ATTRINIT(nvap);
nvap->na_size = tempsize;
nd->nd_repstat = VOP_SETATTR(vp,
&nvap->na_vattr, cred);
}
} else if (vp->v_type == VREG) {
nd->nd_repstat = nfsrv_opencheck(clientid,
stateidp, stp, vp, nd, p, nd->nd_repstat);
}
}
} else {
if (ndp->ni_cnd.cn_flags & HASBUF)
nfsvno_relpathbuf(ndp);
if (ndp->ni_startdir && create == NFSV4OPEN_CREATE) {
vrele(ndp->ni_startdir);
if (ndp->ni_dvp == ndp->ni_vp)
vrele(ndp->ni_dvp);
else
vput(ndp->ni_dvp);
if (ndp->ni_vp)
vput(ndp->ni_vp);
}
}
*vpp = vp;
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(0, 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
}
/*
* Updates the file rev and sets the mtime and ctime
* to the current clock time, returning the va_filerev and va_Xtime
* values.
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 ESTALE to indicate the vnode is VIRF_DOOMED.
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 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
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
nfsvno_updfilerev(struct vnode *vp, struct nfsvattr *nvap,
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, struct thread *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 vattr va;
VATTR_NULL(&va);
Use vfs_timestamp() to set file timestamps rather than invoking getmicrotime() or getnanotime() directly in NFS. Reviewed by: rmacklem, bde MFC after: 1 week
2013-01-18 18:43:38 +00:00
vfs_timestamp(&va.va_mtime);
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
if (NFSVOPISLOCKED(vp) != LK_EXCLUSIVE) {
NFSVOPLOCK(vp, LK_UPGRADE | 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))
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
return (ESTALE);
}
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
(void) VOP_SETATTR(vp, &va, nd->nd_cred);
(void) nfsvno_getattr(vp, nvap, nd, p, 1, NULL);
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
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
}
/*
* Glue routine to nfsv4_fillattr().
*/
int
Modify the experimental NFSv4 server so that it handles crossing of server mount points properly. The functions nfsvno_fillattr() and nfsv4_fillattr() were modified to take the extra arguments that are the mount point, a flag to indicate that it is a file system root and the mounted on fileno. The mount point argument needs to be busy when nfsvno_fillattr() is called, since the vp argument is not locked. Reviewed by: kib MFC after: 2 weeks
2011-04-14 21:49:52 +00:00
nfsvno_fillattr(struct nfsrv_descript *nd, struct mount *mp, struct vnode *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 nfsvattr *nvap, fhandle_t *fhp, int rderror, nfsattrbit_t *attrbitp,
Fix the experimental NFSv4 server so that it uses VOP_PATHCONF() to determine if a file system supports NFSv4 ACLs. Since VOP_PATHCONF() must be called with a locked vnode, the function is called before nfsvno_fillattr() and the result is passed in as an extra argument. MFC after: 2 weeks
2011-04-14 23:46:15 +00:00
struct ucred *cred, struct thread *p, int isdgram, int reterr,
int supports_nfsv4acls, int at_root, uint64_t mounted_on_fileno)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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
struct statfs *sf;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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;
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
sf = NULL;
if (nfsrv_devidcnt > 0 &&
(NFSISSET_ATTRBIT(attrbitp, NFSATTRBIT_SPACEAVAIL) ||
NFSISSET_ATTRBIT(attrbitp, NFSATTRBIT_SPACEFREE) ||
NFSISSET_ATTRBIT(attrbitp, NFSATTRBIT_SPACETOTAL))) {
sf = malloc(sizeof(*sf), M_TEMP, M_WAITOK | M_ZERO);
Fix the pNFS server's reporting of disk space usage for the "#<path>" case. The pNFS server would report the total disk space used and free for all of the DSs, even when certain DSs are assigned to the file system via the "#<path>" suffix used in the "nfsd -p" option argument. This patch fixes this case. It only reports usage for the file system that the argument vnode resides on. This is consistent with the non-pNFS NFSv4 server. In NFSv4 it is possible to have subtrees on other file systems, but these are not included in the usage information for NFSv4. Approved by: re (gjb)
2018-10-09 01:10:50 +00:00
error = nfsrv_pnfsstatfs(sf, 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
if (error != 0) {
free(sf, M_TEMP);
sf = NULL;
}
}
Modify the experimental NFSv4 server so that it handles crossing of server mount points properly. The functions nfsvno_fillattr() and nfsv4_fillattr() were modified to take the extra arguments that are the mount point, a flag to indicate that it is a file system root and the mounted on fileno. The mount point argument needs to be busy when nfsvno_fillattr() is called, since the vp argument is not locked. Reviewed by: kib MFC after: 2 weeks
2011-04-14 21:49:52 +00:00
error = nfsv4_fillattr(nd, mp, vp, NULL, &nvap->na_vattr, fhp, rderror,
Fix the experimental NFSv4 server so that it uses VOP_PATHCONF() to determine if a file system supports NFSv4 ACLs. Since VOP_PATHCONF() must be called with a locked vnode, the function is called before nfsvno_fillattr() and the result is passed in as an extra argument. MFC after: 2 weeks
2011-04-14 23:46:15 +00:00
attrbitp, cred, p, isdgram, reterr, supports_nfsv4acls, at_root,
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
mounted_on_fileno, sf);
free(sf, 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(0, 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);
}
/* Since the Readdir vnode ops vary, put the entire functions in here. */
/*
* nfs readdir service
* - mallocs what it thinks is enough to read
* count rounded up to a multiple of DIRBLKSIZ <= NFS_MAXREADDIR
Harden the experimental NFS server a little, by adding extra checks in the readdir functions for non-positive byte count arguments. For the negative case, set it to the maximum allowable, since it was actually a large positive value (unsigned) on the wire. Also, fix up the readdir function comment a bit. Suggested by: dillon AT apollo.backplane.com MFC after: 2 weeks
2010-04-04 23:19:11 +00:00
* - calls VOP_READDIR()
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
* - loops around building the reply
* if the output generated exceeds count break out of loop
* The NFSM_CLGET macro is used here so that the reply will be packed
* tightly in mbuf clusters.
* - it trims out records with d_fileno == 0
* this doesn't matter for Unix clients, but they might confuse clients
* for other os'.
* - it trims out records with d_type == DT_WHT
* these cannot be seen through NFS (unless we extend the protocol)
* The alternate call nfsrvd_readdirplus() does lookups as well.
* PS: The NFS protocol spec. does not clarify what the "count" byte
* argument is a count of.. just name strings and file id's or the
* entire reply rpc or ...
* I tried just file name and id sizes and it confused the Sun client,
* so I am using the full rpc size now. The "paranoia.." comment refers
* to including the status longwords that are not a part of the dir.
* "entry" structures, but are in the rpc.
*/
int
nfsrvd_readdir(struct nfsrv_descript *nd, int isdgram,
Push down the thread argument in NFS server code, using curthread instead of passing it explicitly. No functional changes Reviewed by: rmacklem (earlier version) MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19419
2019-03-04 13:12:23 +00:00
struct vnode *vp, 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 dirent *dp;
u_int32_t *tl;
int dirlen;
char *cpos, *cend, *rbuf;
struct nfsvattr at;
int nlen, error = 0, getret = 1;
int siz, cnt, fullsiz, eofflag, ncookies;
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
u_int64_t off, toff, verf __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
u_long *cookies = NULL, *cookiep;
struct uio io;
struct iovec iv;
Change NFS readdir() to only ignore cookies preceding the given offset for UFS rather than for all but ZFS. This code was assuming that offsets were monotonically increasing for all file systems except ZFS and that the cookies from a previous call may have been rewound to a block boundary. According to mckusick@ only UFS is known to do this, so only requests against UFS file systems should remove cookies smaller than the given offset. This fixes serving TMPFS over NFS as it too does not have monotonically increasing offsets. The comment around the code also indicated it was specific to UFS. Some of the code using 'not_zfs' is specific to ZFS snapshot handling, so add a 'is_zfs' variable for those cases. It's possible that 'is_zfs' check for VFS_VGET() support may not be specific to ZFS. This needs more research and testing. After this fix TMPFS and other file systems can be served over NFS. To test I compared the results of syncing a /usr/src tree into a tmpfs and serving that over NFS. Before the fix 3589 files were missing on the remote view. After the fix all files were successfully found. Reviewed by: rmacklem Discussed with: mckusick, rmacklem via fs@ Discussed at: http://lists.freebsd.org/pipermail/freebsd-fs/2014-April/019264.html MFC after: 2 weeks Sponsored by: EMC / Isilon Storage Division
2014-07-01 20:00:35 +00:00
int is_ufs;
Push down the thread argument in NFS server code, using curthread instead of passing it explicitly. No functional changes Reviewed by: rmacklem (earlier version) MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19419
2019-03-04 13:12:23 +00:00
struct thread *p = curthread;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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_repstat) {
nfsrv_postopattr(nd, getret, &at);
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 (nd->nd_flag & ND_NFSV2) {
NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
off = fxdr_unsigned(u_quad_t, *tl++);
} else {
NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
off = fxdr_hyper(tl);
tl += 2;
verf = fxdr_hyper(tl);
tl += 2;
}
toff = off;
cnt = fxdr_unsigned(int, *tl);
Harden the experimental NFS server a little, by adding extra checks in the readdir functions for non-positive byte count arguments. For the negative case, set it to the maximum allowable, since it was actually a large positive value (unsigned) on the wire. Also, fix up the readdir function comment a bit. Suggested by: dillon AT apollo.backplane.com MFC after: 2 weeks
2010-04-04 23:19:11 +00:00
if (cnt > NFS_SRVMAXDATA(nd) || cnt < 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
cnt = NFS_SRVMAXDATA(nd);
siz = ((cnt + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1));
fullsiz = siz;
if (nd->nd_flag & ND_NFSV3) {
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->nd_repstat = getret = nfsvno_getattr(vp, &at, nd, p, 1,
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 0
/*
* va_filerev is not sufficient as a cookie verifier,
* since it is not supposed to change when entries are
* removed/added unless that offset cookies returned to
* the client are no longer valid.
*/
if (!nd->nd_repstat && toff && verf != at.na_filerev)
nd->nd_repstat = NFSERR_BAD_COOKIE;
#endif
}
Fix a bug that allows NFS clients to issue READDIR on files. PR: kern/178016 Security: CVE-2013-3266 Security: FreeBSD-SA-13:05.nfsserver
2013-04-29 20:09:44 +00:00
if (!nd->nd_repstat && vp->v_type != VDIR)
nd->nd_repstat = NFSERR_NOTDIR;
Harden the experimental NFS server a little, by adding extra checks in the readdir functions for non-positive byte count arguments. For the negative case, set it to the maximum allowable, since it was actually a large positive value (unsigned) on the wire. Also, fix up the readdir function comment a bit. Suggested by: dillon AT apollo.backplane.com MFC after: 2 weeks
2010-04-04 23:19:11 +00:00
if (nd->nd_repstat == 0 && cnt == 0) {
if (nd->nd_flag & ND_NFSV2)
/* NFSv2 does not have NFSERR_TOOSMALL */
nd->nd_repstat = EPERM;
else
nd->nd_repstat = NFSERR_TOOSMALL;
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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_repstat)
Modify the experimental server so that it uses VOP_ACCESSX(). This is necessary in order to enable NFSv4 ACL support. The argument to nfsvno_accchk() was changed to an accmode_t and the function nfsrv_aclaccess() was no longer needed and, therefore, deleted. Reviewed by: trasz MFC after: 2 weeks
2009-12-25 20:44:19 +00:00
nd->nd_repstat = nfsvno_accchk(vp, VEXEC,
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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_cred, exp, p, NFSACCCHK_NOOVERRIDE,
Modify the experimental server so that it uses VOP_ACCESSX(). This is necessary in order to enable NFSv4 ACL support. The argument to nfsvno_accchk() was changed to an accmode_t and the function nfsrv_aclaccess() was no longer needed and, therefore, deleted. Reviewed by: trasz MFC after: 2 weeks
2009-12-25 20:44:19 +00:00
NFSACCCHK_VPISLOCKED, 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 (nd->nd_repstat) {
vput(vp);
if (nd->nd_flag & ND_NFSV3)
nfsrv_postopattr(nd, getret, &at);
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
}
Change NFS readdir() to only ignore cookies preceding the given offset for UFS rather than for all but ZFS. This code was assuming that offsets were monotonically increasing for all file systems except ZFS and that the cookies from a previous call may have been rewound to a block boundary. According to mckusick@ only UFS is known to do this, so only requests against UFS file systems should remove cookies smaller than the given offset. This fixes serving TMPFS over NFS as it too does not have monotonically increasing offsets. The comment around the code also indicated it was specific to UFS. Some of the code using 'not_zfs' is specific to ZFS snapshot handling, so add a 'is_zfs' variable for those cases. It's possible that 'is_zfs' check for VFS_VGET() support may not be specific to ZFS. This needs more research and testing. After this fix TMPFS and other file systems can be served over NFS. To test I compared the results of syncing a /usr/src tree into a tmpfs and serving that over NFS. Before the fix 3589 files were missing on the remote view. After the fix all files were successfully found. Reviewed by: rmacklem Discussed with: mckusick, rmacklem via fs@ Discussed at: http://lists.freebsd.org/pipermail/freebsd-fs/2014-April/019264.html MFC after: 2 weeks Sponsored by: EMC / Isilon Storage Division
2014-07-01 20:00:35 +00:00
is_ufs = strcmp(vp->v_mount->mnt_vfc->vfc_name, "ufs") == 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
rbuf = malloc(siz, 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
again:
eofflag = 0;
if (cookies) {
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(cookies, 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
cookies = NULL;
}
iv.iov_base = rbuf;
iv.iov_len = siz;
io.uio_iov = &iv;
io.uio_iovcnt = 1;
io.uio_offset = (off_t)off;
io.uio_resid = siz;
io.uio_segflg = UIO_SYSSPACE;
io.uio_rw = UIO_READ;
io.uio_td = NULL;
nd->nd_repstat = VOP_READDIR(vp, &io, nd->nd_cred, &eofflag, &ncookies,
&cookies);
off = (u_int64_t)io.uio_offset;
if (io.uio_resid)
siz -= io.uio_resid;
if (!cookies && !nd->nd_repstat)
nd->nd_repstat = NFSERR_PERM;
if (nd->nd_flag & ND_NFSV3) {
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
getret = nfsvno_getattr(vp, &at, nd, p, 1, 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 (!nd->nd_repstat)
nd->nd_repstat = getret;
}
/*
* Handles the failed cases. nd->nd_repstat == 0 past here.
*/
if (nd->nd_repstat) {
Simplify vnode locking in the expeimental NFS server's readdir functions. In particular, get rid of two bogus VOP_ISLOCKED() calls. Removing the VOP_ISLOCKED() calls is the only actual bug fixed by this patch. Reviewed by: kib MFC after: 2 weeks
2010-12-24 20:24:07 +00:00
vput(vp);
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(rbuf, 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 (cookies)
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(cookies, 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 (nd->nd_flag & ND_NFSV3)
nfsrv_postopattr(nd, getret, &at);
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 nothing read, return eof
* rpc reply
*/
if (siz == 0) {
Simplify vnode locking in the expeimental NFS server's readdir functions. In particular, get rid of two bogus VOP_ISLOCKED() calls. Removing the VOP_ISLOCKED() calls is the only actual bug fixed by this patch. Reviewed by: kib MFC after: 2 weeks
2010-12-24 20:24:07 +00:00
vput(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 (nd->nd_flag & ND_NFSV2) {
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
} else {
nfsrv_postopattr(nd, getret, &at);
NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
txdr_hyper(at.na_filerev, tl);
tl += 2;
}
*tl++ = newnfs_false;
*tl = newnfs_true;
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(rbuf, M_TEMP);
free(cookies, 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
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 degenerate cases of nothing useful read.
* If so go try again
*/
cpos = rbuf;
cend = rbuf + siz;
dp = (struct dirent *)cpos;
cookiep = cookies;
/*
* For some reason FreeBSD's ufs_readdir() chooses to back the
* directory offset up to a block boundary, so it is necessary to
* skip over the records that precede the requested offset. This
* requires the assumption that file offset cookies monotonically
* increase.
*/
while (cpos < cend && ncookies > 0 &&
(dp->d_fileno == 0 || dp->d_type == DT_WHT ||
Change NFS readdir() to only ignore cookies preceding the given offset for UFS rather than for all but ZFS. This code was assuming that offsets were monotonically increasing for all file systems except ZFS and that the cookies from a previous call may have been rewound to a block boundary. According to mckusick@ only UFS is known to do this, so only requests against UFS file systems should remove cookies smaller than the given offset. This fixes serving TMPFS over NFS as it too does not have monotonically increasing offsets. The comment around the code also indicated it was specific to UFS. Some of the code using 'not_zfs' is specific to ZFS snapshot handling, so add a 'is_zfs' variable for those cases. It's possible that 'is_zfs' check for VFS_VGET() support may not be specific to ZFS. This needs more research and testing. After this fix TMPFS and other file systems can be served over NFS. To test I compared the results of syncing a /usr/src tree into a tmpfs and serving that over NFS. Before the fix 3589 files were missing on the remote view. After the fix all files were successfully found. Reviewed by: rmacklem Discussed with: mckusick, rmacklem via fs@ Discussed at: http://lists.freebsd.org/pipermail/freebsd-fs/2014-April/019264.html MFC after: 2 weeks Sponsored by: EMC / Isilon Storage Division
2014-07-01 20:00:35 +00:00
(is_ufs == 1 && ((u_quad_t)(*cookiep)) <= toff))) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
cpos += dp->d_reclen;
dp = (struct dirent *)cpos;
cookiep++;
ncookies--;
}
if (cpos >= cend || ncookies == 0) {
siz = fullsiz;
toff = off;
goto again;
}
Simplify vnode locking in the expeimental NFS server's readdir functions. In particular, get rid of two bogus VOP_ISLOCKED() calls. Removing the VOP_ISLOCKED() calls is the only actual bug fixed by this patch. Reviewed by: kib MFC after: 2 weeks
2010-12-24 20:24:07 +00:00
vput(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
Add support for ext_pgs mbufs to nfsrvd_readdir() and nfsrvd_readdirplus(). This patch code that optionally (based on ND_TLS, never set yet) generates readdir replies in ext_pgs mbufs. To trim the list back, a new function that is ext_pgs aware called nfsm_trimtrailing() replaces newnfs_trimtrailing(). newnfs_trimtrailing() is no longer used, but will be removed in a future commit, since its removal does modify the internal kpi between the NFS modules. This is another in the series of commits that add support to the NFS client and server for building RPC messages in ext_pgs mbufs with anonymous pages. This is useful so that the entire mbuf list does not need to be copied before calling sosend() when NFS over TLS is enabled. Use of ext_pgs mbufs will not be enabled until the kernel RPC is updated to handle TLS.
2020-07-29 22:58:08 +00:00
/*
* If cnt > MCLBYTES and the reply will not be saved, use
* ext_pgs mbufs for TLS.
* For NFSv4.0, we do not know for sure if the reply will
* be saved, so do not use ext_pgs mbufs for NFSv4.0.
*/
if (cnt > MCLBYTES && siz > MCLBYTES &&
(nd->nd_flag & (ND_TLS | ND_EXTPG | ND_SAVEREPLY)) == ND_TLS &&
(nd->nd_flag & (ND_NFSV4 | ND_NFSV41)) != ND_NFSV4)
nd->nd_flag |= ND_EXTPG;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* dirlen is the size of the reply, including all XDR and must
* not exceed cnt. For NFSv2, RFC1094 didn't clearly indicate
* if the XDR should be included in "count", but to be safe, we do.
* (Include the two booleans at the end of the reply in dirlen now.)
*/
if (nd->nd_flag & ND_NFSV3) {
nfsrv_postopattr(nd, getret, &at);
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
txdr_hyper(at.na_filerev, tl);
dirlen = NFSX_V3POSTOPATTR + NFSX_VERF + 2 * NFSX_UNSIGNED;
} else {
dirlen = 2 * NFSX_UNSIGNED;
}
/* Loop through the records and build reply */
while (cpos < cend && ncookies > 0) {
nlen = dp->d_namlen;
if (dp->d_fileno != 0 && dp->d_type != DT_WHT &&
nlen <= NFS_MAXNAMLEN) {
if (nd->nd_flag & ND_NFSV3)
dirlen += (6*NFSX_UNSIGNED + NFSM_RNDUP(nlen));
else
dirlen += (4*NFSX_UNSIGNED + NFSM_RNDUP(nlen));
if (dirlen > cnt) {
eofflag = 0;
break;
}
/*
* Build the directory record xdr from
* the dirent entry.
*/
if (nd->nd_flag & ND_NFSV3) {
NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
*tl++ = newnfs_true;
*tl++ = 0;
} else {
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
*tl++ = newnfs_true;
}
*tl = txdr_unsigned(dp->d_fileno);
(void) nfsm_strtom(nd, dp->d_name, nlen);
if (nd->nd_flag & ND_NFSV3) {
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
*tl++ = 0;
} else
NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
*tl = txdr_unsigned(*cookiep);
}
cpos += dp->d_reclen;
dp = (struct dirent *)cpos;
cookiep++;
ncookies--;
}
if (cpos < cend)
eofflag = 0;
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
*tl++ = newnfs_false;
if (eofflag)
*tl = newnfs_true;
else
*tl = newnfs_false;
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(rbuf, M_TEMP);
free(cookies, 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
out:
NFSEXITCODE2(0, 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 (0);
nfsmout:
vput(vp);
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);
}
/*
* Readdirplus for V3 and Readdir for V4.
*/
int
nfsrvd_readdirplus(struct nfsrv_descript *nd, int isdgram,
Push down the thread argument in NFS server code, using curthread instead of passing it explicitly. No functional changes Reviewed by: rmacklem (earlier version) MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19419
2019-03-04 13:12:23 +00:00
struct vnode *vp, 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 dirent *dp;
u_int32_t *tl;
int dirlen;
char *cpos, *cend, *rbuf;
struct vnode *nvp;
fhandle_t nfh;
struct nfsvattr nva, at, *nvap = &nva;
struct mbuf *mb0, *mb1;
struct nfsreferral *refp;
Modify the experimental nfs server so that it falls back to using VOP_LOOKUP() when VFS_VGET() returns EOPNOTSUPP in the ReaddirPlus RPC. This patch is based upon one by pjd@ for the regular nfs server which has not yet been committed. It is needed when a ZFS volume is exported and ReaddirPlus (which almost always happens for NFSv4) is performed by a client. The patch also simplifies vnode lock handling somewhat. MFC after: 2 weeks
2009-11-23 16:08:15 +00:00
int nlen, r, error = 0, getret = 1, usevget = 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 siz, cnt, fullsiz, eofflag, ncookies, entrycnt;
caddr_t bpos0, bpos1;
u_int64_t off, toff, verf;
u_long *cookies = NULL, *cookiep;
nfsattrbit_t attrbits, rderrbits, savbits;
struct uio io;
struct iovec iv;
Modify the experimental nfs server so that it falls back to using VOP_LOOKUP() when VFS_VGET() returns EOPNOTSUPP in the ReaddirPlus RPC. This patch is based upon one by pjd@ for the regular nfs server which has not yet been committed. It is needed when a ZFS volume is exported and ReaddirPlus (which almost always happens for NFSv4) is performed by a client. The patch also simplifies vnode lock handling somewhat. MFC after: 2 weeks
2009-11-23 16:08:15 +00:00
struct componentname cn;
Change NFS readdir() to only ignore cookies preceding the given offset for UFS rather than for all but ZFS. This code was assuming that offsets were monotonically increasing for all file systems except ZFS and that the cookies from a previous call may have been rewound to a block boundary. According to mckusick@ only UFS is known to do this, so only requests against UFS file systems should remove cookies smaller than the given offset. This fixes serving TMPFS over NFS as it too does not have monotonically increasing offsets. The comment around the code also indicated it was specific to UFS. Some of the code using 'not_zfs' is specific to ZFS snapshot handling, so add a 'is_zfs' variable for those cases. It's possible that 'is_zfs' check for VFS_VGET() support may not be specific to ZFS. This needs more research and testing. After this fix TMPFS and other file systems can be served over NFS. To test I compared the results of syncing a /usr/src tree into a tmpfs and serving that over NFS. Before the fix 3589 files were missing on the remote view. After the fix all files were successfully found. Reviewed by: rmacklem Discussed with: mckusick, rmacklem via fs@ Discussed at: http://lists.freebsd.org/pipermail/freebsd-fs/2014-April/019264.html MFC after: 2 weeks Sponsored by: EMC / Isilon Storage Division
2014-07-01 20:00:35 +00:00
int at_root, is_ufs, is_zfs, needs_unbusy, supports_nfsv4acls;
Modify the experimental NFSv4 server so that it handles crossing of server mount points properly. The functions nfsvno_fillattr() and nfsv4_fillattr() were modified to take the extra arguments that are the mount point, a flag to indicate that it is a file system root and the mounted on fileno. The mount point argument needs to be busy when nfsvno_fillattr() is called, since the vp argument is not locked. Reviewed by: kib MFC after: 2 weeks
2011-04-14 21:49:52 +00:00
struct mount *mp, *new_mp;
uint64_t mounted_on_fileno;
Push down the thread argument in NFS server code, using curthread instead of passing it explicitly. No functional changes Reviewed by: rmacklem (earlier version) MFC after: 2 weeks Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D19419
2019-03-04 13:12:23 +00:00
struct thread *p = curthread;
Add support for ext_pgs mbufs to nfsrvd_readdir() and nfsrvd_readdirplus(). This patch code that optionally (based on ND_TLS, never set yet) generates readdir replies in ext_pgs mbufs. To trim the list back, a new function that is ext_pgs aware called nfsm_trimtrailing() replaces newnfs_trimtrailing(). newnfs_trimtrailing() is no longer used, but will be removed in a future commit, since its removal does modify the internal kpi between the NFS modules. This is another in the series of commits that add support to the NFS client and server for building RPC messages in ext_pgs mbufs with anonymous pages. This is useful so that the entire mbuf list does not need to be copied before calling sosend() when NFS over TLS is enabled. Use of ext_pgs mbufs will not be enabled until the kernel RPC is updated to handle TLS.
2020-07-29 22:58:08 +00:00
int bextpg0, bextpg1, bextpgsiz0, bextpgsiz1;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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_repstat) {
nfsrv_postopattr(nd, getret, &at);
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
}
NFSM_DISSECT(tl, u_int32_t *, 6 * NFSX_UNSIGNED);
off = fxdr_hyper(tl);
toff = off;
tl += 2;
verf = fxdr_hyper(tl);
tl += 2;
siz = fxdr_unsigned(int, *tl++);
cnt = fxdr_unsigned(int, *tl);
/*
* Use the server's maximum data transfer size as the upper bound
* on reply datalen.
*/
Harden the experimental NFS server a little, by adding extra checks in the readdir functions for non-positive byte count arguments. For the negative case, set it to the maximum allowable, since it was actually a large positive value (unsigned) on the wire. Also, fix up the readdir function comment a bit. Suggested by: dillon AT apollo.backplane.com MFC after: 2 weeks
2010-04-04 23:19:11 +00:00
if (cnt > NFS_SRVMAXDATA(nd) || cnt < 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
cnt = NFS_SRVMAXDATA(nd);
/*
* siz is a "hint" of how much directory information (name, fileid,
* cookie) should be in the reply. At least one client "hints" 0,
* so I set it to cnt for that case. I also round it up to the
* next multiple of DIRBLKSIZ.
Improve sanity checking for the dircount hint argument to NFSv3's ReaddirPlus and NFSv4's Readdir operations. The code checked for a zero argument, but did not check for a very large value. This patch clips dircount at the server's maximum data size. MFC after: 1 week
2018-11-20 01:59:57 +00:00
* Since the size of a Readdirplus directory entry reply will always
* be greater than a directory entry returned by VOP_READDIR(), it
* does not make sense to read more than NFS_SRVMAXDATA() via
* VOP_READDIR().
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
*/
Harden the experimental NFS server a little, by adding extra checks in the readdir functions for non-positive byte count arguments. For the negative case, set it to the maximum allowable, since it was actually a large positive value (unsigned) on the wire. Also, fix up the readdir function comment a bit. Suggested by: dillon AT apollo.backplane.com MFC after: 2 weeks
2010-04-04 23:19:11 +00:00
if (siz <= 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
siz = cnt;
Improve sanity checking for the dircount hint argument to NFSv3's ReaddirPlus and NFSv4's Readdir operations. The code checked for a zero argument, but did not check for a very large value. This patch clips dircount at the server's maximum data size. MFC after: 1 week
2018-11-20 01:59:57 +00:00
else if (siz > NFS_SRVMAXDATA(nd))
siz = NFS_SRVMAXDATA(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
siz = ((siz + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1));
if (nd->nd_flag & ND_NFSV4) {
error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
if (error)
goto nfsmout;
NFSSET_ATTRBIT(&savbits, &attrbits);
Fix the NFSv4.0 server so that it does not support NFSv4.1 attributes. During inspection of a packet trace, I noticed that an NFSv4.0 mount reported that it supported attributes that are only defined for NFSv4.1. In practice, this bug appears to be benign, since NFSv4.0 clients will not use attributes that were added for NFSv4.1. However, this was not correct and this patch fixes the NFSv4.0 server so that it only supports attributes defined for NFSv4.0. It also adds a definition for NFSv4.1 attributes that can only be set, although it is only defined as 0 for now. This is anticipation of the addition of support for the NFSv4.1 mode+mask attribute soon. MFC after: 2 weeks
2019-04-19 03:36:22 +00:00
NFSCLRNOTFILLABLE_ATTRBIT(&attrbits, 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
NFSZERO_ATTRBIT(&rderrbits);
NFSSETBIT_ATTRBIT(&rderrbits, NFSATTRBIT_RDATTRERROR);
} else {
NFSZERO_ATTRBIT(&attrbits);
}
fullsiz = siz;
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->nd_repstat = getret = nfsvno_getattr(vp, &at, nd, p, 1, NULL);
Ignore the cookie verifier for NFSv4.1 when the cookie is 0. RFC5661 states that the cookie verifier should be 0 when the cookie is 0. However, the wording is somewhat unclear and a recent discussion on the nfsv4@ietf.org mailing list indicated that the NFSv4 server should ignore the cookie verifier's value when the dirctory offset cookie is 0. This patch deletes the check for this that would return NFSERR_BAD_COOKIE when the verifier was not 0. This was found during testing of the ESXi client against the NFSv4.1 server. Reported by: daniel@ftml.net (via packet trace) MFC after: 2 weeks
2018-07-11 23:23:29 +00:00
#if 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_repstat) {
if (off && verf != at.na_filerev) {
/*
* va_filerev is not sufficient as a cookie verifier,
* since it is not supposed to change when entries are
* removed/added unless that offset cookies returned to
* the client are no longer valid.
*/
if (nd->nd_flag & ND_NFSV4) {
nd->nd_repstat = NFSERR_NOTSAME;
} else {
nd->nd_repstat = NFSERR_BAD_COOKIE;
}
}
}
Ignore the cookie verifier for NFSv4.1 when the cookie is 0. RFC5661 states that the cookie verifier should be 0 when the cookie is 0. However, the wording is somewhat unclear and a recent discussion on the nfsv4@ietf.org mailing list indicated that the NFSv4 server should ignore the cookie verifier's value when the dirctory offset cookie is 0. This patch deletes the check for this that would return NFSERR_BAD_COOKIE when the verifier was not 0. This was found during testing of the ESXi client against the NFSv4.1 server. Reported by: daniel@ftml.net (via packet trace) MFC after: 2 weeks
2018-07-11 23:23:29 +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
if (!nd->nd_repstat && vp->v_type != VDIR)
nd->nd_repstat = NFSERR_NOTDIR;
if (!nd->nd_repstat && cnt == 0)
nd->nd_repstat = NFSERR_TOOSMALL;
if (!nd->nd_repstat)
Modify the experimental server so that it uses VOP_ACCESSX(). This is necessary in order to enable NFSv4 ACL support. The argument to nfsvno_accchk() was changed to an accmode_t and the function nfsrv_aclaccess() was no longer needed and, therefore, deleted. Reviewed by: trasz MFC after: 2 weeks
2009-12-25 20:44:19 +00:00
nd->nd_repstat = nfsvno_accchk(vp, VEXEC,
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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_cred, exp, p, NFSACCCHK_NOOVERRIDE,
Modify the experimental server so that it uses VOP_ACCESSX(). This is necessary in order to enable NFSv4 ACL support. The argument to nfsvno_accchk() was changed to an accmode_t and the function nfsrv_aclaccess() was no longer needed and, therefore, deleted. Reviewed by: trasz MFC after: 2 weeks
2009-12-25 20:44:19 +00:00
NFSACCCHK_VPISLOCKED, 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 (nd->nd_repstat) {
vput(vp);
if (nd->nd_flag & ND_NFSV3)
nfsrv_postopattr(nd, getret, &at);
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
}
Change NFS readdir() to only ignore cookies preceding the given offset for UFS rather than for all but ZFS. This code was assuming that offsets were monotonically increasing for all file systems except ZFS and that the cookies from a previous call may have been rewound to a block boundary. According to mckusick@ only UFS is known to do this, so only requests against UFS file systems should remove cookies smaller than the given offset. This fixes serving TMPFS over NFS as it too does not have monotonically increasing offsets. The comment around the code also indicated it was specific to UFS. Some of the code using 'not_zfs' is specific to ZFS snapshot handling, so add a 'is_zfs' variable for those cases. It's possible that 'is_zfs' check for VFS_VGET() support may not be specific to ZFS. This needs more research and testing. After this fix TMPFS and other file systems can be served over NFS. To test I compared the results of syncing a /usr/src tree into a tmpfs and serving that over NFS. Before the fix 3589 files were missing on the remote view. After the fix all files were successfully found. Reviewed by: rmacklem Discussed with: mckusick, rmacklem via fs@ Discussed at: http://lists.freebsd.org/pipermail/freebsd-fs/2014-April/019264.html MFC after: 2 weeks Sponsored by: EMC / Isilon Storage Division
2014-07-01 20:00:35 +00:00
is_ufs = strcmp(vp->v_mount->mnt_vfc->vfc_name, "ufs") == 0;
is_zfs = strcmp(vp->v_mount->mnt_vfc->vfc_name, "zfs") == 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
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
rbuf = malloc(siz, 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
again:
eofflag = 0;
if (cookies) {
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(cookies, 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
cookies = NULL;
}
iv.iov_base = rbuf;
iv.iov_len = siz;
io.uio_iov = &iv;
io.uio_iovcnt = 1;
io.uio_offset = (off_t)off;
io.uio_resid = siz;
io.uio_segflg = UIO_SYSSPACE;
io.uio_rw = UIO_READ;
io.uio_td = NULL;
nd->nd_repstat = VOP_READDIR(vp, &io, nd->nd_cred, &eofflag, &ncookies,
&cookies);
off = (u_int64_t)io.uio_offset;
if (io.uio_resid)
siz -= io.uio_resid;
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
getret = nfsvno_getattr(vp, &at, nd, p, 1, 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 (!cookies && !nd->nd_repstat)
nd->nd_repstat = NFSERR_PERM;
if (!nd->nd_repstat)
nd->nd_repstat = getret;
if (nd->nd_repstat) {
Modify the experimental nfs server so that it falls back to using VOP_LOOKUP() when VFS_VGET() returns EOPNOTSUPP in the ReaddirPlus RPC. This patch is based upon one by pjd@ for the regular nfs server which has not yet been committed. It is needed when a ZFS volume is exported and ReaddirPlus (which almost always happens for NFSv4) is performed by a client. The patch also simplifies vnode lock handling somewhat. MFC after: 2 weeks
2009-11-23 16:08:15 +00:00
vput(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 (cookies)
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(cookies, M_TEMP);
free(rbuf, 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 (nd->nd_flag & ND_NFSV3)
nfsrv_postopattr(nd, getret, &at);
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 nothing read, return eof
* rpc reply
*/
if (siz == 0) {
Modify the experimental nfs server so that it falls back to using VOP_LOOKUP() when VFS_VGET() returns EOPNOTSUPP in the ReaddirPlus RPC. This patch is based upon one by pjd@ for the regular nfs server which has not yet been committed. It is needed when a ZFS volume is exported and ReaddirPlus (which almost always happens for NFSv4) is performed by a client. The patch also simplifies vnode lock handling somewhat. MFC after: 2 weeks
2009-11-23 16:08:15 +00:00
vput(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 (nd->nd_flag & ND_NFSV3)
nfsrv_postopattr(nd, getret, &at);
NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
txdr_hyper(at.na_filerev, tl);
tl += 2;
*tl++ = newnfs_false;
*tl = newnfs_true;
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(cookies, M_TEMP);
free(rbuf, 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
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 degenerate cases of nothing useful read.
* If so go try again
*/
cpos = rbuf;
cend = rbuf + siz;
dp = (struct dirent *)cpos;
cookiep = cookies;
/*
* For some reason FreeBSD's ufs_readdir() chooses to back the
* directory offset up to a block boundary, so it is necessary to
* skip over the records that precede the requested offset. This
* requires the assumption that file offset cookies monotonically
* increase.
*/
while (cpos < cend && ncookies > 0 &&
(dp->d_fileno == 0 || dp->d_type == DT_WHT ||
Change NFS readdir() to only ignore cookies preceding the given offset for UFS rather than for all but ZFS. This code was assuming that offsets were monotonically increasing for all file systems except ZFS and that the cookies from a previous call may have been rewound to a block boundary. According to mckusick@ only UFS is known to do this, so only requests against UFS file systems should remove cookies smaller than the given offset. This fixes serving TMPFS over NFS as it too does not have monotonically increasing offsets. The comment around the code also indicated it was specific to UFS. Some of the code using 'not_zfs' is specific to ZFS snapshot handling, so add a 'is_zfs' variable for those cases. It's possible that 'is_zfs' check for VFS_VGET() support may not be specific to ZFS. This needs more research and testing. After this fix TMPFS and other file systems can be served over NFS. To test I compared the results of syncing a /usr/src tree into a tmpfs and serving that over NFS. Before the fix 3589 files were missing on the remote view. After the fix all files were successfully found. Reviewed by: rmacklem Discussed with: mckusick, rmacklem via fs@ Discussed at: http://lists.freebsd.org/pipermail/freebsd-fs/2014-April/019264.html MFC after: 2 weeks Sponsored by: EMC / Isilon Storage Division
2014-07-01 20:00:35 +00:00
(is_ufs == 1 && ((u_quad_t)(*cookiep)) <= toff) ||
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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_flag & ND_NFSV4) &&
((dp->d_namlen == 1 && dp->d_name[0] == '.') ||
(dp->d_namlen==2 && dp->d_name[0]=='.' && dp->d_name[1]=='.'))))) {
cpos += dp->d_reclen;
dp = (struct dirent *)cpos;
cookiep++;
ncookies--;
}
if (cpos >= cend || ncookies == 0) {
siz = fullsiz;
toff = off;
goto again;
}
Modify readdirplus in the experimental NFS server in a manner analogous to r216633 for the regular server. This change busies the file system so that VFS_VGET() is guaranteed to be using the correct mount point even during a forced dismount attempt. Since nfsd_fhtovp() is not called immediately before readdirplus, the patch is actually a clone of pjd@'s nfs_serv.c.4.patch instead of the one committed in r216633. Reviewed by: kib MFC after: 10 days
2011-01-09 02:10:54 +00:00
/*
* Busy the file system so that the mount point won't go away
* and, as such, VFS_VGET() can be used safely.
*/
mp = vp->v_mount;
vfs_ref(mp);
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);
Modify readdirplus in the experimental NFS server in a manner analogous to r216633 for the regular server. This change busies the file system so that VFS_VGET() is guaranteed to be using the correct mount point even during a forced dismount attempt. Since nfsd_fhtovp() is not called immediately before readdirplus, the patch is actually a clone of pjd@'s nfs_serv.c.4.patch instead of the one committed in r216633. Reviewed by: kib MFC after: 10 days
2011-01-09 02:10:54 +00:00
nd->nd_repstat = vfs_busy(mp, 0);
vfs_rel(mp);
if (nd->nd_repstat != 0) {
vrele(vp);
free(cookies, M_TEMP);
free(rbuf, M_TEMP);
if (nd->nd_flag & ND_NFSV3)
nfsrv_postopattr(nd, getret, &at);
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;
Modify readdirplus in the experimental NFS server in a manner analogous to r216633 for the regular server. This change busies the file system so that VFS_VGET() is guaranteed to be using the correct mount point even during a forced dismount attempt. Since nfsd_fhtovp() is not called immediately before readdirplus, the patch is actually a clone of pjd@'s nfs_serv.c.4.patch instead of the one committed in r216633. Reviewed by: kib MFC after: 10 days
2011-01-09 02:10:54 +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
An intermittent problem with NFSv4 exporting of ZFS snapshots was reported to the freebsd-fs mailing list. I believe the problem was caused by the Readdir operation using VFS_VGET() for a snapshot file entry instead of VOP_LOOKUP(). This would not occur for NFSv3, since it will do a VFS_VGET() of "." which fails with ENOTSUPP at the beginning of the directory, whereas NFSv4 does not check "." or "..". This patch adds a call to VFS_VGET() for the directory being read to check for ENOTSUPP. I also observed that the mount_on_fileid and fsid attributes were not correct at the snapshot's auto mountpoints when looking at packet traces for the Readdir. This patch fixes the attributes by doing a check for different v_mount structure, even if the vnode v_mountedhere is not set. Reported by: jas@cse.yorku.ca Tested by: jas@cse.yorku.ca Reviewed by: asomers MFC after: 1 week
2013-12-24 22:24:17 +00:00
/*
Commit the 64-bit inode project. Extend the ino_t, dev_t, nlink_t types to 64-bit ints. Modify struct dirent layout to add d_off, increase the size of d_fileno to 64-bits, increase the size of d_namlen to 16-bits, and change the required alignment. Increase struct statfs f_mntfromname[] and f_mntonname[] array length MNAMELEN to 1024. ABI breakage is mitigated by providing compatibility using versioned symbols, ingenious use of the existing padding in structures, and by employing other tricks. Unfortunately, not everything can be fixed, especially outside the base system. For instance, third-party APIs which pass struct stat around are broken in backward and forward incompatible ways. Kinfo sysctl MIBs ABI is changed in backward-compatible way, but there is no general mechanism to handle other sysctl MIBS which return structures where the layout has changed. It was considered that the breakage is either in the management interfaces, where we usually allow ABI slip, or is not important. Struct xvnode changed layout, no compat shims are provided. For struct xtty, dev_t tty device member was reduced to uint32_t. It was decided that keeping ABI compat in this case is more useful than reporting 64-bit dev_t, for the sake of pstat. Update note: strictly follow the instructions in UPDATING. Build and install the new kernel with COMPAT_FREEBSD11 option enabled, then reboot, and only then install new world. Credits: The 64-bit inode project, also known as ino64, started life many years ago as a project by Gleb Kurtsou (gleb). Kirk McKusick (mckusick) then picked up and updated the patch, and acted as a flag-waver. Feedback, suggestions, and discussions were carried by Ed Maste (emaste), John Baldwin (jhb), Jilles Tjoelker (jilles), and Rick Macklem (rmacklem). Kris Moore (kris) performed an initial ports investigation followed by an exp-run by Antoine Brodin (antoine). Essential and all-embracing testing was done by Peter Holm (pho). The heavy lifting of coordinating all these efforts and bringing the project to completion were done by Konstantin Belousov (kib). Sponsored by: The FreeBSD Foundation (emaste, kib) Differential revision: https://reviews.freebsd.org/D10439
2017-05-23 09:29:05 +00:00
* Check to see if entries in this directory can be safely acquired
* via VFS_VGET() or if a switch to VOP_LOOKUP() is required.
* ZFS snapshot directories need VOP_LOOKUP(), so that any
* automount of the snapshot directory that is required will
* be done.
* This needs to be done here for NFSv4, since NFSv4 never does
* a VFS_VGET() for "." or "..".
An intermittent problem with NFSv4 exporting of ZFS snapshots was reported to the freebsd-fs mailing list. I believe the problem was caused by the Readdir operation using VFS_VGET() for a snapshot file entry instead of VOP_LOOKUP(). This would not occur for NFSv3, since it will do a VFS_VGET() of "." which fails with ENOTSUPP at the beginning of the directory, whereas NFSv4 does not check "." or "..". This patch adds a call to VFS_VGET() for the directory being read to check for ENOTSUPP. I also observed that the mount_on_fileid and fsid attributes were not correct at the snapshot's auto mountpoints when looking at packet traces for the Readdir. This patch fixes the attributes by doing a check for different v_mount structure, even if the vnode v_mountedhere is not set. Reported by: jas@cse.yorku.ca Tested by: jas@cse.yorku.ca Reviewed by: asomers MFC after: 1 week
2013-12-24 22:24:17 +00:00
*/
Commit the 64-bit inode project. Extend the ino_t, dev_t, nlink_t types to 64-bit ints. Modify struct dirent layout to add d_off, increase the size of d_fileno to 64-bits, increase the size of d_namlen to 16-bits, and change the required alignment. Increase struct statfs f_mntfromname[] and f_mntonname[] array length MNAMELEN to 1024. ABI breakage is mitigated by providing compatibility using versioned symbols, ingenious use of the existing padding in structures, and by employing other tricks. Unfortunately, not everything can be fixed, especially outside the base system. For instance, third-party APIs which pass struct stat around are broken in backward and forward incompatible ways. Kinfo sysctl MIBs ABI is changed in backward-compatible way, but there is no general mechanism to handle other sysctl MIBS which return structures where the layout has changed. It was considered that the breakage is either in the management interfaces, where we usually allow ABI slip, or is not important. Struct xvnode changed layout, no compat shims are provided. For struct xtty, dev_t tty device member was reduced to uint32_t. It was decided that keeping ABI compat in this case is more useful than reporting 64-bit dev_t, for the sake of pstat. Update note: strictly follow the instructions in UPDATING. Build and install the new kernel with COMPAT_FREEBSD11 option enabled, then reboot, and only then install new world. Credits: The 64-bit inode project, also known as ino64, started life many years ago as a project by Gleb Kurtsou (gleb). Kirk McKusick (mckusick) then picked up and updated the patch, and acted as a flag-waver. Feedback, suggestions, and discussions were carried by Ed Maste (emaste), John Baldwin (jhb), Jilles Tjoelker (jilles), and Rick Macklem (rmacklem). Kris Moore (kris) performed an initial ports investigation followed by an exp-run by Antoine Brodin (antoine). Essential and all-embracing testing was done by Peter Holm (pho). The heavy lifting of coordinating all these efforts and bringing the project to completion were done by Konstantin Belousov (kib). Sponsored by: The FreeBSD Foundation (emaste, kib) Differential revision: https://reviews.freebsd.org/D10439
2017-05-23 09:29:05 +00:00
if (is_zfs == 1) {
r = VFS_VGET(mp, at.na_fileid, LK_SHARED, &nvp);
if (r == EOPNOTSUPP) {
usevget = 0;
cn.cn_nameiop = LOOKUP;
cn.cn_lkflags = LK_SHARED | LK_RETRY;
cn.cn_cred = nd->nd_cred;
cn.cn_thread = p;
} else if (r == 0)
vput(nvp);
}
An intermittent problem with NFSv4 exporting of ZFS snapshots was reported to the freebsd-fs mailing list. I believe the problem was caused by the Readdir operation using VFS_VGET() for a snapshot file entry instead of VOP_LOOKUP(). This would not occur for NFSv3, since it will do a VFS_VGET() of "." which fails with ENOTSUPP at the beginning of the directory, whereas NFSv4 does not check "." or "..". This patch adds a call to VFS_VGET() for the directory being read to check for ENOTSUPP. I also observed that the mount_on_fileid and fsid attributes were not correct at the snapshot's auto mountpoints when looking at packet traces for the Readdir. This patch fixes the attributes by doing a check for different v_mount structure, even if the vnode v_mountedhere is not set. Reported by: jas@cse.yorku.ca Tested by: jas@cse.yorku.ca Reviewed by: asomers MFC after: 1 week
2013-12-24 22:24:17 +00:00
Add support for ext_pgs mbufs to nfsrvd_readdir() and nfsrvd_readdirplus(). This patch code that optionally (based on ND_TLS, never set yet) generates readdir replies in ext_pgs mbufs. To trim the list back, a new function that is ext_pgs aware called nfsm_trimtrailing() replaces newnfs_trimtrailing(). newnfs_trimtrailing() is no longer used, but will be removed in a future commit, since its removal does modify the internal kpi between the NFS modules. This is another in the series of commits that add support to the NFS client and server for building RPC messages in ext_pgs mbufs with anonymous pages. This is useful so that the entire mbuf list does not need to be copied before calling sosend() when NFS over TLS is enabled. Use of ext_pgs mbufs will not be enabled until the kernel RPC is updated to handle TLS.
2020-07-29 22:58:08 +00:00
/*
* If the reply is likely to exceed MCLBYTES and the reply will
* not be saved, use ext_pgs mbufs for TLS.
* It is difficult to predict how large each entry will be and
* how many entries have been read, so just assume the directory
* entries grow by a factor of 4 when attributes are included.
* For NFSv4.0, we do not know for sure if the reply will
* be saved, so do not use ext_pgs mbufs for NFSv4.0.
*/
if (cnt > MCLBYTES && siz > MCLBYTES / 4 &&
(nd->nd_flag & (ND_TLS | ND_EXTPG | ND_SAVEREPLY)) == ND_TLS &&
(nd->nd_flag & (ND_NFSV4 | ND_NFSV41)) != ND_NFSV4)
nd->nd_flag |= ND_EXTPG;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* Save this position, in case there is an error before one entry
* is created.
*/
mb0 = nd->nd_mb;
bpos0 = nd->nd_bpos;
Add support for ext_pgs mbufs to nfsrvd_readdir() and nfsrvd_readdirplus(). This patch code that optionally (based on ND_TLS, never set yet) generates readdir replies in ext_pgs mbufs. To trim the list back, a new function that is ext_pgs aware called nfsm_trimtrailing() replaces newnfs_trimtrailing(). newnfs_trimtrailing() is no longer used, but will be removed in a future commit, since its removal does modify the internal kpi between the NFS modules. This is another in the series of commits that add support to the NFS client and server for building RPC messages in ext_pgs mbufs with anonymous pages. This is useful so that the entire mbuf list does not need to be copied before calling sosend() when NFS over TLS is enabled. Use of ext_pgs mbufs will not be enabled until the kernel RPC is updated to handle TLS.
2020-07-29 22:58:08 +00:00
bextpg0 = nd->nd_bextpg;
bextpgsiz0 = nd->nd_bextpgsiz;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
* Fill in the first part of the reply.
* dirlen is the reply length in bytes and cannot exceed cnt.
* (Include the two booleans at the end of the reply in dirlen now,
* so we recognize when we have exceeded cnt.)
*/
if (nd->nd_flag & ND_NFSV3) {
dirlen = NFSX_V3POSTOPATTR + NFSX_VERF + 2 * NFSX_UNSIGNED;
nfsrv_postopattr(nd, getret, &at);
} else {
dirlen = NFSX_VERF + 2 * NFSX_UNSIGNED;
}
NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
txdr_hyper(at.na_filerev, tl);
/*
* Save this position, in case there is an empty reply needed.
*/
mb1 = nd->nd_mb;
bpos1 = nd->nd_bpos;
Add support for ext_pgs mbufs to nfsrvd_readdir() and nfsrvd_readdirplus(). This patch code that optionally (based on ND_TLS, never set yet) generates readdir replies in ext_pgs mbufs. To trim the list back, a new function that is ext_pgs aware called nfsm_trimtrailing() replaces newnfs_trimtrailing(). newnfs_trimtrailing() is no longer used, but will be removed in a future commit, since its removal does modify the internal kpi between the NFS modules. This is another in the series of commits that add support to the NFS client and server for building RPC messages in ext_pgs mbufs with anonymous pages. This is useful so that the entire mbuf list does not need to be copied before calling sosend() when NFS over TLS is enabled. Use of ext_pgs mbufs will not be enabled until the kernel RPC is updated to handle TLS.
2020-07-29 22:58:08 +00:00
bextpg1 = nd->nd_bextpg;
bextpgsiz1 = nd->nd_bextpgsiz;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 through the records and build reply */
entrycnt = 0;
while (cpos < cend && ncookies > 0 && dirlen < cnt) {
nlen = dp->d_namlen;
if (dp->d_fileno != 0 && dp->d_type != DT_WHT &&
nlen <= NFS_MAXNAMLEN &&
((nd->nd_flag & ND_NFSV3) || nlen > 2 ||
(nlen==2 && (dp->d_name[0]!='.' || dp->d_name[1]!='.'))
|| (nlen == 1 && dp->d_name[0] != '.'))) {
/*
* Save the current position in the reply, in case
* this entry exceeds cnt.
*/
mb1 = nd->nd_mb;
bpos1 = nd->nd_bpos;
Add support for ext_pgs mbufs to nfsrvd_readdir() and nfsrvd_readdirplus(). This patch code that optionally (based on ND_TLS, never set yet) generates readdir replies in ext_pgs mbufs. To trim the list back, a new function that is ext_pgs aware called nfsm_trimtrailing() replaces newnfs_trimtrailing(). newnfs_trimtrailing() is no longer used, but will be removed in a future commit, since its removal does modify the internal kpi between the NFS modules. This is another in the series of commits that add support to the NFS client and server for building RPC messages in ext_pgs mbufs with anonymous pages. This is useful so that the entire mbuf list does not need to be copied before calling sosend() when NFS over TLS is enabled. Use of ext_pgs mbufs will not be enabled until the kernel RPC is updated to handle TLS.
2020-07-29 22:58:08 +00:00
bextpg1 = nd->nd_bextpg;
bextpgsiz1 = nd->nd_bextpgsiz;
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
/*
* For readdir_and_lookup get the vnode using
* the file number.
*/
nvp = NULL;
refp = NULL;
r = 0;
Modify the experimental NFSv4 server so that it handles crossing of server mount points properly. The functions nfsvno_fillattr() and nfsv4_fillattr() were modified to take the extra arguments that are the mount point, a flag to indicate that it is a file system root and the mounted on fileno. The mount point argument needs to be busy when nfsvno_fillattr() is called, since the vp argument is not locked. Reviewed by: kib MFC after: 2 weeks
2011-04-14 21:49:52 +00:00
at_root = 0;
needs_unbusy = 0;
new_mp = mp;
mounted_on_fileno = (uint64_t)dp->d_fileno;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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_NFSV3) ||
NFSNONZERO_ATTRBIT(&savbits)) {
if (nd->nd_flag & ND_NFSV4)
refp = nfsv4root_getreferral(NULL,
vp, dp->d_fileno);
Modify the experimental nfs server so that it falls back to using VOP_LOOKUP() when VFS_VGET() returns EOPNOTSUPP in the ReaddirPlus RPC. This patch is based upon one by pjd@ for the regular nfs server which has not yet been committed. It is needed when a ZFS volume is exported and ReaddirPlus (which almost always happens for NFSv4) is performed by a client. The patch also simplifies vnode lock handling somewhat. MFC after: 2 weeks
2009-11-23 16:08:15 +00:00
if (refp == NULL) {
if (usevget)
Modify readdirplus in the experimental NFS server in a manner analogous to r216633 for the regular server. This change busies the file system so that VFS_VGET() is guaranteed to be using the correct mount point even during a forced dismount attempt. Since nfsd_fhtovp() is not called immediately before readdirplus, the patch is actually a clone of pjd@'s nfs_serv.c.4.patch instead of the one committed in r216633. Reviewed by: kib MFC after: 10 days
2011-01-09 02:10:54 +00:00
r = VFS_VGET(mp, dp->d_fileno,
LK_SHARED, &nvp);
Modify the experimental nfs server so that it falls back to using VOP_LOOKUP() when VFS_VGET() returns EOPNOTSUPP in the ReaddirPlus RPC. This patch is based upon one by pjd@ for the regular nfs server which has not yet been committed. It is needed when a ZFS volume is exported and ReaddirPlus (which almost always happens for NFSv4) is performed by a client. The patch also simplifies vnode lock handling somewhat. MFC after: 2 weeks
2009-11-23 16:08:15 +00:00
else
r = EOPNOTSUPP;
if (r == EOPNOTSUPP) {
Commit the 64-bit inode project. Extend the ino_t, dev_t, nlink_t types to 64-bit ints. Modify struct dirent layout to add d_off, increase the size of d_fileno to 64-bits, increase the size of d_namlen to 16-bits, and change the required alignment. Increase struct statfs f_mntfromname[] and f_mntonname[] array length MNAMELEN to 1024. ABI breakage is mitigated by providing compatibility using versioned symbols, ingenious use of the existing padding in structures, and by employing other tricks. Unfortunately, not everything can be fixed, especially outside the base system. For instance, third-party APIs which pass struct stat around are broken in backward and forward incompatible ways. Kinfo sysctl MIBs ABI is changed in backward-compatible way, but there is no general mechanism to handle other sysctl MIBS which return structures where the layout has changed. It was considered that the breakage is either in the management interfaces, where we usually allow ABI slip, or is not important. Struct xvnode changed layout, no compat shims are provided. For struct xtty, dev_t tty device member was reduced to uint32_t. It was decided that keeping ABI compat in this case is more useful than reporting 64-bit dev_t, for the sake of pstat. Update note: strictly follow the instructions in UPDATING. Build and install the new kernel with COMPAT_FREEBSD11 option enabled, then reboot, and only then install new world. Credits: The 64-bit inode project, also known as ino64, started life many years ago as a project by Gleb Kurtsou (gleb). Kirk McKusick (mckusick) then picked up and updated the patch, and acted as a flag-waver. Feedback, suggestions, and discussions were carried by Ed Maste (emaste), John Baldwin (jhb), Jilles Tjoelker (jilles), and Rick Macklem (rmacklem). Kris Moore (kris) performed an initial ports investigation followed by an exp-run by Antoine Brodin (antoine). Essential and all-embracing testing was done by Peter Holm (pho). The heavy lifting of coordinating all these efforts and bringing the project to completion were done by Konstantin Belousov (kib). Sponsored by: The FreeBSD Foundation (emaste, kib) Differential revision: https://reviews.freebsd.org/D10439
2017-05-23 09:29:05 +00:00
if (usevget) {
usevget = 0;
cn.cn_nameiop = LOOKUP;
cn.cn_lkflags =
LK_SHARED |
LK_RETRY;
cn.cn_cred =
nd->nd_cred;
cn.cn_thread = p;
}
Modify the experimental nfs server so that it falls back to using VOP_LOOKUP() when VFS_VGET() returns EOPNOTSUPP in the ReaddirPlus RPC. This patch is based upon one by pjd@ for the regular nfs server which has not yet been committed. It is needed when a ZFS volume is exported and ReaddirPlus (which almost always happens for NFSv4) is performed by a client. The patch also simplifies vnode lock handling somewhat. MFC after: 2 weeks
2009-11-23 16:08:15 +00:00
cn.cn_nameptr = dp->d_name;
cn.cn_namelen = nlen;
cn.cn_flags = ISLASTCN |
Remove the support for using non-mpsafe filesystem modules. In particular, do not lock Giant conditionally when calling into the filesystem module, remove the VFS_LOCK_GIANT() and related macros. Stop handling buffers belonging to non-mpsafe filesystems. The VFS_VERSION is bumped to indicate the interface change which does not result in the interface signatures changes. Conducted and reviewed by: attilio Tested by: pho
2012-10-22 17:50:54 +00:00
NOFOLLOW | LOCKLEAF;
Modify the experimental nfs server so that it falls back to using VOP_LOOKUP() when VFS_VGET() returns EOPNOTSUPP in the ReaddirPlus RPC. This patch is based upon one by pjd@ for the regular nfs server which has not yet been committed. It is needed when a ZFS volume is exported and ReaddirPlus (which almost always happens for NFSv4) is performed by a client. The patch also simplifies vnode lock handling somewhat. MFC after: 2 weeks
2009-11-23 16:08:15 +00:00
if (nlen == 2 &&
dp->d_name[0] == '.' &&
dp->d_name[1] == '.')
cn.cn_flags |=
ISDOTDOT;
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_SHARED)
Simplify vnode locking in the expeimental NFS server's readdir functions. In particular, get rid of two bogus VOP_ISLOCKED() calls. Removing the VOP_ISLOCKED() calls is the only actual bug fixed by this patch. Reviewed by: kib MFC after: 2 weeks
2010-12-24 20:24:07 +00:00
!= 0) {
nd->nd_repstat = EPERM;
break;
}
Modify the experimental NFS server in a manner analagous to r214049 for the regular NFS server, so that it will not do a VOP_LOOKUP() of ".." when at the root of a file system when performing a ReaddirPlus RPC. MFC after: 10 days
2010-10-21 18:49:12 +00:00
if ((vp->v_vflag & VV_ROOT) != 0
&& (cn.cn_flags & ISDOTDOT)
!= 0) {
vref(vp);
nvp = vp;
r = 0;
Add a VOP_UNLOCK() for the directory, when that is not what VOP_LOOKUP() returned. This fixes a bug in the experimental NFS server for the case where VFS_VGET() fails returning EOPNOTSUPP in the ReaddirPlus RPC, forcing the use of VOP_LOOKUP() instead. MFC after: 2 weeks
2011-04-09 23:55:27 +00:00
} else {
Modify the experimental NFS server in a manner analagous to r214049 for the regular NFS server, so that it will not do a VOP_LOOKUP() of ".." when at the root of a file system when performing a ReaddirPlus RPC. MFC after: 10 days
2010-10-21 18:49:12 +00:00
r = VOP_LOOKUP(vp, &nvp,
&cn);
Add a VOP_UNLOCK() for the directory, when that is not what VOP_LOOKUP() returned. This fixes a bug in the experimental NFS server for the case where VFS_VGET() fails returning EOPNOTSUPP in the ReaddirPlus RPC, forcing the use of VOP_LOOKUP() instead. MFC after: 2 weeks
2011-04-09 23:55:27 +00:00
if (vp != nvp)
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 a VOP_UNLOCK() for the directory, when that is not what VOP_LOOKUP() returned. This fixes a bug in the experimental NFS server for the case where VFS_VGET() fails returning EOPNOTSUPP in the ReaddirPlus RPC, forcing the use of VOP_LOOKUP() instead. MFC after: 2 weeks
2011-04-09 23:55:27 +00:00
}
Modify the experimental nfs server so that it falls back to using VOP_LOOKUP() when VFS_VGET() returns EOPNOTSUPP in the ReaddirPlus RPC. This patch is based upon one by pjd@ for the regular nfs server which has not yet been committed. It is needed when a ZFS volume is exported and ReaddirPlus (which almost always happens for NFSv4) is performed by a client. The patch also simplifies vnode lock handling somewhat. MFC after: 2 weeks
2009-11-23 16:08:15 +00:00
}
Modify the experimental NFSv4 server so that it handles crossing of server mount points properly. The functions nfsvno_fillattr() and nfsv4_fillattr() were modified to take the extra arguments that are the mount point, a flag to indicate that it is a file system root and the mounted on fileno. The mount point argument needs to be busy when nfsvno_fillattr() is called, since the vp argument is not locked. Reviewed by: kib MFC after: 2 weeks
2011-04-14 21:49:52 +00:00
/*
* For NFSv4, check to see if nvp is
* a mount point and get the mount
* point vnode, as required.
*/
if (r == 0 &&
nfsrv_enable_crossmntpt != 0 &&
(nd->nd_flag & ND_NFSV4) != 0 &&
nvp->v_type == VDIR &&
nvp->v_mountedhere != NULL) {
new_mp = nvp->v_mountedhere;
r = vfs_busy(new_mp, 0);
vput(nvp);
nvp = NULL;
if (r == 0) {
r = VFS_ROOT(new_mp,
LK_SHARED, &nvp);
needs_unbusy = 1;
if (r == 0)
at_root = 1;
}
}
Modify the experimental nfs server so that it falls back to using VOP_LOOKUP() when VFS_VGET() returns EOPNOTSUPP in the ReaddirPlus RPC. This patch is based upon one by pjd@ for the regular nfs server which has not yet been committed. It is needed when a ZFS volume is exported and ReaddirPlus (which almost always happens for NFSv4) is performed by a client. The patch also simplifies vnode lock handling somewhat. MFC after: 2 weeks
2009-11-23 16:08:15 +00:00
}
nfsrvd_readdirplus: for some errors, do not fail the entire request Instead, a failing entry is skipped. This change consist of two logical changes. A failure to vget or lookup an entry is considered to be a result of a concurrent removal, which is the only reasonable explanation given that the filesystem is busied. So, the entry would be silently skipped. In the case of a failure to get attributes of an entry for an NFSv3 request, the entry would be silently skipped. There can be legitimate reasons for the failure, but NFSv3 does not provide any means to report the error, so we have two options: either fail the whole request or ignore the failed entry. Traditionally, the old NFS server used the latter option, so the code is reverted to it. Making the whole directory unreadable because of a single entry seems to be unpractical. Additionally, some bits of code are slightly re-arranged to account for the new control flow and to honor style(9). Reviewed by: rmacklem Sponsored by: Panzura Differential Revision: https://reviews.freebsd.org/D15424
2018-10-22 15:33:05 +00:00
/*
* If we failed to look up the entry, then it
* has become invalid, most likely removed.
*/
if (r != 0) {
if (needs_unbusy)
vfs_unbusy(new_mp);
goto invalid;
}
KASSERT(refp != NULL || nvp != NULL,
("%s: undetected lookup error", __func__));
if (refp == NULL &&
((nd->nd_flag & ND_NFSV3) ||
NFSNONZERO_ATTRBIT(&attrbits))) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
r = nfsvno_getfh(nvp, &nfh, p);
if (!r)
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
r = nfsvno_getattr(nvp, nvap, nd, p,
1, &attrbits);
Change NFS readdir() to only ignore cookies preceding the given offset for UFS rather than for all but ZFS. This code was assuming that offsets were monotonically increasing for all file systems except ZFS and that the cookies from a previous call may have been rewound to a block boundary. According to mckusick@ only UFS is known to do this, so only requests against UFS file systems should remove cookies smaller than the given offset. This fixes serving TMPFS over NFS as it too does not have monotonically increasing offsets. The comment around the code also indicated it was specific to UFS. Some of the code using 'not_zfs' is specific to ZFS snapshot handling, so add a 'is_zfs' variable for those cases. It's possible that 'is_zfs' check for VFS_VGET() support may not be specific to ZFS. This needs more research and testing. After this fix TMPFS and other file systems can be served over NFS. To test I compared the results of syncing a /usr/src tree into a tmpfs and serving that over NFS. Before the fix 3589 files were missing on the remote view. After the fix all files were successfully found. Reviewed by: rmacklem Discussed with: mckusick, rmacklem via fs@ Discussed at: http://lists.freebsd.org/pipermail/freebsd-fs/2014-April/019264.html MFC after: 2 weeks Sponsored by: EMC / Isilon Storage Division
2014-07-01 20:00:35 +00:00
if (r == 0 && is_zfs == 1 &&
An intermittent problem with NFSv4 exporting of ZFS snapshots was reported to the freebsd-fs mailing list. I believe the problem was caused by the Readdir operation using VFS_VGET() for a snapshot file entry instead of VOP_LOOKUP(). This would not occur for NFSv3, since it will do a VFS_VGET() of "." which fails with ENOTSUPP at the beginning of the directory, whereas NFSv4 does not check "." or "..". This patch adds a call to VFS_VGET() for the directory being read to check for ENOTSUPP. I also observed that the mount_on_fileid and fsid attributes were not correct at the snapshot's auto mountpoints when looking at packet traces for the Readdir. This patch fixes the attributes by doing a check for different v_mount structure, even if the vnode v_mountedhere is not set. Reported by: jas@cse.yorku.ca Tested by: jas@cse.yorku.ca Reviewed by: asomers MFC after: 1 week
2013-12-24 22:24:17 +00:00
nfsrv_enable_crossmntpt != 0 &&
(nd->nd_flag & ND_NFSV4) != 0 &&
nvp->v_type == VDIR &&
vp->v_mount != nvp->v_mount) {
/*
* For a ZFS snapshot, there is a
* pseudo mount that does not set
* v_mountedhere, so it needs to
* be detected via a different
* mount structure.
*/
at_root = 1;
if (new_mp == mp)
new_mp = nvp->v_mount;
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
}
nfsrvd_readdirplus: for some errors, do not fail the entire request Instead, a failing entry is skipped. This change consist of two logical changes. A failure to vget or lookup an entry is considered to be a result of a concurrent removal, which is the only reasonable explanation given that the filesystem is busied. So, the entry would be silently skipped. In the case of a failure to get attributes of an entry for an NFSv3 request, the entry would be silently skipped. There can be legitimate reasons for the failure, but NFSv3 does not provide any means to report the error, so we have two options: either fail the whole request or ignore the failed entry. Traditionally, the old NFS server used the latter option, so the code is reverted to it. Making the whole directory unreadable because of a single entry seems to be unpractical. Additionally, some bits of code are slightly re-arranged to account for the new control flow and to honor style(9). Reviewed by: rmacklem Sponsored by: Panzura Differential Revision: https://reviews.freebsd.org/D15424
2018-10-22 15:33:05 +00:00
/*
* If we failed to get attributes of the entry,
* then just skip it for NFSv3 (the traditional
* behavior in the old NFS server).
* For NFSv4 the behavior is controlled by
* RDATTRERROR: we either ignore the error or
* fail the request.
* Note that RDATTRERROR is never set for NFSv3.
*/
if (r != 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 (!NFSISSET_ATTRBIT(&attrbits,
NFSATTRBIT_RDATTRERROR)) {
nfsrvd_readdirplus: for some errors, do not fail the entire request Instead, a failing entry is skipped. This change consist of two logical changes. A failure to vget or lookup an entry is considered to be a result of a concurrent removal, which is the only reasonable explanation given that the filesystem is busied. So, the entry would be silently skipped. In the case of a failure to get attributes of an entry for an NFSv3 request, the entry would be silently skipped. There can be legitimate reasons for the failure, but NFSv3 does not provide any means to report the error, so we have two options: either fail the whole request or ignore the failed entry. Traditionally, the old NFS server used the latter option, so the code is reverted to it. Making the whole directory unreadable because of a single entry seems to be unpractical. Additionally, some bits of code are slightly re-arranged to account for the new control flow and to honor style(9). Reviewed by: rmacklem Sponsored by: Panzura Differential Revision: https://reviews.freebsd.org/D15424
2018-10-22 15:33:05 +00:00
vput(nvp);
Modify the experimental NFSv4 server so that it handles crossing of server mount points properly. The functions nfsvno_fillattr() and nfsv4_fillattr() were modified to take the extra arguments that are the mount point, a flag to indicate that it is a file system root and the mounted on fileno. The mount point argument needs to be busy when nfsvno_fillattr() is called, since the vp argument is not locked. Reviewed by: kib MFC after: 2 weeks
2011-04-14 21:49:52 +00:00
if (needs_unbusy != 0)
vfs_unbusy(new_mp);
nfsrvd_readdirplus: for some errors, do not fail the entire request Instead, a failing entry is skipped. This change consist of two logical changes. A failure to vget or lookup an entry is considered to be a result of a concurrent removal, which is the only reasonable explanation given that the filesystem is busied. So, the entry would be silently skipped. In the case of a failure to get attributes of an entry for an NFSv3 request, the entry would be silently skipped. There can be legitimate reasons for the failure, but NFSv3 does not provide any means to report the error, so we have two options: either fail the whole request or ignore the failed entry. Traditionally, the old NFS server used the latter option, so the code is reverted to it. Making the whole directory unreadable because of a single entry seems to be unpractical. Additionally, some bits of code are slightly re-arranged to account for the new control flow and to honor style(9). Reviewed by: rmacklem Sponsored by: Panzura Differential Revision: https://reviews.freebsd.org/D15424
2018-10-22 15:33:05 +00:00
if ((nd->nd_flag & ND_NFSV3))
goto invalid;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 = r;
break;
}
}
}
/*
* Build the directory record xdr
*/
if (nd->nd_flag & ND_NFSV3) {
NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
*tl++ = newnfs_true;
*tl++ = 0;
*tl = txdr_unsigned(dp->d_fileno);
dirlen += nfsm_strtom(nd, dp->d_name, nlen);
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
*tl++ = 0;
*tl = txdr_unsigned(*cookiep);
nfsrv_postopattr(nd, 0, nvap);
dirlen += nfsm_fhtom(nd,(u_int8_t *)&nfh,0,1);
dirlen += (5*NFSX_UNSIGNED+NFSX_V3POSTOPATTR);
if (nvp != NULL)
vput(nvp);
} else {
NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
*tl++ = newnfs_true;
*tl++ = 0;
*tl = txdr_unsigned(*cookiep);
dirlen += nfsm_strtom(nd, dp->d_name, nlen);
Fix the experimental NFSv4 server so that it uses VOP_PATHCONF() to determine if a file system supports NFSv4 ACLs. Since VOP_PATHCONF() must be called with a locked vnode, the function is called before nfsvno_fillattr() and the result is passed in as an extra argument. MFC after: 2 weeks
2011-04-14 23:46:15 +00:00
if (nvp != NULL) {
supports_nfsv4acls =
nfs_supportsnfsv4acls(nvp);
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(nvp);
Fix the experimental NFSv4 server so that it uses VOP_PATHCONF() to determine if a file system supports NFSv4 ACLs. Since VOP_PATHCONF() must be called with a locked vnode, the function is called before nfsvno_fillattr() and the result is passed in as an extra argument. MFC after: 2 weeks
2011-04-14 23:46:15 +00:00
} else
supports_nfsv4acls = 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 (refp != NULL) {
dirlen += nfsrv_putreferralattr(nd,
&savbits, refp, 0,
&nd->nd_repstat);
if (nd->nd_repstat) {
if (nvp != NULL)
vrele(nvp);
Modify the experimental NFSv4 server so that it handles crossing of server mount points properly. The functions nfsvno_fillattr() and nfsv4_fillattr() were modified to take the extra arguments that are the mount point, a flag to indicate that it is a file system root and the mounted on fileno. The mount point argument needs to be busy when nfsvno_fillattr() is called, since the vp argument is not locked. Reviewed by: kib MFC after: 2 weeks
2011-04-14 21:49:52 +00:00
if (needs_unbusy != 0)
vfs_unbusy(new_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
break;
}
} else if (r) {
Modify the experimental NFSv4 server so that it handles crossing of server mount points properly. The functions nfsvno_fillattr() and nfsv4_fillattr() were modified to take the extra arguments that are the mount point, a flag to indicate that it is a file system root and the mounted on fileno. The mount point argument needs to be busy when nfsvno_fillattr() is called, since the vp argument is not locked. Reviewed by: kib MFC after: 2 weeks
2011-04-14 21:49:52 +00:00
dirlen += nfsvno_fillattr(nd, new_mp,
nvp, nvap, &nfh, r, &rderrbits,
Fix the experimental NFSv4 server so that it uses VOP_PATHCONF() to determine if a file system supports NFSv4 ACLs. Since VOP_PATHCONF() must be called with a locked vnode, the function is called before nfsvno_fillattr() and the result is passed in as an extra argument. MFC after: 2 weeks
2011-04-14 23:46:15 +00:00
nd->nd_cred, p, isdgram, 0,
supports_nfsv4acls, at_root,
Modify the experimental NFSv4 server so that it handles crossing of server mount points properly. The functions nfsvno_fillattr() and nfsv4_fillattr() were modified to take the extra arguments that are the mount point, a flag to indicate that it is a file system root and the mounted on fileno. The mount point argument needs to be busy when nfsvno_fillattr() is called, since the vp argument is not locked. Reviewed by: kib MFC after: 2 weeks
2011-04-14 21:49:52 +00:00
mounted_on_fileno);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 {
Modify the experimental NFSv4 server so that it handles crossing of server mount points properly. The functions nfsvno_fillattr() and nfsv4_fillattr() were modified to take the extra arguments that are the mount point, a flag to indicate that it is a file system root and the mounted on fileno. The mount point argument needs to be busy when nfsvno_fillattr() is called, since the vp argument is not locked. Reviewed by: kib MFC after: 2 weeks
2011-04-14 21:49:52 +00:00
dirlen += nfsvno_fillattr(nd, new_mp,
nvp, nvap, &nfh, r, &attrbits,
Fix the experimental NFSv4 server so that it uses VOP_PATHCONF() to determine if a file system supports NFSv4 ACLs. Since VOP_PATHCONF() must be called with a locked vnode, the function is called before nfsvno_fillattr() and the result is passed in as an extra argument. MFC after: 2 weeks
2011-04-14 23:46:15 +00:00
nd->nd_cred, p, isdgram, 0,
supports_nfsv4acls, at_root,
Modify the experimental NFSv4 server so that it handles crossing of server mount points properly. The functions nfsvno_fillattr() and nfsv4_fillattr() were modified to take the extra arguments that are the mount point, a flag to indicate that it is a file system root and the mounted on fileno. The mount point argument needs to be busy when nfsvno_fillattr() is called, since the vp argument is not locked. Reviewed by: kib MFC after: 2 weeks
2011-04-14 21:49:52 +00:00
mounted_on_fileno);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 (nvp != NULL)
vrele(nvp);
dirlen += (3 * NFSX_UNSIGNED);
}
Modify the experimental NFSv4 server so that it handles crossing of server mount points properly. The functions nfsvno_fillattr() and nfsv4_fillattr() were modified to take the extra arguments that are the mount point, a flag to indicate that it is a file system root and the mounted on fileno. The mount point argument needs to be busy when nfsvno_fillattr() is called, since the vp argument is not locked. Reviewed by: kib MFC after: 2 weeks
2011-04-14 21:49:52 +00:00
if (needs_unbusy != 0)
vfs_unbusy(new_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 (dirlen <= cnt)
entrycnt++;
}
nfsrvd_readdirplus: for some errors, do not fail the entire request Instead, a failing entry is skipped. This change consist of two logical changes. A failure to vget or lookup an entry is considered to be a result of a concurrent removal, which is the only reasonable explanation given that the filesystem is busied. So, the entry would be silently skipped. In the case of a failure to get attributes of an entry for an NFSv3 request, the entry would be silently skipped. There can be legitimate reasons for the failure, but NFSv3 does not provide any means to report the error, so we have two options: either fail the whole request or ignore the failed entry. Traditionally, the old NFS server used the latter option, so the code is reverted to it. Making the whole directory unreadable because of a single entry seems to be unpractical. Additionally, some bits of code are slightly re-arranged to account for the new control flow and to honor style(9). Reviewed by: rmacklem Sponsored by: Panzura Differential Revision: https://reviews.freebsd.org/D15424
2018-10-22 15:33:05 +00:00
invalid:
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
cpos += dp->d_reclen;
dp = (struct dirent *)cpos;
cookiep++;
ncookies--;
}
Simplify vnode locking in the expeimental NFS server's readdir functions. In particular, get rid of two bogus VOP_ISLOCKED() calls. Removing the VOP_ISLOCKED() calls is the only actual bug fixed by this patch. Reviewed by: kib MFC after: 2 weeks
2010-12-24 20:24:07 +00:00
vrele(vp);
Modify readdirplus in the experimental NFS server in a manner analogous to r216633 for the regular server. This change busies the file system so that VFS_VGET() is guaranteed to be using the correct mount point even during a forced dismount attempt. Since nfsd_fhtovp() is not called immediately before readdirplus, the patch is actually a clone of pjd@'s nfs_serv.c.4.patch instead of the one committed in r216633. Reviewed by: kib MFC after: 10 days
2011-01-09 02:10:54 +00:00
vfs_unbusy(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 dirlen > cnt, we must strip off the last entry. If that
* results in an empty reply, report NFSERR_TOOSMALL.
*/
if (dirlen > cnt || nd->nd_repstat) {
if (!nd->nd_repstat && entrycnt == 0)
nd->nd_repstat = NFSERR_TOOSMALL;
The new NFSv3 server did not generate directory postop attributes for the reply to ReaddirPlus when the server failed within the loop that calls VFS_VGET(). This failure is most likely an error return from VFS_VGET() caused by a bogus d_fileno that was truncated to 32bits. This patch fixes the server so that it will return directory postop attributes for the failure. It does not fix the underlying issue caused by d_fileno being uint32_t when a file system like ZFS generates a fileno that is greater than 32bits. Reported by: jpaetzel Reviewed by: jpaetzel MFC after: 1 month
2014-07-04 22:47:07 +00:00
if (nd->nd_repstat) {
Add support for ext_pgs mbufs to nfsrvd_readdir() and nfsrvd_readdirplus(). This patch code that optionally (based on ND_TLS, never set yet) generates readdir replies in ext_pgs mbufs. To trim the list back, a new function that is ext_pgs aware called nfsm_trimtrailing() replaces newnfs_trimtrailing(). newnfs_trimtrailing() is no longer used, but will be removed in a future commit, since its removal does modify the internal kpi between the NFS modules. This is another in the series of commits that add support to the NFS client and server for building RPC messages in ext_pgs mbufs with anonymous pages. This is useful so that the entire mbuf list does not need to be copied before calling sosend() when NFS over TLS is enabled. Use of ext_pgs mbufs will not be enabled until the kernel RPC is updated to handle TLS.
2020-07-29 22:58:08 +00:00
nfsm_trimtrailing(nd, mb0, bpos0, bextpg0, bextpgsiz0);
The new NFSv3 server did not generate directory postop attributes for the reply to ReaddirPlus when the server failed within the loop that calls VFS_VGET(). This failure is most likely an error return from VFS_VGET() caused by a bogus d_fileno that was truncated to 32bits. This patch fixes the server so that it will return directory postop attributes for the failure. It does not fix the underlying issue caused by d_fileno being uint32_t when a file system like ZFS generates a fileno that is greater than 32bits. Reported by: jpaetzel Reviewed by: jpaetzel MFC after: 1 month
2014-07-04 22:47:07 +00:00
if (nd->nd_flag & ND_NFSV3)
nfsrv_postopattr(nd, getret, &at);
} else
Add support for ext_pgs mbufs to nfsrvd_readdir() and nfsrvd_readdirplus(). This patch code that optionally (based on ND_TLS, never set yet) generates readdir replies in ext_pgs mbufs. To trim the list back, a new function that is ext_pgs aware called nfsm_trimtrailing() replaces newnfs_trimtrailing(). newnfs_trimtrailing() is no longer used, but will be removed in a future commit, since its removal does modify the internal kpi between the NFS modules. This is another in the series of commits that add support to the NFS client and server for building RPC messages in ext_pgs mbufs with anonymous pages. This is useful so that the entire mbuf list does not need to be copied before calling sosend() when NFS over TLS is enabled. Use of ext_pgs mbufs will not be enabled until the kernel RPC is updated to handle TLS.
2020-07-29 22:58:08 +00:00
nfsm_trimtrailing(nd, mb1, bpos1, bextpg1, bextpgsiz1);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
eofflag = 0;
} else if (cpos < cend)
eofflag = 0;
if (!nd->nd_repstat) {
NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
*tl++ = newnfs_false;
if (eofflag)
*tl = newnfs_true;
else
*tl = newnfs_false;
}
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(cookies, M_TEMP);
free(rbuf, 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
out:
NFSEXITCODE2(0, 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 (0);
nfsmout:
vput(vp);
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);
}
/*
* Get the settable attributes out of the mbuf list.
* (Return 0 or EBADRPC)
*/
int
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
nfsrv_sattr(struct nfsrv_descript *nd, vnode_t vp, struct nfsvattr *nvap,
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 *attrbitp, NFSACL_T *aclp, struct thread *p)
{
u_int32_t *tl;
struct nfsv2_sattr *sp;
int error = 0, toclient = 0;
switch (nd->nd_flag & (ND_NFSV2 | ND_NFSV3 | ND_NFSV4)) {
case ND_NFSV2:
NFSM_DISSECT(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
/*
* Some old clients didn't fill in the high order 16bits.
* --> check the low order 2 bytes for 0xffff
*/
if ((fxdr_unsigned(int, sp->sa_mode) & 0xffff) != 0xffff)
nvap->na_mode = nfstov_mode(sp->sa_mode);
if (sp->sa_uid != newnfs_xdrneg1)
nvap->na_uid = fxdr_unsigned(uid_t, sp->sa_uid);
if (sp->sa_gid != newnfs_xdrneg1)
nvap->na_gid = fxdr_unsigned(gid_t, sp->sa_gid);
if (sp->sa_size != newnfs_xdrneg1)
nvap->na_size = fxdr_unsigned(u_quad_t, sp->sa_size);
if (sp->sa_atime.nfsv2_sec != newnfs_xdrneg1) {
#ifdef notyet
fxdr_nfsv2time(&sp->sa_atime, &nvap->na_atime);
#else
nvap->na_atime.tv_sec =
fxdr_unsigned(u_int32_t,sp->sa_atime.nfsv2_sec);
nvap->na_atime.tv_nsec = 0;
#endif
}
if (sp->sa_mtime.nfsv2_sec != newnfs_xdrneg1)
fxdr_nfsv2time(&sp->sa_mtime, &nvap->na_mtime);
break;
case ND_NFSV3:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (*tl == newnfs_true) {
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
nvap->na_mode = nfstov_mode(*tl);
}
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (*tl == newnfs_true) {
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
nvap->na_uid = fxdr_unsigned(uid_t, *tl);
}
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (*tl == newnfs_true) {
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
nvap->na_gid = fxdr_unsigned(gid_t, *tl);
}
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (*tl == newnfs_true) {
NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
nvap->na_size = fxdr_hyper(tl);
}
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
switch (fxdr_unsigned(int, *tl)) {
case NFSV3SATTRTIME_TOCLIENT:
NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
fxdr_nfsv3time(tl, &nvap->na_atime);
toclient = 1;
break;
case NFSV3SATTRTIME_TOSERVER:
Use vfs_timestamp() to set file timestamps rather than invoking getmicrotime() or getnanotime() directly in NFS. Reviewed by: rmacklem, bde MFC after: 1 week
2013-01-18 18:43:38 +00:00
vfs_timestamp(&nvap->na_atime);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nvap->na_vaflags |= VA_UTIMES_NULL;
break;
Cleanup unnecessary semicolons from the kernel. Found with devel/coccinelle.
2016-04-10 23:07:00 +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
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
switch (fxdr_unsigned(int, *tl)) {
case NFSV3SATTRTIME_TOCLIENT:
NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
fxdr_nfsv3time(tl, &nvap->na_mtime);
nvap->na_vaflags &= ~VA_UTIMES_NULL;
break;
case NFSV3SATTRTIME_TOSERVER:
Use vfs_timestamp() to set file timestamps rather than invoking getmicrotime() or getnanotime() directly in NFS. Reviewed by: rmacklem, bde MFC after: 1 week
2013-01-18 18:43:38 +00:00
vfs_timestamp(&nvap->na_mtime);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 (!toclient)
nvap->na_vaflags |= VA_UTIMES_NULL;
break;
Cleanup unnecessary semicolons from the kernel. Found with devel/coccinelle.
2016-04-10 23:07:00 +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
break;
case ND_NFSV4:
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
error = nfsv4_sattr(nd, vp, nvap, attrbitp, aclp, p);
Cleanup unnecessary semicolons from the kernel. Found with devel/coccinelle.
2016-04-10 23:07:00 +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
nfsmout:
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);
}
/*
* Handle the setable attributes for V4.
* Returns NFSERR_BADXDR if it can't be parsed, 0 otherwise.
*/
int
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
nfsv4_sattr(struct nfsrv_descript *nd, vnode_t vp, struct nfsvattr *nvap,
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 *attrbitp, NFSACL_T *aclp, struct thread *p)
{
u_int32_t *tl;
int attrsum = 0;
int i, j;
int error, attrsize, bitpos, aclsize, aceerr, retnotsup = 0;
Add support for the ModeSetMasked attribute to the NFSv4.1 server. I do not know of an extant NFSv4.1 client that currently does a Setattr operation for the ModeSetMasked, but it has been discussed on the linux-nfs mailing list. This patch adds support for doing a Setattr of ModeSetMasked, so that it will work for any future NFSv4.1 client that chooses to do so. Tested via a hacked FreeBSD NFSv4.1 client. MFC after: 2 weeks
2019-04-19 23:35:08 +00:00
int moderet, toclient = 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
u_char *cp, namestr[NFSV4_SMALLSTR + 1];
uid_t uid;
gid_t gid;
Add support for the ModeSetMasked attribute to the NFSv4.1 server. I do not know of an extant NFSv4.1 client that currently does a Setattr operation for the ModeSetMasked, but it has been discussed on the linux-nfs mailing list. This patch adds support for doing a Setattr of ModeSetMasked, so that it will work for any future NFSv4.1 client that chooses to do so. Tested via a hacked FreeBSD NFSv4.1 client. MFC after: 2 weeks
2019-04-19 23:35:08 +00:00
u_short mode, mask; /* Same type as va_mode. */
struct vattr va;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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_getattrbits(nd, attrbitp, NULL, &retnotsup);
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 nfsmout;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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);
attrsize = fxdr_unsigned(int, *tl);
/*
* Loop around getting the setable attributes. If an unsupported
* one is found, set nd_repstat == NFSERR_ATTRNOTSUPP and return.
*/
if (retnotsup) {
nd->nd_repstat = NFSERR_ATTRNOTSUPP;
bitpos = NFSATTRBIT_MAX;
} else {
bitpos = 0;
}
Add support for the ModeSetMasked attribute to the NFSv4.1 server. I do not know of an extant NFSv4.1 client that currently does a Setattr operation for the ModeSetMasked, but it has been discussed on the linux-nfs mailing list. This patch adds support for doing a Setattr of ModeSetMasked, so that it will work for any future NFSv4.1 client that chooses to do so. Tested via a hacked FreeBSD NFSv4.1 client. MFC after: 2 weeks
2019-04-19 23:35:08 +00:00
moderet = 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
for (; bitpos < NFSATTRBIT_MAX; bitpos++) {
if (attrsum > attrsize) {
error = NFSERR_BADXDR;
goto nfsmout;
}
if (NFSISSET_ATTRBIT(attrbitp, bitpos))
switch (bitpos) {
case NFSATTRBIT_SIZE:
NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
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 (vp != NULL && vp->v_type != VREG) {
error = (vp->v_type == VDIR) ? NFSERR_ISDIR :
NFSERR_INVAL;
goto nfsmout;
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nvap->na_size = fxdr_hyper(tl);
attrsum += NFSX_HYPER;
break;
case NFSATTRBIT_ACL:
error = nfsrv_dissectacl(nd, aclp, &aceerr, &aclsize,
p);
if (error)
goto nfsmout;
if (aceerr && !nd->nd_repstat)
Small acl patch to return the aclerror that comes back from nfsrv_dissectacl(). This fixes a problem where ATTRNOTSUPP was being returned instead of BADOWNER. Reviewed by: rmacklem Approved by: zml (mentor) MFC after: 2 weeks
2011-07-16 08:04:57 +00:00
nd->nd_repstat = aceerr;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
attrsum += aclsize;
break;
case NFSATTRBIT_ARCHIVE:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (!nd->nd_repstat)
nd->nd_repstat = NFSERR_ATTRNOTSUPP;
attrsum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_HIDDEN:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (!nd->nd_repstat)
nd->nd_repstat = NFSERR_ATTRNOTSUPP;
attrsum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_MIMETYPE:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
i = fxdr_unsigned(int, *tl);
error = nfsm_advance(nd, NFSM_RNDUP(i), -1);
if (error)
goto nfsmout;
if (!nd->nd_repstat)
nd->nd_repstat = NFSERR_ATTRNOTSUPP;
attrsum += (NFSX_UNSIGNED + NFSM_RNDUP(i));
break;
case NFSATTRBIT_MODE:
Add support for the ModeSetMasked attribute to the NFSv4.1 server. I do not know of an extant NFSv4.1 client that currently does a Setattr operation for the ModeSetMasked, but it has been discussed on the linux-nfs mailing list. This patch adds support for doing a Setattr of ModeSetMasked, so that it will work for any future NFSv4.1 client that chooses to do so. Tested via a hacked FreeBSD NFSv4.1 client. MFC after: 2 weeks
2019-04-19 23:35:08 +00:00
moderet = NFSERR_INVAL; /* Can't do MODESETMASKED. */
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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);
nvap->na_mode = nfstov_mode(*tl);
attrsum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_OWNER:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
j = fxdr_unsigned(int, *tl);
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 (j < 0) {
error = NFSERR_BADXDR;
goto nfsmout;
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 (j > NFSV4_SMALLSTR)
cp = malloc(j + 1, M_NFSSTRING, M_WAITOK);
else
cp = namestr;
error = nfsrv_mtostr(nd, cp, j);
if (error) {
if (j > NFSV4_SMALLSTR)
free(cp, M_NFSSTRING);
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 nfsmout;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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_repstat) {
Drop unused 'p' argument to nfsv4_strtouid(). MFC after: 2 weeks Sponsored by: DARPA, AFRL
2019-03-12 15:02:52 +00:00
nd->nd_repstat = nfsv4_strtouid(nd, cp, j,
&uid);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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_repstat)
nvap->na_uid = uid;
}
if (j > NFSV4_SMALLSTR)
free(cp, M_NFSSTRING);
attrsum += (NFSX_UNSIGNED + NFSM_RNDUP(j));
break;
case NFSATTRBIT_OWNERGROUP:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
j = fxdr_unsigned(int, *tl);
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 (j < 0) {
error = NFSERR_BADXDR;
goto nfsmout;
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 (j > NFSV4_SMALLSTR)
cp = malloc(j + 1, M_NFSSTRING, M_WAITOK);
else
cp = namestr;
error = nfsrv_mtostr(nd, cp, j);
if (error) {
if (j > NFSV4_SMALLSTR)
free(cp, M_NFSSTRING);
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 nfsmout;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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_repstat) {
Drop unused 'p' argument to nfsv4_strtogid(). MFC after: 2 weeks Sponsored by: DARPA, AFRL
2019-03-12 15:07:47 +00:00
nd->nd_repstat = nfsv4_strtogid(nd, cp, j,
&gid);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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_repstat)
nvap->na_gid = gid;
}
if (j > NFSV4_SMALLSTR)
free(cp, M_NFSSTRING);
attrsum += (NFSX_UNSIGNED + NFSM_RNDUP(j));
break;
case NFSATTRBIT_SYSTEM:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
if (!nd->nd_repstat)
nd->nd_repstat = NFSERR_ATTRNOTSUPP;
attrsum += NFSX_UNSIGNED;
break;
case NFSATTRBIT_TIMEACCESSSET:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
attrsum += NFSX_UNSIGNED;
if (fxdr_unsigned(int, *tl)==NFSV4SATTRTIME_TOCLIENT) {
NFSM_DISSECT(tl, u_int32_t *, NFSX_V4TIME);
fxdr_nfsv4time(tl, &nvap->na_atime);
toclient = 1;
attrsum += NFSX_V4TIME;
} else {
Use vfs_timestamp() to set file timestamps rather than invoking getmicrotime() or getnanotime() directly in NFS. Reviewed by: rmacklem, bde MFC after: 1 week
2013-01-18 18:43:38 +00:00
vfs_timestamp(&nvap->na_atime);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nvap->na_vaflags |= VA_UTIMES_NULL;
}
break;
case NFSATTRBIT_TIMEBACKUP:
NFSM_DISSECT(tl, u_int32_t *, NFSX_V4TIME);
if (!nd->nd_repstat)
nd->nd_repstat = NFSERR_ATTRNOTSUPP;
attrsum += NFSX_V4TIME;
break;
case NFSATTRBIT_TIMECREATE:
NFSM_DISSECT(tl, u_int32_t *, NFSX_V4TIME);
Fix the NFS server so that it sets va_birthtime. r362490 marked that the NFSv4 attribute TimeCreate (va_birthtime) is supported, but it did not change the NFS server code to actually do it. As such, errors could occur when unrolling a tarball onto an NFSv4 mounted volume, since setting TimeCreate would fail with a NFSERR_ATTRNOTSUPP reply. This patch fixes the server so that it does TimeCreate and also makes sure that TimeCreate will not be set for a DS file for a pNFS server. A separate commit will add a check to the NFSv4 client for support of the TimeCreate attribute before attempting to set it, to avoid a problem when mounting a server that does not support the attribute. The failures will still occur for r362490 or later kernels that do not have this patch, since they indicate support for the attribute, but do not actually support the attribute.
2020-07-26 23:03:41 +00:00
fxdr_nfsv4time(tl, &nvap->na_btime);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
attrsum += NFSX_V4TIME;
break;
case NFSATTRBIT_TIMEMODIFYSET:
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
attrsum += NFSX_UNSIGNED;
if (fxdr_unsigned(int, *tl)==NFSV4SATTRTIME_TOCLIENT) {
NFSM_DISSECT(tl, u_int32_t *, NFSX_V4TIME);
fxdr_nfsv4time(tl, &nvap->na_mtime);
nvap->na_vaflags &= ~VA_UTIMES_NULL;
attrsum += NFSX_V4TIME;
} else {
Use vfs_timestamp() to set file timestamps rather than invoking getmicrotime() or getnanotime() directly in NFS. Reviewed by: rmacklem, bde MFC after: 1 week
2013-01-18 18:43:38 +00:00
vfs_timestamp(&nvap->na_mtime);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 (!toclient)
nvap->na_vaflags |= VA_UTIMES_NULL;
}
break;
Add support for the ModeSetMasked attribute to the NFSv4.1 server. I do not know of an extant NFSv4.1 client that currently does a Setattr operation for the ModeSetMasked, but it has been discussed on the linux-nfs mailing list. This patch adds support for doing a Setattr of ModeSetMasked, so that it will work for any future NFSv4.1 client that chooses to do so. Tested via a hacked FreeBSD NFSv4.1 client. MFC after: 2 weeks
2019-04-19 23:35:08 +00:00
case NFSATTRBIT_MODESETMASKED:
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
mode = fxdr_unsigned(u_short, *tl++);
mask = fxdr_unsigned(u_short, *tl);
/*
* vp == NULL implies an Open/Create operation.
* This attribute can only be used for Setattr and
* only for NFSv4.1 or higher.
* If moderet != 0, a mode attribute has also been
* specified and this attribute cannot be done in the
* same Setattr operation.
*/
if ((nd->nd_flag & ND_NFSV41) == 0)
nd->nd_repstat = NFSERR_ATTRNOTSUPP;
else if ((mode & ~07777) != 0 || (mask & ~07777) != 0 ||
vp == NULL)
nd->nd_repstat = NFSERR_INVAL;
else if (moderet == 0)
moderet = VOP_GETATTR(vp, &va, nd->nd_cred);
if (moderet == 0)
nvap->na_mode = (mode & mask) |
(va.va_mode & ~mask);
else
nd->nd_repstat = moderet;
attrsum += 2 * NFSX_UNSIGNED;
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
default:
nd->nd_repstat = NFSERR_ATTRNOTSUPP;
/*
* set bitpos so we drop out of the loop.
*/
bitpos = NFSATTRBIT_MAX;
break;
Cleanup unnecessary semicolons from the kernel. Found with devel/coccinelle.
2016-04-10 23:07:00 +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
}
/*
* some clients pad the attrlist, so we need to skip over the
* padding.
*/
if (attrsum > attrsize) {
error = NFSERR_BADXDR;
} else {
attrsize = NFSM_RNDUP(attrsize);
if (attrsum < attrsize)
error = nfsm_advance(nd, attrsize - attrsum, -1);
}
nfsmout:
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);
}
/*
* Check/setup export credentials.
*/
int
nfsd_excred(struct nfsrv_descript *nd, struct nfsexstuff *exp,
nfsd: Fix when NFSERR_WRONGSEC may be replied to NFSv4 clients Commit d224f05fcfc1 pre-parsed the next operation number for the put file handle operations. This patch uses this next operation number, plus the type of the file handle being set by the put file handle operation, to implement the rules in RFC5661 Sec. 2.6 with respect to replying NFSERR_WRONGSEC. This patch also adds a check to see if NFSERR_WRONGSEC should be replied when about to perform Lookup, Lookupp or Open with a file name component, so that the NFSERR_WRONGSEC reply is done for these operations, as required by RFC5661 Sec. 2.6. This patch does not have any practical effect for the FreeBSD NFSv4 client and I believe that the same is true for the Linux client, since NFSERR_WRONGSEC is considered a fatal error at this time. MFC after: 2 weeks
2021-06-05 23:53:07 +00:00
struct ucred *credanon, bool testsec)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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: Fix when NFSERR_WRONGSEC may be replied to NFSv4 clients Commit d224f05fcfc1 pre-parsed the next operation number for the put file handle operations. This patch uses this next operation number, plus the type of the file handle being set by the put file handle operation, to implement the rules in RFC5661 Sec. 2.6 with respect to replying NFSERR_WRONGSEC. This patch also adds a check to see if NFSERR_WRONGSEC should be replied when about to perform Lookup, Lookupp or Open with a file name component, so that the NFSERR_WRONGSEC reply is done for these operations, as required by RFC5661 Sec. 2.6. This patch does not have any practical effect for the FreeBSD NFSv4 client and I believe that the same is true for the Linux client, since NFSERR_WRONGSEC is considered a fatal error at this time. MFC after: 2 weeks
2021-06-05 23:53:07 +00:00
int 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/setup credentials.
*/
if (nd->nd_flag & ND_GSS)
Apply changes to the experimental nfs server so that it uses the security flavors as exported in FreeBSD-CURRENT. This allows it to use a slightly modified mountd.c instead of a different utility. Approved by: kib (mentor)
2009-05-14 21:39:08 +00:00
exp->nes_exflag &= ~MNT_EXPORTANON;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
/*
Apply changes to the experimental nfs server so that it uses the security flavors as exported in FreeBSD-CURRENT. This allows it to use a slightly modified mountd.c instead of a different utility. Approved by: kib (mentor)
2009-05-14 21:39:08 +00:00
* Check to see if the operation is allowed for this security flavor.
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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: Fix when NFSERR_WRONGSEC may be replied to NFSv4 clients Commit d224f05fcfc1 pre-parsed the next operation number for the put file handle operations. This patch uses this next operation number, plus the type of the file handle being set by the put file handle operation, to implement the rules in RFC5661 Sec. 2.6 with respect to replying NFSERR_WRONGSEC. This patch also adds a check to see if NFSERR_WRONGSEC should be replied when about to perform Lookup, Lookupp or Open with a file name component, so that the NFSERR_WRONGSEC reply is done for these operations, as required by RFC5661 Sec. 2.6. This patch does not have any practical effect for the FreeBSD NFSv4 client and I believe that the same is true for the Linux client, since NFSERR_WRONGSEC is considered a fatal error at this time. MFC after: 2 weeks
2021-06-05 23:53:07 +00:00
error = 0;
if (testsec) {
error = nfsvno_testexp(nd, exp);
if (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
}
/*
* Check to see if the file system is exported V4 only.
*/
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 (NFSVNO_EXV4ONLY(exp) && !(nd->nd_flag & ND_NFSV4)) {
error = NFSERR_PROGNOTV4;
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, map the user credentials.
* (Note that ND_AUTHNONE will only be set for an NFSv3
* Fsinfo RPC. If set for anything else, this code might need
* to change.)
*/
Add kernel support to the NFS server for the "-manage-gids" option that will be added to the nfsuserd daemon in a future commit. It modifies the cache used by NFSv4 for name<-->id translation (both username/uid and group/gid) to support this. When "-manage-gids" is set, the server looks up each uid for the RPC and uses the list of groups cached in the server instead of the list of groups provided in the RPC request. The cached group list is acquired for the cache by the nfsuserd daemon via getgrouplist(3). This avoids the 16 groups limit for the list in the RPC request. Since the cache is now used for every RPC when "-manage-gids" is enabled, the code also modifies the cache to use a separate mutex for each hash list instead of a single global mutex. Suggested by: jpaetzel Tested by: jpaetzel MFC after: 2 weeks
2015-11-30 21:54:27 +00:00
if (NFSVNO_EXPORTED(exp)) {
if (((nd->nd_flag & ND_GSS) == 0 && nd->nd_cred->cr_uid == 0) ||
NFSVNO_EXPORTANON(exp) ||
(nd->nd_flag & ND_AUTHNONE) != 0) {
nd->nd_cred->cr_uid = credanon->cr_uid;
nd->nd_cred->cr_gid = credanon->cr_gid;
crsetgroups(nd->nd_cred, credanon->cr_ngroups,
credanon->cr_groups);
} else if ((nd->nd_flag & ND_GSS) == 0) {
/*
* If using AUTH_SYS, call nfsrv_getgrpscred() to see
* if there is a replacement credential with a group
* list set up by "nfsuserd -manage-gids".
* If there is no replacement, nfsrv_getgrpscred()
* simply returns its argument.
*/
nd->nd_cred = nfsrv_getgrpscred(nd->nd_cred);
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 exports.
*/
int
nfsvno_checkexp(struct mount *mp, struct sockaddr *nam, struct nfsexstuff *exp,
struct ucred **credp)
{
Fix export_args ex_flags field so that is 64bits, the same as mnt_flags. Since mnt_flags was upgraded to 64bits there has been a quirk in "struct export_args", since it hold a copy of mnt_flags in ex_flags, which is an "int" (32bits). This happens to currently work, since all the flag bits used in ex_flags are defined in the low order 32bits. However, new export flags cannot be defined. Also, ex_anon is a "struct xucred", which limits it to 16 additional groups. This patch revises "struct export_args" to make ex_flags 64bits and replaces ex_anon with ex_uid, ex_ngroups and ex_groups (which points to a groups list, so it can be malloc'd up to NGROUPS in size. This requires that the VFS_CHECKEXP() arguments change, so I also modified the last "secflavors" argument to be an array pointer, so that the secflavors could be copied in VFS_CHECKEXP() while the export entry is locked. (Without this patch VFS_CHECKEXP() returns a pointer to the secflavors array and then it is used after being unlocked, which is potentially a problem if the exports entry is changed. In practice this does not occur when mountd is run with "-S", but I think it is worth fixing.) This patch also deleted the vfs_oexport_conv() function, since do_mount_update() does the conversion, as required by the old vfs_cmount() calls. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:10:18 +00:00
int 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
error = VFS_CHECKEXP(mp, nam, &exp->nes_exflag, credp,
Fix export_args ex_flags field so that is 64bits, the same as mnt_flags. Since mnt_flags was upgraded to 64bits there has been a quirk in "struct export_args", since it hold a copy of mnt_flags in ex_flags, which is an "int" (32bits). This happens to currently work, since all the flag bits used in ex_flags are defined in the low order 32bits. However, new export flags cannot be defined. Also, ex_anon is a "struct xucred", which limits it to 16 additional groups. This patch revises "struct export_args" to make ex_flags 64bits and replaces ex_anon with ex_uid, ex_ngroups and ex_groups (which points to a groups list, so it can be malloc'd up to NGROUPS in size. This requires that the VFS_CHECKEXP() arguments change, so I also modified the last "secflavors" argument to be an array pointer, so that the secflavors could be copied in VFS_CHECKEXP() while the export entry is locked. (Without this patch VFS_CHECKEXP() returns a pointer to the secflavors array and then it is used after being unlocked, which is potentially a problem if the exports entry is changed. In practice this does not occur when mountd is run with "-S", but I think it is worth fixing.) This patch also deleted the vfs_oexport_conv() function, since do_mount_update() does the conversion, as required by the old vfs_cmount() calls. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:10:18 +00:00
&exp->nes_numsecflavor, exp->nes_secflavors);
Apply changes to the experimental nfs server so that it uses the security flavors as exported in FreeBSD-CURRENT. This allows it to use a slightly modified mountd.c instead of a different utility. Approved by: kib (mentor)
2009-05-14 21:39:08 +00:00
if (error) {
if (nfs_rootfhset) {
exp->nes_exflag = 0;
exp->nes_numsecflavor = 0;
error = 0;
}
Add a sanity check for nes_numsecflavor to the NFS server. Ryan Moeller reported crashes in the NFS server that appear to be caused by stack corruption in nfsrv_compound(). It appears that the stack got corrupted just after a NFSv4.1 Lookup that crosses a server mount point. Although it is just a "theory" at this point, the most obvious way the stack could get corrupted would be if nfsvno_checkexp() somehow acquires an export with a bogus nes_numsecflavor value. This would cause the copying of the secflavors to run off the end of the array, which is allocated on the stack below where the corruption occurs. This sanity check is simple to do and would stop the stack corruption if the theory is correct. Otherwise, doing the sanity check seems to be a reasonable safety belt to add to the code. Reported by: freqlabs MFC after: 2 weeks
2020-04-17 02:21:46 +00:00
} else if (exp->nes_numsecflavor < 1 || exp->nes_numsecflavor >
MAXSECFLAVORS) {
printf("nfsvno_checkexp: numsecflavors out of range\n");
exp->nes_numsecflavor = 0;
error = EACCES;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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
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 a vnode for a file handle and export stuff.
*/
int
nfsvno_fhtovp(struct mount *mp, fhandle_t *fhp, struct sockaddr *nam,
Modify the experimental NFS server so that it uses LK_SHARED for RPC operations when it can. Since VFS_FHTOVP() currently always gets an exclusively locked vnode and is usually called at the beginning of each RPC, the RPCs for a given vnode will still be serialized. As such, passing a lock type argument to VFS_FHTOVP() would be preferable to doing the vn_lock() with LK_DOWNGRADE after the VFS_FHTOVP() call. Reviewed by: kib MFC after: 2 weeks
2010-12-25 21:56:25 +00:00
int lktype, struct vnode **vpp, struct nfsexstuff *exp,
struct ucred **credp)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 export_args ex_flags field so that is 64bits, the same as mnt_flags. Since mnt_flags was upgraded to 64bits there has been a quirk in "struct export_args", since it hold a copy of mnt_flags in ex_flags, which is an "int" (32bits). This happens to currently work, since all the flag bits used in ex_flags are defined in the low order 32bits. However, new export flags cannot be defined. Also, ex_anon is a "struct xucred", which limits it to 16 additional groups. This patch revises "struct export_args" to make ex_flags 64bits and replaces ex_anon with ex_uid, ex_ngroups and ex_groups (which points to a groups list, so it can be malloc'd up to NGROUPS in size. This requires that the VFS_CHECKEXP() arguments change, so I also modified the last "secflavors" argument to be an array pointer, so that the secflavors could be copied in VFS_CHECKEXP() while the export entry is locked. (Without this patch VFS_CHECKEXP() returns a pointer to the secflavors array and then it is used after being unlocked, which is potentially a problem if the exports entry is changed. In practice this does not occur when mountd is run with "-S", but I think it is worth fixing.) This patch also deleted the vfs_oexport_conv() function, since do_mount_update() does the conversion, as required by the old vfs_cmount() calls. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:10:18 +00:00
int 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
Modify nfsvno_fhtovp() to ensure that it always sets the credp argument. Returning without credp set could result in a caller doing crfree() on garbage. Reviewed by: kan Approved by: kib (mentor)
2009-05-11 18:45:04 +00:00
*credp = NULL;
Apply changes to the experimental nfs server so that it uses the security flavors as exported in FreeBSD-CURRENT. This allows it to use a slightly modified mountd.c instead of a different utility. Approved by: kib (mentor)
2009-05-14 21:39:08 +00:00
exp->nes_numsecflavor = 0;
Remove the support for using non-mpsafe filesystem modules. In particular, do not lock Giant conditionally when calling into the filesystem module, remove the VFS_LOCK_GIANT() and related macros. Stop handling buffers belonging to non-mpsafe filesystems. The VFS_VERSION is bumped to indicate the interface change which does not result in the interface signatures changes. Conducted and reviewed by: attilio Tested by: pho
2012-10-22 17:50:54 +00:00
error = VFS_FHTOVP(mp, &fhp->fh_fid, lktype, vpp);
Patch the experimental NFS server in a manner analagous to r205661 for the regular NFS server, to ensure that ESTALE is returned to the client for all errors returned by VFS_FHTOVP(). MFC after: 2 weeks
2010-03-26 01:35:19 +00:00
if (error != 0)
/* Make sure the server replies ESTALE to the client. */
error = ESTALE;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 (nam && !error) {
error = VFS_CHECKEXP(mp, nam, &exp->nes_exflag, credp,
Fix export_args ex_flags field so that is 64bits, the same as mnt_flags. Since mnt_flags was upgraded to 64bits there has been a quirk in "struct export_args", since it hold a copy of mnt_flags in ex_flags, which is an "int" (32bits). This happens to currently work, since all the flag bits used in ex_flags are defined in the low order 32bits. However, new export flags cannot be defined. Also, ex_anon is a "struct xucred", which limits it to 16 additional groups. This patch revises "struct export_args" to make ex_flags 64bits and replaces ex_anon with ex_uid, ex_ngroups and ex_groups (which points to a groups list, so it can be malloc'd up to NGROUPS in size. This requires that the VFS_CHECKEXP() arguments change, so I also modified the last "secflavors" argument to be an array pointer, so that the secflavors could be copied in VFS_CHECKEXP() while the export entry is locked. (Without this patch VFS_CHECKEXP() returns a pointer to the secflavors array and then it is used after being unlocked, which is potentially a problem if the exports entry is changed. In practice this does not occur when mountd is run with "-S", but I think it is worth fixing.) This patch also deleted the vfs_oexport_conv() function, since do_mount_update() does the conversion, as required by the old vfs_cmount() calls. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:10:18 +00:00
&exp->nes_numsecflavor, exp->nes_secflavors);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 (nfs_rootfhset) {
exp->nes_exflag = 0;
Apply changes to the experimental nfs server so that it uses the security flavors as exported in FreeBSD-CURRENT. This allows it to use a slightly modified mountd.c instead of a different utility. Approved by: kib (mentor)
2009-05-14 21:39:08 +00:00
exp->nes_numsecflavor = 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 = 0;
} else {
vput(*vpp);
}
Add a sanity check for nes_numsecflavor to the NFS server. Ryan Moeller reported crashes in the NFS server that appear to be caused by stack corruption in nfsrv_compound(). It appears that the stack got corrupted just after a NFSv4.1 Lookup that crosses a server mount point. Although it is just a "theory" at this point, the most obvious way the stack could get corrupted would be if nfsvno_checkexp() somehow acquires an export with a bogus nes_numsecflavor value. This would cause the copying of the secflavors to run off the end of the array, which is allocated on the stack below where the corruption occurs. This sanity check is simple to do and would stop the stack corruption if the theory is correct. Otherwise, doing the sanity check seems to be a reasonable safety belt to add to the code. Reported by: freqlabs MFC after: 2 weeks
2020-04-17 02:21:46 +00:00
} else if (exp->nes_numsecflavor < 1 || exp->nes_numsecflavor >
MAXSECFLAVORS) {
printf("nfsvno_fhtovp: numsecflavors out of range\n");
exp->nes_numsecflavor = 0;
error = EACCES;
vput(*vpp);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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
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);
}
/*
* nfsd_fhtovp() - convert a fh to a vnode ptr
* - look up fsid in mount list (if not found ret error)
* - get vp and export rights by calling nfsvno_fhtovp()
* - if cred->cr_uid == 0 or MNT_EXPORTANON set it to credanon
* for AUTH_SYS
Since the VFS_LOCK_GIANT() code in the experimental NFS server is broken and the major file systems are now all mpsafe, modify the server so that it will only export mpsafe file systems. This was discussed on freebsd-fs@ and removes a fair bit of crufty code. MFC after: 12 days
2011-01-06 19:50:11 +00:00
* - if mpp != NULL, return the mount point so that it can
* be used for vn_finished_write() by the caller
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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
Modify the experimental NFS server so that it uses LK_SHARED for RPC operations when it can. Since VFS_FHTOVP() currently always gets an exclusively locked vnode and is usually called at the beginning of each RPC, the RPCs for a given vnode will still be serialized. As such, passing a lock type argument to VFS_FHTOVP() would be preferable to doing the vn_lock() with LK_DOWNGRADE after the VFS_FHTOVP() call. Reviewed by: kib MFC after: 2 weeks
2010-12-25 21:56:25 +00:00
nfsd_fhtovp(struct nfsrv_descript *nd, struct nfsrvfh *nfp, int lktype,
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 vnode **vpp, struct nfsexstuff *exp,
nfsd: Fix when NFSERR_WRONGSEC may be replied to NFSv4 clients Commit d224f05fcfc1 pre-parsed the next operation number for the put file handle operations. This patch uses this next operation number, plus the type of the file handle being set by the put file handle operation, to implement the rules in RFC5661 Sec. 2.6 with respect to replying NFSERR_WRONGSEC. This patch also adds a check to see if NFSERR_WRONGSEC should be replied when about to perform Lookup, Lookupp or Open with a file name component, so that the NFSERR_WRONGSEC reply is done for these operations, as required by RFC5661 Sec. 2.6. This patch does not have any practical effect for the FreeBSD NFSv4 client and I believe that the same is true for the Linux client, since NFSERR_WRONGSEC is considered a fatal error at this time. MFC after: 2 weeks
2021-06-05 23:53:07 +00:00
struct mount **mpp, int startwrite, int nextop)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 server: improve use of the VFS KPI In particular, do not assume that vn_start_write() returns the same mp as it was passed in, or never returns error. Also be more accurate to return NULL vp and mp when error occured, to catch wrong control flow easier. Stop checking for NULL mp before calling vn_finished_write(), NULL mp is handled transparently by the function. Reviewed by: rmacklem Tested by: pho MFC after: 1 week Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D27881
2021-01-01 15:35:44 +00:00
struct mount *mp, *mpw;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 *credanon;
fhandle_t *fhp;
nfs server: improve use of the VFS KPI In particular, do not assume that vn_start_write() returns the same mp as it was passed in, or never returns error. Also be more accurate to return NULL vp and mp when error occured, to catch wrong control flow easier. Stop checking for NULL mp before calling vn_finished_write(), NULL mp is handled transparently by the function. Reviewed by: rmacklem Tested by: pho MFC after: 1 week Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D27881
2021-01-01 15:35:44 +00:00
int 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
nfs server: improve use of the VFS KPI In particular, do not assume that vn_start_write() returns the same mp as it was passed in, or never returns error. Also be more accurate to return NULL vp and mp when error occured, to catch wrong control flow easier. Stop checking for NULL mp before calling vn_finished_write(), NULL mp is handled transparently by the function. Reviewed by: rmacklem Tested by: pho MFC after: 1 week Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D27881
2021-01-01 15:35:44 +00:00
if (mpp != NULL)
*mpp = NULL;
*vpp = 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 = (fhandle_t *)nfp->nfsrvfh_data;
Fix the experimental NFS server to use vfs_busyfs() instead of vfs_getvfs() so that the mount point is busied for the VFS_FHTOVP() call. This is analagous to r185432 for the regular NFS server. Reviewed by: kib MFC after: 12 days
2011-01-05 18:46:05 +00:00
mp = vfs_busyfs(&fhp->fh_fsid);
if (mp == 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
nd->nd_repstat = ESTALE;
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
}
Revamp the old NFS server's File Handle Affinity (FHA) code so that it will work with either the old or new server. The FHA code keeps a cache of currently active file handles for NFSv2 and v3 requests, so that read and write requests for the same file are directed to the same group of threads (reads) or thread (writes). It does not currently work for NFSv4 requests. They are more complex, and will take more work to support. This improves read-ahead performance, especially with ZFS, if the FHA tuning parameters are configured appropriately. Without the FHA code, concurrent reads that are part of a sequential read from a file will be directed to separate NFS threads. This has the effect of confusing the ZFS zfetch (prefetch) code and makes sequential reads significantly slower with clients like Linux that do a lot of prefetching. The FHA code has also been updated to direct write requests to nearby file offsets to the same thread in the same way it batches reads, and the FHA code will now also send writes to multiple threads when needed. This improves sequential write performance in ZFS, because writes to a file are now more ordered. Since NFS writes (generally less than 64K) are smaller than the typical ZFS record size (usually 128K), out of order NFS writes to the same block can trigger a read in ZFS. Sending them down the same thread increases the odds of their being in order. In order for multiple write threads per file in the FHA code to be useful, writes in the NFS server have been changed to use a LK_SHARED vnode lock, and upgrade that to LK_EXCLUSIVE if the filesystem doesn't allow multiple writers to a file at once. ZFS is currently the only filesystem that allows multiple writers to a file, because it has internal file range locking. This change does not affect the NFSv4 code. This improves random write performance to a single file in ZFS, since we can now have multiple writers inside ZFS at one time. I have changed the default tuning parameters to a 22 bit (4MB) window size (from 256K) and unlimited commands per thread as a result of my benchmarking with ZFS. The FHA code has been updated to allow configuring the tuning parameters from loader tunable variables in addition to sysctl variables. The read offset window calculation has been slightly modified as well. Instead of having separate bins, each file handle has a rolling window of bin_shift size. This minimizes glitches in throughput when shifting from one bin to another. sys/conf/files: Add nfs_fha_new.c and nfs_fha_old.c. Compile nfs_fha.c when either the old or the new NFS server is built. sys/fs/nfs/nfsport.h, sys/fs/nfs/nfs_commonport.c: Bring in changes from Rick Macklem to newnfs_realign that allow it to operate in blocking (M_WAITOK) or non-blocking (M_NOWAIT) mode. sys/fs/nfs/nfs_commonsubs.c, sys/fs/nfs/nfs_var.h: Bring in a change from Rick Macklem to allow telling nfsm_dissect() whether or not to wait for mallocs. sys/fs/nfs/nfsm_subs.h: Bring in changes from Rick Macklem to create a new nfsm_dissect_nonblock() inline function and NFSM_DISSECT_NONBLOCK() macro. sys/fs/nfs/nfs_commonkrpc.c, sys/fs/nfsclient/nfs_clkrpc.c: Add the malloc wait flag to a newnfs_realign() call. sys/fs/nfsserver/nfs_nfsdkrpc.c: Setup the new NFS server's RPC thread pool so that it will call the FHA code. Add the malloc flag argument to newnfs_realign(). Unstaticize newnfs_nfsv3_procid[] so that we can use it in the FHA code. sys/fs/nfsserver/nfs_nfsdsocket.c: In nfsrvd_dorpc(), add NFSPROC_WRITE to the list of RPC types that use the LK_SHARED lock type. sys/fs/nfsserver/nfs_nfsdport.c: In nfsd_fhtovp(), if we're starting a write, check to see whether the underlying filesystem supports shared writes. If not, upgrade the lock type from LK_SHARED to LK_EXCLUSIVE. sys/nfsserver/nfs_fha.c: Remove all code that is specific to the NFS server implementation. Anything that is server-specific is now accessed through a callback supplied by that server's FHA shim in the new softc. There are now separate sysctls and tunables for the FHA implementations for the old and new NFS servers. The new NFS server has its tunables under vfs.nfsd.fha, the old NFS server's tunables are under vfs.nfsrv.fha as before. In fha_extract_info(), use callouts for all server-specific code. Getting file handles and offsets is now done in the individual server's shim module. In fha_hash_entry_choose_thread(), change the way we decide whether two reads are in proximity to each other. Previously, the calculation was a simple shift operation to see whether the offsets were in the same power of 2 bucket. The issue was that there would be a bucket (and therefore thread) transition, even if the reads were in close proximity. When there is a thread transition, reads wind up going somewhat out of order, and ZFS gets confused. The new calculation simply tries to see whether the offsets are within 1 << bin_shift of each other. If they are, the reads will be sent to the same thread. The effect of this change is that for sequential reads, if the client doesn't exceed the max_reqs_per_nfsd parameter and the bin_shift is set to a reasonable value (22, or 4MB works well in my tests), the reads in any sequential stream will largely be confined to a single thread. Change fha_assign() so that it takes a softc argument. It is now called from the individual server's shim code, which will pass in the softc. Change fhe_stats_sysctl() so that it takes a softc parameter. It is now called from the individual server's shim code. Add the current offset to the list of things printed out about each active thread. Change the num_reads and num_writes counters in the fha_hash_entry structure to 32-bit values, and rename them num_rw and num_exclusive, respectively, to reflect their changed usage. Add an enable sysctl and tunable that allows the user to disable the FHA code (when vfs.XXX.fha.enable = 0). This is useful for before/after performance comparisons. nfs_fha.h: Move most structure definitions out of nfs_fha.c and into the header file, so that the individual server shims can see them. Change the default bin_shift to 22 (4MB) instead of 18 (256K). Allow unlimited commands per thread. sys/nfsserver/nfs_fha_old.c, sys/nfsserver/nfs_fha_old.h, sys/fs/nfsserver/nfs_fha_new.c, sys/fs/nfsserver/nfs_fha_new.h: Add shims for the old and new NFS servers to interface with the FHA code, and callbacks for the The shims contain all of the code and definitions that are specific to the NFS servers. They setup the server-specific callbacks and set the server name for the sysctl and loader tunable variables. sys/nfsserver/nfs_srvkrpc.c: Configure the RPC code to call fhaold_assign() instead of fha_assign(). sys/modules/nfsd/Makefile: Add nfs_fha.c and nfs_fha_new.c. sys/modules/nfsserver/Makefile: Add nfs_fha_old.c. Reviewed by: rmacklem Sponsored by: Spectra Logic MFC after: 2 weeks
2013-04-17 21:00:22 +00:00
if (startwrite) {
nfs server: improve use of the VFS KPI In particular, do not assume that vn_start_write() returns the same mp as it was passed in, or never returns error. Also be more accurate to return NULL vp and mp when error occured, to catch wrong control flow easier. Stop checking for NULL mp before calling vn_finished_write(), NULL mp is handled transparently by the function. Reviewed by: rmacklem Tested by: pho MFC after: 1 week Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D27881
2021-01-01 15:35:44 +00:00
mpw = mp;
error = vn_start_write(NULL, &mpw, V_WAIT);
if (error != 0) {
mpw = NULL;
vfs_unbusy(mp);
nd->nd_repstat = ESTALE;
goto out;
}
Revamp the old NFS server's File Handle Affinity (FHA) code so that it will work with either the old or new server. The FHA code keeps a cache of currently active file handles for NFSv2 and v3 requests, so that read and write requests for the same file are directed to the same group of threads (reads) or thread (writes). It does not currently work for NFSv4 requests. They are more complex, and will take more work to support. This improves read-ahead performance, especially with ZFS, if the FHA tuning parameters are configured appropriately. Without the FHA code, concurrent reads that are part of a sequential read from a file will be directed to separate NFS threads. This has the effect of confusing the ZFS zfetch (prefetch) code and makes sequential reads significantly slower with clients like Linux that do a lot of prefetching. The FHA code has also been updated to direct write requests to nearby file offsets to the same thread in the same way it batches reads, and the FHA code will now also send writes to multiple threads when needed. This improves sequential write performance in ZFS, because writes to a file are now more ordered. Since NFS writes (generally less than 64K) are smaller than the typical ZFS record size (usually 128K), out of order NFS writes to the same block can trigger a read in ZFS. Sending them down the same thread increases the odds of their being in order. In order for multiple write threads per file in the FHA code to be useful, writes in the NFS server have been changed to use a LK_SHARED vnode lock, and upgrade that to LK_EXCLUSIVE if the filesystem doesn't allow multiple writers to a file at once. ZFS is currently the only filesystem that allows multiple writers to a file, because it has internal file range locking. This change does not affect the NFSv4 code. This improves random write performance to a single file in ZFS, since we can now have multiple writers inside ZFS at one time. I have changed the default tuning parameters to a 22 bit (4MB) window size (from 256K) and unlimited commands per thread as a result of my benchmarking with ZFS. The FHA code has been updated to allow configuring the tuning parameters from loader tunable variables in addition to sysctl variables. The read offset window calculation has been slightly modified as well. Instead of having separate bins, each file handle has a rolling window of bin_shift size. This minimizes glitches in throughput when shifting from one bin to another. sys/conf/files: Add nfs_fha_new.c and nfs_fha_old.c. Compile nfs_fha.c when either the old or the new NFS server is built. sys/fs/nfs/nfsport.h, sys/fs/nfs/nfs_commonport.c: Bring in changes from Rick Macklem to newnfs_realign that allow it to operate in blocking (M_WAITOK) or non-blocking (M_NOWAIT) mode. sys/fs/nfs/nfs_commonsubs.c, sys/fs/nfs/nfs_var.h: Bring in a change from Rick Macklem to allow telling nfsm_dissect() whether or not to wait for mallocs. sys/fs/nfs/nfsm_subs.h: Bring in changes from Rick Macklem to create a new nfsm_dissect_nonblock() inline function and NFSM_DISSECT_NONBLOCK() macro. sys/fs/nfs/nfs_commonkrpc.c, sys/fs/nfsclient/nfs_clkrpc.c: Add the malloc wait flag to a newnfs_realign() call. sys/fs/nfsserver/nfs_nfsdkrpc.c: Setup the new NFS server's RPC thread pool so that it will call the FHA code. Add the malloc flag argument to newnfs_realign(). Unstaticize newnfs_nfsv3_procid[] so that we can use it in the FHA code. sys/fs/nfsserver/nfs_nfsdsocket.c: In nfsrvd_dorpc(), add NFSPROC_WRITE to the list of RPC types that use the LK_SHARED lock type. sys/fs/nfsserver/nfs_nfsdport.c: In nfsd_fhtovp(), if we're starting a write, check to see whether the underlying filesystem supports shared writes. If not, upgrade the lock type from LK_SHARED to LK_EXCLUSIVE. sys/nfsserver/nfs_fha.c: Remove all code that is specific to the NFS server implementation. Anything that is server-specific is now accessed through a callback supplied by that server's FHA shim in the new softc. There are now separate sysctls and tunables for the FHA implementations for the old and new NFS servers. The new NFS server has its tunables under vfs.nfsd.fha, the old NFS server's tunables are under vfs.nfsrv.fha as before. In fha_extract_info(), use callouts for all server-specific code. Getting file handles and offsets is now done in the individual server's shim module. In fha_hash_entry_choose_thread(), change the way we decide whether two reads are in proximity to each other. Previously, the calculation was a simple shift operation to see whether the offsets were in the same power of 2 bucket. The issue was that there would be a bucket (and therefore thread) transition, even if the reads were in close proximity. When there is a thread transition, reads wind up going somewhat out of order, and ZFS gets confused. The new calculation simply tries to see whether the offsets are within 1 << bin_shift of each other. If they are, the reads will be sent to the same thread. The effect of this change is that for sequential reads, if the client doesn't exceed the max_reqs_per_nfsd parameter and the bin_shift is set to a reasonable value (22, or 4MB works well in my tests), the reads in any sequential stream will largely be confined to a single thread. Change fha_assign() so that it takes a softc argument. It is now called from the individual server's shim code, which will pass in the softc. Change fhe_stats_sysctl() so that it takes a softc parameter. It is now called from the individual server's shim code. Add the current offset to the list of things printed out about each active thread. Change the num_reads and num_writes counters in the fha_hash_entry structure to 32-bit values, and rename them num_rw and num_exclusive, respectively, to reflect their changed usage. Add an enable sysctl and tunable that allows the user to disable the FHA code (when vfs.XXX.fha.enable = 0). This is useful for before/after performance comparisons. nfs_fha.h: Move most structure definitions out of nfs_fha.c and into the header file, so that the individual server shims can see them. Change the default bin_shift to 22 (4MB) instead of 18 (256K). Allow unlimited commands per thread. sys/nfsserver/nfs_fha_old.c, sys/nfsserver/nfs_fha_old.h, sys/fs/nfsserver/nfs_fha_new.c, sys/fs/nfsserver/nfs_fha_new.h: Add shims for the old and new NFS servers to interface with the FHA code, and callbacks for the The shims contain all of the code and definitions that are specific to the NFS servers. They setup the server-specific callbacks and set the server name for the sysctl and loader tunable variables. sys/nfsserver/nfs_srvkrpc.c: Configure the RPC code to call fhaold_assign() instead of fha_assign(). sys/modules/nfsd/Makefile: Add nfs_fha.c and nfs_fha_new.c. sys/modules/nfsserver/Makefile: Add nfs_fha_old.c. Reviewed by: rmacklem Sponsored by: Spectra Logic MFC after: 2 weeks
2013-04-17 21:00:22 +00:00
if (lktype == LK_SHARED && !(MNT_SHARED_WRITES(mp)))
lktype = LK_EXCLUSIVE;
nfs server: improve use of the VFS KPI In particular, do not assume that vn_start_write() returns the same mp as it was passed in, or never returns error. Also be more accurate to return NULL vp and mp when error occured, to catch wrong control flow easier. Stop checking for NULL mp before calling vn_finished_write(), NULL mp is handled transparently by the function. Reviewed by: rmacklem Tested by: pho MFC after: 1 week Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D27881
2021-01-01 15:35:44 +00:00
} else
mpw = NULL;
Modify the experimental NFS server so that it uses LK_SHARED for RPC operations when it can. Since VFS_FHTOVP() currently always gets an exclusively locked vnode and is usually called at the beginning of each RPC, the RPCs for a given vnode will still be serialized. As such, passing a lock type argument to VFS_FHTOVP() would be preferable to doing the vn_lock() with LK_DOWNGRADE after the VFS_FHTOVP() call. Reviewed by: kib MFC after: 2 weeks
2010-12-25 21:56:25 +00:00
nd->nd_repstat = nfsvno_fhtovp(mp, fhp, nd->nd_nam, lktype, vpp, 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
&credanon);
Fix the experimental NFS server to use vfs_busyfs() instead of vfs_getvfs() so that the mount point is busied for the VFS_FHTOVP() call. This is analagous to r185432 for the regular NFS server. Reviewed by: kib MFC after: 12 days
2011-01-05 18:46:05 +00:00
vfs_unbusy(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
/*
* For NFSv4 without a pseudo root fs, unexported file handles
* can be returned, so that Lookup works everywhere.
*/
if (!nd->nd_repstat && exp->nes_exflag == 0 &&
!(nd->nd_flag & ND_NFSV4)) {
vput(*vpp);
nfs server: improve use of the VFS KPI In particular, do not assume that vn_start_write() returns the same mp as it was passed in, or never returns error. Also be more accurate to return NULL vp and mp when error occured, to catch wrong control flow easier. Stop checking for NULL mp before calling vn_finished_write(), NULL mp is handled transparently by the function. Reviewed by: rmacklem Tested by: pho MFC after: 1 week Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D27881
2021-01-01 15:35:44 +00:00
*vpp = 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
nd->nd_repstat = EACCES;
}
/*
* Personally, I've never seen any point in requiring a
* reserved port#, since only in the rare case where the
NFS: spelling fixes on comments. No funcional change.
2016-04-29 16:07:25 +00:00
* clients are all boxes with secure system privileges,
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
* does it provide any enhanced security, but... some people
* believe it to be useful and keep putting this code back in.
* (There is also some "security checker" out there that
* complains if the nfs server doesn't enforce this.)
* However, note the following:
* RFC3530 (NFSv4) specifies that a reserved port# not be
* required.
* RFC2623 recommends that, if a reserved port# is checked for,
* that there be a way to turn that off--> ifdef'd.
*/
#ifdef NFS_REQRSVPORT
if (!nd->nd_repstat) {
Apply changes to the experimental nfs server so that it uses the security flavors as exported in FreeBSD-CURRENT. This allows it to use a slightly modified mountd.c instead of a different utility. Approved by: kib (mentor)
2009-05-14 21:39:08 +00:00
struct sockaddr_in *saddr;
struct sockaddr_in6 *saddr6;
saddr = NFSSOCKADDR(nd->nd_nam, struct sockaddr_in *);
saddr6 = NFSSOCKADDR(nd->nd_nam, struct sockaddr_in6 *);
if (!(nd->nd_flag & ND_NFSV4) &&
((saddr->sin_family == AF_INET &&
ntohs(saddr->sin_port) >= IPPORT_RESERVED) ||
(saddr6->sin6_family == AF_INET6 &&
ntohs(saddr6->sin6_port) >= IPPORT_RESERVED))) {
vput(*vpp);
nd->nd_repstat = (NFSERR_AUTHERR | AUTH_TOOWEAK);
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 /* NFS_REQRSVPORT */
/*
* Check/setup credentials.
*/
if (!nd->nd_repstat) {
nd->nd_saveduid = nd->nd_cred->cr_uid;
nfsd: Fix when NFSERR_WRONGSEC may be replied to NFSv4 clients Commit d224f05fcfc1 pre-parsed the next operation number for the put file handle operations. This patch uses this next operation number, plus the type of the file handle being set by the put file handle operation, to implement the rules in RFC5661 Sec. 2.6 with respect to replying NFSERR_WRONGSEC. This patch also adds a check to see if NFSERR_WRONGSEC should be replied when about to perform Lookup, Lookupp or Open with a file name component, so that the NFSERR_WRONGSEC reply is done for these operations, as required by RFC5661 Sec. 2.6. This patch does not have any practical effect for the FreeBSD NFSv4 client and I believe that the same is true for the Linux client, since NFSERR_WRONGSEC is considered a fatal error at this time. MFC after: 2 weeks
2021-06-05 23:53:07 +00:00
nd->nd_repstat = nfsd_excred(nd, exp, credanon,
nfsrv_checkwrongsec(nd, nextop, (*vpp)->v_type));
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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_repstat)
vput(*vpp);
}
Do not embed struct ucred into larger netcred parent structures. Credential might need to hang around longer than its parent and be used outside of mnt_explock scope controlling netcred lifetime. Use separate reference-counted ucred allocated separately instead. While there, extend mnt_explock coverage in vfs_stdexpcheck and clean-up some unused declarations in new NFS code. Reported by: John Hickey PR: kern/133439 Reviewed by: dfr, kib
2009-05-09 18:09:17 +00:00
if (credanon != NULL)
crfree(credanon);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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_repstat) {
nfs server: improve use of the VFS KPI In particular, do not assume that vn_start_write() returns the same mp as it was passed in, or never returns error. Also be more accurate to return NULL vp and mp when error occured, to catch wrong control flow easier. Stop checking for NULL mp before calling vn_finished_write(), NULL mp is handled transparently by the function. Reviewed by: rmacklem Tested by: pho MFC after: 1 week Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D27881
2021-01-01 15:35:44 +00:00
vn_finished_write(mpw);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
*vpp = NULL;
nfs server: improve use of the VFS KPI In particular, do not assume that vn_start_write() returns the same mp as it was passed in, or never returns error. Also be more accurate to return NULL vp and mp when error occured, to catch wrong control flow easier. Stop checking for NULL mp before calling vn_finished_write(), NULL mp is handled transparently by the function. Reviewed by: rmacklem Tested by: pho MFC after: 1 week Sponsored by: The FreeBSD Foundation Differential Revision: https://reviews.freebsd.org/D27881
2021-01-01 15:35:44 +00:00
} else if (mpp != NULL) {
*mpp = mpw;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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(0, 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
}
/*
* glue for fp.
*/
Merge Capsicum overhaul: - Capability is no longer separate descriptor type. Now every descriptor has set of its own capability rights. - The cap_new(2) system call is left, but it is no longer documented and should not be used in new code. - The new syscall cap_rights_limit(2) should be used instead of cap_new(2), which limits capability rights of the given descriptor without creating a new one. - The cap_getrights(2) syscall is renamed to cap_rights_get(2). - If CAP_IOCTL capability right is present we can further reduce allowed ioctls list with the new cap_ioctls_limit(2) syscall. List of allowed ioctls can be retrived with cap_ioctls_get(2) syscall. - If CAP_FCNTL capability right is present we can further reduce fcntls that can be used with the new cap_fcntls_limit(2) syscall and retrive them with cap_fcntls_get(2). - To support ioctl and fcntl white-listing the filedesc structure was heavly modified. - The audit subsystem, kdump and procstat tools were updated to recognize new syscalls. - Capability rights were revised and eventhough I tried hard to provide backward API and ABI compatibility there are some incompatible changes that are described in detail below: CAP_CREATE old behaviour: - Allow for openat(2)+O_CREAT. - Allow for linkat(2). - Allow for symlinkat(2). CAP_CREATE new behaviour: - Allow for openat(2)+O_CREAT. Added CAP_LINKAT: - Allow for linkat(2). ABI: Reuses CAP_RMDIR bit. - Allow to be target for renameat(2). Added CAP_SYMLINKAT: - Allow for symlinkat(2). Removed CAP_DELETE. Old behaviour: - Allow for unlinkat(2) when removing non-directory object. - Allow to be source for renameat(2). Removed CAP_RMDIR. Old behaviour: - Allow for unlinkat(2) when removing directory. Added CAP_RENAMEAT: - Required for source directory for the renameat(2) syscall. Added CAP_UNLINKAT (effectively it replaces CAP_DELETE and CAP_RMDIR): - Allow for unlinkat(2) on any object. - Required if target of renameat(2) exists and will be removed by this call. Removed CAP_MAPEXEC. CAP_MMAP old behaviour: - Allow for mmap(2) with any combination of PROT_NONE, PROT_READ and PROT_WRITE. CAP_MMAP new behaviour: - Allow for mmap(2)+PROT_NONE. Added CAP_MMAP_R: - Allow for mmap(PROT_READ). Added CAP_MMAP_W: - Allow for mmap(PROT_WRITE). Added CAP_MMAP_X: - Allow for mmap(PROT_EXEC). Added CAP_MMAP_RW: - Allow for mmap(PROT_READ | PROT_WRITE). Added CAP_MMAP_RX: - Allow for mmap(PROT_READ | PROT_EXEC). Added CAP_MMAP_WX: - Allow for mmap(PROT_WRITE | PROT_EXEC). Added CAP_MMAP_RWX: - Allow for mmap(PROT_READ | PROT_WRITE | PROT_EXEC). Renamed CAP_MKDIR to CAP_MKDIRAT. Renamed CAP_MKFIFO to CAP_MKFIFOAT. Renamed CAP_MKNODE to CAP_MKNODEAT. CAP_READ old behaviour: - Allow pread(2). - Disallow read(2), readv(2) (if there is no CAP_SEEK). CAP_READ new behaviour: - Allow read(2), readv(2). - Disallow pread(2) (CAP_SEEK was also required). CAP_WRITE old behaviour: - Allow pwrite(2). - Disallow write(2), writev(2) (if there is no CAP_SEEK). CAP_WRITE new behaviour: - Allow write(2), writev(2). - Disallow pwrite(2) (CAP_SEEK was also required). Added convinient defines: #define CAP_PREAD (CAP_SEEK | CAP_READ) #define CAP_PWRITE (CAP_SEEK | CAP_WRITE) #define CAP_MMAP_R (CAP_MMAP | CAP_SEEK | CAP_READ) #define CAP_MMAP_W (CAP_MMAP | CAP_SEEK | CAP_WRITE) #define CAP_MMAP_X (CAP_MMAP | CAP_SEEK | 0x0000000000000008ULL) #define CAP_MMAP_RW (CAP_MMAP_R | CAP_MMAP_W) #define CAP_MMAP_RX (CAP_MMAP_R | CAP_MMAP_X) #define CAP_MMAP_WX (CAP_MMAP_W | CAP_MMAP_X) #define CAP_MMAP_RWX (CAP_MMAP_R | CAP_MMAP_W | CAP_MMAP_X) #define CAP_RECV CAP_READ #define CAP_SEND CAP_WRITE #define CAP_SOCK_CLIENT \ (CAP_CONNECT | CAP_GETPEERNAME | CAP_GETSOCKNAME | CAP_GETSOCKOPT | \ CAP_PEELOFF | CAP_RECV | CAP_SEND | CAP_SETSOCKOPT | CAP_SHUTDOWN) #define CAP_SOCK_SERVER \ (CAP_ACCEPT | CAP_BIND | CAP_GETPEERNAME | CAP_GETSOCKNAME | \ CAP_GETSOCKOPT | CAP_LISTEN | CAP_PEELOFF | CAP_RECV | CAP_SEND | \ CAP_SETSOCKOPT | CAP_SHUTDOWN) Added defines for backward API compatibility: #define CAP_MAPEXEC CAP_MMAP_X #define CAP_DELETE CAP_UNLINKAT #define CAP_MKDIR CAP_MKDIRAT #define CAP_RMDIR CAP_UNLINKAT #define CAP_MKFIFO CAP_MKFIFOAT #define CAP_MKNOD CAP_MKNODAT #define CAP_SOCK_ALL (CAP_SOCK_CLIENT | CAP_SOCK_SERVER) Sponsored by: The FreeBSD Foundation Reviewed by: Christoph Mallon <christoph.mallon@gmx.de> Many aspects discussed with: rwatson, benl, jonathan ABI compatibility discussed with: kib
2013-03-02 00:53:12 +00:00
static 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
fp_getfvp(struct thread *p, int fd, struct file **fpp, struct vnode **vpp)
{
struct filedesc *fdp;
struct file *fp;
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
fdp = p->td_proc->p_fd;
Merge Capsicum overhaul: - Capability is no longer separate descriptor type. Now every descriptor has set of its own capability rights. - The cap_new(2) system call is left, but it is no longer documented and should not be used in new code. - The new syscall cap_rights_limit(2) should be used instead of cap_new(2), which limits capability rights of the given descriptor without creating a new one. - The cap_getrights(2) syscall is renamed to cap_rights_get(2). - If CAP_IOCTL capability right is present we can further reduce allowed ioctls list with the new cap_ioctls_limit(2) syscall. List of allowed ioctls can be retrived with cap_ioctls_get(2) syscall. - If CAP_FCNTL capability right is present we can further reduce fcntls that can be used with the new cap_fcntls_limit(2) syscall and retrive them with cap_fcntls_get(2). - To support ioctl and fcntl white-listing the filedesc structure was heavly modified. - The audit subsystem, kdump and procstat tools were updated to recognize new syscalls. - Capability rights were revised and eventhough I tried hard to provide backward API and ABI compatibility there are some incompatible changes that are described in detail below: CAP_CREATE old behaviour: - Allow for openat(2)+O_CREAT. - Allow for linkat(2). - Allow for symlinkat(2). CAP_CREATE new behaviour: - Allow for openat(2)+O_CREAT. Added CAP_LINKAT: - Allow for linkat(2). ABI: Reuses CAP_RMDIR bit. - Allow to be target for renameat(2). Added CAP_SYMLINKAT: - Allow for symlinkat(2). Removed CAP_DELETE. Old behaviour: - Allow for unlinkat(2) when removing non-directory object. - Allow to be source for renameat(2). Removed CAP_RMDIR. Old behaviour: - Allow for unlinkat(2) when removing directory. Added CAP_RENAMEAT: - Required for source directory for the renameat(2) syscall. Added CAP_UNLINKAT (effectively it replaces CAP_DELETE and CAP_RMDIR): - Allow for unlinkat(2) on any object. - Required if target of renameat(2) exists and will be removed by this call. Removed CAP_MAPEXEC. CAP_MMAP old behaviour: - Allow for mmap(2) with any combination of PROT_NONE, PROT_READ and PROT_WRITE. CAP_MMAP new behaviour: - Allow for mmap(2)+PROT_NONE. Added CAP_MMAP_R: - Allow for mmap(PROT_READ). Added CAP_MMAP_W: - Allow for mmap(PROT_WRITE). Added CAP_MMAP_X: - Allow for mmap(PROT_EXEC). Added CAP_MMAP_RW: - Allow for mmap(PROT_READ | PROT_WRITE). Added CAP_MMAP_RX: - Allow for mmap(PROT_READ | PROT_EXEC). Added CAP_MMAP_WX: - Allow for mmap(PROT_WRITE | PROT_EXEC). Added CAP_MMAP_RWX: - Allow for mmap(PROT_READ | PROT_WRITE | PROT_EXEC). Renamed CAP_MKDIR to CAP_MKDIRAT. Renamed CAP_MKFIFO to CAP_MKFIFOAT. Renamed CAP_MKNODE to CAP_MKNODEAT. CAP_READ old behaviour: - Allow pread(2). - Disallow read(2), readv(2) (if there is no CAP_SEEK). CAP_READ new behaviour: - Allow read(2), readv(2). - Disallow pread(2) (CAP_SEEK was also required). CAP_WRITE old behaviour: - Allow pwrite(2). - Disallow write(2), writev(2) (if there is no CAP_SEEK). CAP_WRITE new behaviour: - Allow write(2), writev(2). - Disallow pwrite(2) (CAP_SEEK was also required). Added convinient defines: #define CAP_PREAD (CAP_SEEK | CAP_READ) #define CAP_PWRITE (CAP_SEEK | CAP_WRITE) #define CAP_MMAP_R (CAP_MMAP | CAP_SEEK | CAP_READ) #define CAP_MMAP_W (CAP_MMAP | CAP_SEEK | CAP_WRITE) #define CAP_MMAP_X (CAP_MMAP | CAP_SEEK | 0x0000000000000008ULL) #define CAP_MMAP_RW (CAP_MMAP_R | CAP_MMAP_W) #define CAP_MMAP_RX (CAP_MMAP_R | CAP_MMAP_X) #define CAP_MMAP_WX (CAP_MMAP_W | CAP_MMAP_X) #define CAP_MMAP_RWX (CAP_MMAP_R | CAP_MMAP_W | CAP_MMAP_X) #define CAP_RECV CAP_READ #define CAP_SEND CAP_WRITE #define CAP_SOCK_CLIENT \ (CAP_CONNECT | CAP_GETPEERNAME | CAP_GETSOCKNAME | CAP_GETSOCKOPT | \ CAP_PEELOFF | CAP_RECV | CAP_SEND | CAP_SETSOCKOPT | CAP_SHUTDOWN) #define CAP_SOCK_SERVER \ (CAP_ACCEPT | CAP_BIND | CAP_GETPEERNAME | CAP_GETSOCKNAME | \ CAP_GETSOCKOPT | CAP_LISTEN | CAP_PEELOFF | CAP_RECV | CAP_SEND | \ CAP_SETSOCKOPT | CAP_SHUTDOWN) Added defines for backward API compatibility: #define CAP_MAPEXEC CAP_MMAP_X #define CAP_DELETE CAP_UNLINKAT #define CAP_MKDIR CAP_MKDIRAT #define CAP_RMDIR CAP_UNLINKAT #define CAP_MKFIFO CAP_MKFIFOAT #define CAP_MKNOD CAP_MKNODAT #define CAP_SOCK_ALL (CAP_SOCK_CLIENT | CAP_SOCK_SERVER) Sponsored by: The FreeBSD Foundation Reviewed by: Christoph Mallon <christoph.mallon@gmx.de> Many aspects discussed with: rwatson, benl, jonathan ABI compatibility discussed with: kib
2013-03-02 00:53:12 +00:00
if (fd < 0 || fd >= fdp->fd_nfiles ||
(fp = fdp->fd_ofiles[fd].fde_file) == 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
error = EBADF;
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
*fpp = fp;
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
}
/*
Apply changes to the experimental nfs server so that it uses the security flavors as exported in FreeBSD-CURRENT. This allows it to use a slightly modified mountd.c instead of a different utility. Approved by: kib (mentor)
2009-05-14 21:39:08 +00:00
* Called from nfssvc() to update the exports list. Just 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
* vfs_export(). This has to be done, since the v4 root fake fs isn't
* in the mount list.
*/
int
nfsrv_v4rootexport(void *argp, struct ucred *cred, struct thread *p)
{
struct nfsex_args *nfsexargp = (struct nfsex_args *)argp;
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
struct nameidata nd;
fhandle_t fh;
error = vfs_export(&nfsv4root_mnt, &nfsexargp->export);
Delete some cruft from the experimental NFS server that was only used by the OpenBSD port for its pseudo-fs. MFC after: 2 weeks
2011-01-02 21:34:01 +00:00
if ((nfsexargp->export.ex_flags & MNT_DELEXPORT) != 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
nfs_rootfhset = 0;
Delete some cruft from the experimental NFS server that was only used by the OpenBSD port for its pseudo-fs. MFC after: 2 weeks
2011-01-02 21:34:01 +00:00
else if (error == 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 (nfsexargp->fspec == NULL) {
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
/*
* If fspec != NULL, this is the v4root path.
*/
Remove the support for using non-mpsafe filesystem modules. In particular, do not lock Giant conditionally when calling into the filesystem module, remove the VFS_LOCK_GIANT() and related macros. Stop handling buffers belonging to non-mpsafe filesystems. The VFS_VERSION is bumped to indicate the interface change which does not result in the interface signatures changes. Conducted and reviewed by: attilio Tested by: pho
2012-10-22 17:50:54 +00:00
NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE,
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsexargp->fspec, p);
if ((error = namei(&nd)) != 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
error = nfsvno_getfh(nd.ni_vp, &fh, p);
vrele(nd.ni_vp);
if (!error) {
nfs_rootfh.nfsrvfh_len = NFSX_MYFH;
NFSBCOPY((caddr_t)&fh,
nfs_rootfh.nfsrvfh_data,
sizeof (fhandle_t));
nfs_rootfhset = 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
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);
}
/*
* This function needs to test to see if the system is near its limit
* for memory allocation via malloc() or mget() and return True iff
* either of these resources are near their limit.
* XXX (For now, this is just a stub.)
*/
int nfsrv_testmalloclimit = 0;
int
nfsrv_mallocmget_limit(void)
{
static int printmesg = 0;
static int testval = 1;
if (nfsrv_testmalloclimit && (testval++ % 1000) == 0) {
if ((printmesg++ % 100) == 0)
printf("nfsd: malloc/mget near limit\n");
return (1);
}
return (0);
}
/*
* BSD specific initialization of a mount point.
*/
void
nfsd_mntinit(void)
{
static int inited = 0;
if (inited)
return;
inited = 1;
nfsv4root_mnt.mnt_flag = (MNT_RDONLY | MNT_EXPORTED);
TAILQ_INIT(&nfsv4root_mnt.mnt_nvnodelist);
vfs: add per-mount vnode lazy list and use it for deferred inactive + msync This obviates the need to scan the entire active list looking for vnodes of interest. msync is handled by adding all vnodes with write count to the lazy list. deferred inactive directly adds vnodes as it sets the VI_DEFINACT flag. Vnodes get dequeued from the list when their hold count reaches 0. Newly added MNT_VNODE_FOREACH_LAZY* macros support filtering so that spurious locking is avoided in the common case. Reviewed by: jeff Tested by: pho (in a larger patch, previous version) Differential Revision: https://reviews.freebsd.org/D22995
2020-01-13 02:34:02 +00:00
TAILQ_INIT(&nfsv4root_mnt.mnt_lazyvnodelist);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
nfsv4root_mnt.mnt_export = NULL;
TAILQ_INIT(&nfsv4root_opt);
TAILQ_INIT(&nfsv4root_newopt);
nfsv4root_mnt.mnt_opt = &nfsv4root_opt;
nfsv4root_mnt.mnt_optnew = &nfsv4root_newopt;
nfsv4root_mnt.mnt_nvnodelistsize = 0;
vfs: add per-mount vnode lazy list and use it for deferred inactive + msync This obviates the need to scan the entire active list looking for vnodes of interest. msync is handled by adding all vnodes with write count to the lazy list. deferred inactive directly adds vnodes as it sets the VI_DEFINACT flag. Vnodes get dequeued from the list when their hold count reaches 0. Newly added MNT_VNODE_FOREACH_LAZY* macros support filtering so that spurious locking is avoided in the common case. Reviewed by: jeff Tested by: pho (in a larger patch, previous version) Differential Revision: https://reviews.freebsd.org/D22995
2020-01-13 02:34:02 +00:00
nfsv4root_mnt.mnt_lazyvnodelistsize = 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
}
/*
* Get a vnode for a file handle, without checking exports, etc.
*/
struct vnode *
nfsvno_getvp(fhandle_t *fhp)
{
struct mount *mp;
struct vnode *vp;
int error;
Fix the experimental NFS server to use vfs_busyfs() instead of vfs_getvfs() so that the mount point is busied for the VFS_FHTOVP() call. This is analagous to r185432 for the regular NFS server. Reviewed by: kib MFC after: 12 days
2011-01-05 18:46:05 +00:00
mp = vfs_busyfs(&fhp->fh_fsid);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 (mp == NULL)
return (NULL);
Add a lock flags argument to the VFS_FHTOVP() file system method, so that callers can indicate the minimum vnode locking requirement. This will allow some file systems to choose to return a LK_SHARED locked vnode when LK_SHARED is specified for the flags argument. This patch only adds the flag. It does not change any file system to use it and all callers specify LK_EXCLUSIVE, so file system semantics are not changed. Reviewed by: kib
2011-05-22 01:07:54 +00:00
error = VFS_FHTOVP(mp, &fhp->fh_fid, LK_EXCLUSIVE, &vp);
Fix the experimental NFS server to use vfs_busyfs() instead of vfs_getvfs() so that the mount point is busied for the VFS_FHTOVP() call. This is analagous to r185432 for the regular NFS server. Reviewed by: kib MFC after: 12 days
2011-01-05 18:46:05 +00:00
vfs_unbusy(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)
return (NULL);
return (vp);
}
/*
* Do a local VOP_ADVLOCK().
*/
int
nfsvno_advlock(struct vnode *vp, int ftype, u_int64_t first,
Fix the acquisition of local locks via VOP_ADVLOCK() by the experimental nfsv4 server. It was setting the a_id argument to a fixed value, but that wasn't sufficient for FreeBSD8. Instead, set l_pid and l_sysid to 0 plus set the F_REMOTE flag to indicate that these fields are used to check for same lock owner. Since, for NFSv4, a lockowner is a ClientID plus an up to 1024byte name, it can't be put in l_sysid easily. I also renamed the p variable to td, since it's a thread ptr. Approved by: kib (mentor)
2009-05-17 19:33:48 +00:00
u_int64_t end, struct thread *td)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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
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
struct flock fl;
u_int64_t tlen;
Fix the experimental NFSv4 server so that it performs local VOP_ADVLOCK() unlock operations correctly. It was passing in F_SETLK instead of F_UNLCK as the operation for the unlock case. This only affected operation when local locking (vfs.newnfs.enable_locallocks=1) was enabled. MFC after: 1 week
2010-09-19 01:05:19 +00:00
if (nfsrv_dolocallocks == 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;
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_UNLOCKED(vp, "nfsvno_advlock: vp locked");
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 the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
fl.l_whence = SEEK_SET;
fl.l_type = ftype;
fl.l_start = (off_t)first;
if (end == NFS64BITSSET) {
fl.l_len = 0;
} else {
tlen = end - first;
fl.l_len = (off_t)tlen;
}
/*
Fix the acquisition of local locks via VOP_ADVLOCK() by the experimental nfsv4 server. It was setting the a_id argument to a fixed value, but that wasn't sufficient for FreeBSD8. Instead, set l_pid and l_sysid to 0 plus set the F_REMOTE flag to indicate that these fields are used to check for same lock owner. Since, for NFSv4, a lockowner is a ClientID plus an up to 1024byte name, it can't be put in l_sysid easily. I also renamed the p variable to td, since it's a thread ptr. Approved by: kib (mentor)
2009-05-17 19:33:48 +00:00
* For FreeBSD8, the l_pid and l_sysid must be set to the same
* values for all calls, so that all locks will be held by the
* nfsd server. (The nfsd server handles conflicts between the
* various clients.)
* Since an NFSv4 lockowner is a ClientID plus an array of up to 1024
* bytes, so it can't be put in l_sysid.
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 sys/fs/nfsserver/nfs_nfsdport.c to use nlm_acquire_next_sysid() to set the l_sysid for locks correctly. Approved by: kib (mentor)
2009-05-21 01:50:27 +00:00
if (nfsv4_sysid == 0)
nfsv4_sysid = nlm_acquire_next_sysid();
Fix the acquisition of local locks via VOP_ADVLOCK() by the experimental nfsv4 server. It was setting the a_id argument to a fixed value, but that wasn't sufficient for FreeBSD8. Instead, set l_pid and l_sysid to 0 plus set the F_REMOTE flag to indicate that these fields are used to check for same lock owner. Since, for NFSv4, a lockowner is a ClientID plus an up to 1024byte name, it can't be put in l_sysid easily. I also renamed the p variable to td, since it's a thread ptr. Approved by: kib (mentor)
2009-05-17 19:33:48 +00:00
fl.l_pid = (pid_t)0;
Modify sys/fs/nfsserver/nfs_nfsdport.c to use nlm_acquire_next_sysid() to set the l_sysid for locks correctly. Approved by: kib (mentor)
2009-05-21 01:50:27 +00:00
fl.l_sysid = (int)nfsv4_sysid;
Fix the acquisition of local locks via VOP_ADVLOCK() by the experimental nfsv4 server. It was setting the a_id argument to a fixed value, but that wasn't sufficient for FreeBSD8. Instead, set l_pid and l_sysid to 0 plus set the F_REMOTE flag to indicate that these fields are used to check for same lock owner. Since, for NFSv4, a lockowner is a ClientID plus an up to 1024byte name, it can't be put in l_sysid easily. I also renamed the p variable to td, since it's a thread ptr. Approved by: kib (mentor)
2009-05-17 19:33:48 +00:00
Fix the experimental NFSv4 server so that it performs local VOP_ADVLOCK() unlock operations correctly. It was passing in F_SETLK instead of F_UNLCK as the operation for the unlock case. This only affected operation when local locking (vfs.newnfs.enable_locallocks=1) was enabled. MFC after: 1 week
2010-09-19 01:05:19 +00:00
if (ftype == F_UNLCK)
error = VOP_ADVLOCK(vp, (caddr_t)td->td_proc, F_UNLCK, &fl,
(F_POSIX | F_REMOTE));
else
error = VOP_ADVLOCK(vp, (caddr_t)td->td_proc, F_SETLK, &fl,
(F_POSIX | F_REMOTE));
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);
}
/*
* Check the nfsv4 root exports.
*/
int
nfsvno_v4rootexport(struct nfsrv_descript *nd)
{
struct ucred *credanon;
Fix export_args ex_flags field so that is 64bits, the same as mnt_flags. Since mnt_flags was upgraded to 64bits there has been a quirk in "struct export_args", since it hold a copy of mnt_flags in ex_flags, which is an "int" (32bits). This happens to currently work, since all the flag bits used in ex_flags are defined in the low order 32bits. However, new export flags cannot be defined. Also, ex_anon is a "struct xucred", which limits it to 16 additional groups. This patch revises "struct export_args" to make ex_flags 64bits and replaces ex_anon with ex_uid, ex_ngroups and ex_groups (which points to a groups list, so it can be malloc'd up to NGROUPS in size. This requires that the VFS_CHECKEXP() arguments change, so I also modified the last "secflavors" argument to be an array pointer, so that the secflavors could be copied in VFS_CHECKEXP() while the export entry is locked. (Without this patch VFS_CHECKEXP() returns a pointer to the secflavors array and then it is used after being unlocked, which is potentially a problem if the exports entry is changed. In practice this does not occur when mountd is run with "-S", but I think it is worth fixing.) This patch also deleted the vfs_oexport_conv() function, since do_mount_update() does the conversion, as required by the old vfs_cmount() calls. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:10:18 +00:00
int error = 0, numsecflavor, secflavors[MAXSECFLAVORS], i;
uint64_t exflags;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 = vfs_stdcheckexp(&nfsv4root_mnt, nd->nd_nam, &exflags,
Fix export_args ex_flags field so that is 64bits, the same as mnt_flags. Since mnt_flags was upgraded to 64bits there has been a quirk in "struct export_args", since it hold a copy of mnt_flags in ex_flags, which is an "int" (32bits). This happens to currently work, since all the flag bits used in ex_flags are defined in the low order 32bits. However, new export flags cannot be defined. Also, ex_anon is a "struct xucred", which limits it to 16 additional groups. This patch revises "struct export_args" to make ex_flags 64bits and replaces ex_anon with ex_uid, ex_ngroups and ex_groups (which points to a groups list, so it can be malloc'd up to NGROUPS in size. This requires that the VFS_CHECKEXP() arguments change, so I also modified the last "secflavors" argument to be an array pointer, so that the secflavors could be copied in VFS_CHECKEXP() while the export entry is locked. (Without this patch VFS_CHECKEXP() returns a pointer to the secflavors array and then it is used after being unlocked, which is potentially a problem if the exports entry is changed. In practice this does not occur when mountd is run with "-S", but I think it is worth fixing.) This patch also deleted the vfs_oexport_conv() function, since do_mount_update() does the conversion, as required by the old vfs_cmount() calls. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:10:18 +00:00
&credanon, &numsecflavor, secflavors);
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 (error) {
error = NFSERR_PROGUNAVAIL;
goto out;
}
Do not embed struct ucred into larger netcred parent structures. Credential might need to hang around longer than its parent and be used outside of mnt_explock scope controlling netcred lifetime. Use separate reference-counted ucred allocated separately instead. While there, extend mnt_explock coverage in vfs_stdexpcheck and clean-up some unused declarations in new NFS code. Reported by: John Hickey PR: kern/133439 Reviewed by: dfr, kib
2009-05-09 18:09:17 +00:00
if (credanon != NULL)
crfree(credanon);
Apply changes to the experimental nfs server so that it uses the security flavors as exported in FreeBSD-CURRENT. This allows it to use a slightly modified mountd.c instead of a different utility. Approved by: kib (mentor)
2009-05-14 21:39:08 +00:00
for (i = 0; i < numsecflavor; i++) {
if (secflavors[i] == AUTH_SYS)
nd->nd_flag |= ND_EXAUTHSYS;
else if (secflavors[i] == RPCSEC_GSS_KRB5)
nd->nd_flag |= ND_EXGSS;
else if (secflavors[i] == RPCSEC_GSS_KRB5I)
nd->nd_flag |= ND_EXGSSINTEGRITY;
else if (secflavors[i] == RPCSEC_GSS_KRB5P)
nd->nd_flag |= ND_EXGSSPRIVACY;
}
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
Add flags to enable NFS over TLS to the NFS client and server. An Internet Draft titled "Towards Remote Procedure Call Encryption By Default" (soon to be an RFC I think) describes how Sun RPC is to use TLS with NFS as a specific application case. Various commits prepared the NFS code to use KERN_TLS, mainly enabling use of ext_pgs mbufs for large RPC messages. r364475 added TLS support to the kernel RPC. This commit (which is the final one for kernel changes required to do NFS over TLS) adds support for three export flags: MNT_EXTLS - Requires a TLS connection. MNT_EXTLSCERT - Requires a TLS connection where the client presents a valid X.509 certificate during TLS handshake. MNT_EXTLSCERTUSER - Requires a TLS connection where the client presents a valid X.509 certificate with "user@domain" in the otherName field of the SubjectAltName during TLS handshake. Without these export options, clients are permitted, but not required, to use TLS. For the client, a new nmount(2) option called "tls" makes the client do a STARTTLS Null RPC and TLS handshake for all TCP connections used for the mount. The CLSET_TLS client control option is used to indicate to the kernel RPC that this should be done. Unless the above export flags or "tls" option is used, semantics should not change for the NFS client nor server. For NFS over TLS to work, the userspace daemons rpctlscd(8) { for client } or rpctlssd(8) daemon { for server } must be running.
2020-08-27 23:57:30 +00:00
/* And set ND_EXxx flags for TLS. */
if ((exflags & MNT_EXTLS) != 0) {
nd->nd_flag |= ND_EXTLS;
if ((exflags & MNT_EXTLSCERT) != 0)
nd->nd_flag |= ND_EXTLSCERT;
if ((exflags & MNT_EXTLSCERTUSER) != 0)
nd->nd_flag |= ND_EXTLSCERTUSER;
}
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
}
/*
NFS: spelling fixes on comments. No funcional change.
2016-04-29 16:07:25 +00:00
* Nfs server pseudo system call for the nfsd's
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
*/
/*
* MPSAFE
*/
static int
nfssvc_nfsd(struct thread *td, struct nfssvc_args *uap)
{
struct file *fp;
Modify the experimental nfs server to use the new nfsd_nfsd_args structure for nfsd. Includes a change that clarifies the use of an empty principal name string to indicate AUTH_SYS only. Approved by: kib (mentor)
2009-05-12 16:04:51 +00:00
struct nfsd_addsock_args sockarg;
struct nfsd_nfsd_args nfsdarg;
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 nfsd_nfsd_oargs onfsdarg;
struct nfsd_pnfsd_args pnfsdarg;
struct vnode *vp, *nvp, *curdvp;
struct pnfsdsfile *pf;
struct nfsdevice *ds, *fds;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
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 buflen, error, ret;
char *buf, *cp, *cp2, *cp3;
char fname[PNFS_FILENAME_LEN + 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 (uap->flag & NFSSVC_NFSDADDSOCK) {
Modify the experimental nfs server to use the new nfsd_nfsd_args structure for nfsd. Includes a change that clarifies the use of an empty principal name string to indicate AUTH_SYS only. Approved by: kib (mentor)
2009-05-12 16:04:51 +00:00
error = copyin(uap->argp, (caddr_t)&sockarg, sizeof (sockarg));
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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)
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;
Second-to-last commit implementing Capsicum capabilities in the FreeBSD kernel for FreeBSD 9.0: Add a new capability mask argument to fget(9) and friends, allowing system call code to declare what capabilities are required when an integer file descriptor is converted into an in-kernel struct file *. With options CAPABILITIES compiled into the kernel, this enforces capability protection; without, this change is effectively a no-op. Some cases require special handling, such as mmap(2), which must preserve information about the maximum rights at the time of mapping in the memory map so that they can later be enforced in mprotect(2) -- this is done by narrowing the rights in the existing max_protection field used for similar purposes with file permissions. In namei(9), we assert that the code is not reached from within capability mode, as we're not yet ready to enforce namespace capabilities there. This will follow in a later commit. Update two capability names: CAP_EVENT and CAP_KEVENT become CAP_POST_KEVENT and CAP_POLL_KEVENT to more accurately indicate what they represent. Approved by: re (bz) Submitted by: jonathan Sponsored by: Google Inc
2011-08-11 12:30:23 +00:00
/*
* Since we don't know what rights might be required,
* pretend that we need them all. It is better to be too
* careful than too reckless.
*/
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
error = fget(td, sockarg.sock,
Convert remaining cap_rights_init users to cap_rights_init_one semantic patch: @@ expression rights, r; @@ - cap_rights_init(&rights, r) + cap_rights_init_one(&rights, r)
2021-01-12 12:11:26 +00:00
cap_rights_init_one(&rights, CAP_SOCK_SERVER), &fp);
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
if (error != 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
if (fp->f_type != DTYPE_SOCKET) {
fdrop(fp, td);
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
}
error = nfsrvd_addsock(fp);
fdrop(fp, td);
} else if (uap->flag & NFSSVC_NFSDNFSD) {
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 (uap->argp == NULL) {
error = EINVAL;
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
if ((uap->flag & NFSSVC_NEWSTRUCT) == 0) {
error = copyin(uap->argp, &onfsdarg, sizeof(onfsdarg));
if (error == 0) {
nfsdarg.principal = onfsdarg.principal;
nfsdarg.minthreads = onfsdarg.minthreads;
nfsdarg.maxthreads = onfsdarg.maxthreads;
nfsdarg.version = 1;
nfsdarg.addr = NULL;
nfsdarg.addrlen = 0;
nfsdarg.dnshost = NULL;
nfsdarg.dnshostlen = 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
nfsdarg.dspath = NULL;
nfsdarg.dspathlen = 0;
nfsdarg.mdspath = NULL;
nfsdarg.mdspathlen = 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
nfsdarg.mirrorcnt = 1;
}
} else
error = copyin(uap->argp, &nfsdarg, sizeof(nfsdarg));
Modify the experimental nfs server to use the new nfsd_nfsd_args structure for nfsd. Includes a change that clarifies the use of an empty principal name string to indicate AUTH_SYS only. Approved by: kib (mentor)
2009-05-12 16:04:51 +00:00
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;
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 (nfsdarg.addrlen > 0 && nfsdarg.addrlen < 10000 &&
nfsdarg.dnshostlen > 0 && nfsdarg.dnshostlen < 10000 &&
nfsdarg.dspathlen > 0 && nfsdarg.dspathlen < 10000 &&
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
nfsdarg.mdspathlen > 0 && nfsdarg.mdspathlen < 10000 &&
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
nfsdarg.mirrorcnt >= 1 &&
nfsdarg.mirrorcnt <= NFSDEV_MAXMIRRORS &&
nfsdarg.addr != NULL && nfsdarg.dnshost != 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
nfsdarg.dspath != NULL && nfsdarg.mdspath != 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
NFSD_DEBUG(1, "addrlen=%d dspathlen=%d dnslen=%d"
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
" mdspathlen=%d mirrorcnt=%d\n", nfsdarg.addrlen,
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
nfsdarg.dspathlen, nfsdarg.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
nfsdarg.mdspathlen, nfsdarg.mirrorcnt);
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
cp = malloc(nfsdarg.addrlen + 1, M_TEMP, M_WAITOK);
error = copyin(nfsdarg.addr, cp, nfsdarg.addrlen);
if (error != 0) {
free(cp, M_TEMP);
goto out;
}
cp[nfsdarg.addrlen] = '\0'; /* Ensure nul term. */
nfsdarg.addr = cp;
cp = malloc(nfsdarg.dnshostlen + 1, M_TEMP, M_WAITOK);
error = copyin(nfsdarg.dnshost, cp, nfsdarg.dnshostlen);
if (error != 0) {
free(nfsdarg.addr, M_TEMP);
free(cp, M_TEMP);
goto out;
}
cp[nfsdarg.dnshostlen] = '\0'; /* Ensure nul term. */
nfsdarg.dnshost = cp;
cp = malloc(nfsdarg.dspathlen + 1, M_TEMP, M_WAITOK);
error = copyin(nfsdarg.dspath, cp, nfsdarg.dspathlen);
if (error != 0) {
free(nfsdarg.addr, M_TEMP);
free(nfsdarg.dnshost, M_TEMP);
free(cp, M_TEMP);
goto out;
}
cp[nfsdarg.dspathlen] = '\0'; /* Ensure nul term. */
nfsdarg.dspath = cp;
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
cp = malloc(nfsdarg.mdspathlen + 1, M_TEMP, M_WAITOK);
error = copyin(nfsdarg.mdspath, cp, nfsdarg.mdspathlen);
if (error != 0) {
free(nfsdarg.addr, M_TEMP);
free(nfsdarg.dnshost, M_TEMP);
free(nfsdarg.dspath, M_TEMP);
free(cp, M_TEMP);
goto out;
}
cp[nfsdarg.mdspathlen] = '\0'; /* Ensure nul term. */
nfsdarg.mdspath = cp;
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 {
nfsdarg.addr = NULL;
nfsdarg.addrlen = 0;
nfsdarg.dnshost = NULL;
nfsdarg.dnshostlen = 0;
nfsdarg.dspath = NULL;
nfsdarg.dspathlen = 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
nfsdarg.mdspath = NULL;
nfsdarg.mdspathlen = 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
nfsdarg.mirrorcnt = 1;
}
Modify the experimental nfs server to use the new nfsd_nfsd_args structure for nfsd. Includes a change that clarifies the use of an empty principal name string to indicate AUTH_SYS only. Approved by: kib (mentor)
2009-05-12 16:04:51 +00:00
error = nfsrvd_nfsd(td, &nfsdarg);
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(nfsdarg.addr, M_TEMP);
free(nfsdarg.dnshost, M_TEMP);
free(nfsdarg.dspath, 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
free(nfsdarg.mdspath, M_TEMP);
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 (uap->flag & NFSSVC_PNFSDS) {
error = copyin(uap->argp, &pnfsdarg, sizeof(pnfsdarg));
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 (error == 0 && (pnfsdarg.op == PNFSDOP_DELDSSERVER ||
pnfsdarg.op == PNFSDOP_FORCEDELDS)) {
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
cp = malloc(PATH_MAX + 1, M_TEMP, M_WAITOK);
error = copyinstr(pnfsdarg.dspath, cp, PATH_MAX + 1,
NULL);
if (error == 0)
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
error = nfsrv_deldsserver(pnfsdarg.op, cp, td);
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(cp, M_TEMP);
} else if (error == 0 && pnfsdarg.op == PNFSDOP_COPYMR) {
cp = malloc(PATH_MAX + 1, M_TEMP, M_WAITOK);
buflen = sizeof(*pf) * NFSDEV_MAXMIRRORS;
buf = malloc(buflen, M_TEMP, M_WAITOK);
error = copyinstr(pnfsdarg.mdspath, cp, PATH_MAX + 1,
NULL);
NFSD_DEBUG(4, "pnfsdcopymr cp mdspath=%d\n", error);
if (error == 0 && pnfsdarg.dspath != NULL) {
cp2 = malloc(PATH_MAX + 1, M_TEMP, M_WAITOK);
error = copyinstr(pnfsdarg.dspath, cp2,
PATH_MAX + 1, NULL);
NFSD_DEBUG(4, "pnfsdcopymr cp dspath=%d\n",
error);
} else
cp2 = NULL;
if (error == 0 && pnfsdarg.curdspath != NULL) {
cp3 = malloc(PATH_MAX + 1, M_TEMP, M_WAITOK);
error = copyinstr(pnfsdarg.curdspath, cp3,
PATH_MAX + 1, NULL);
NFSD_DEBUG(4, "pnfsdcopymr cp curdspath=%d\n",
error);
} else
cp3 = NULL;
curdvp = NULL;
fds = NULL;
if (error == 0)
error = nfsrv_mdscopymr(cp, cp2, cp3, buf,
&buflen, fname, td, &vp, &nvp, &pf, &ds,
&fds);
NFSD_DEBUG(4, "nfsrv_mdscopymr=%d\n", error);
if (error == 0) {
if (pf->dsf_dir >= nfsrv_dsdirsize) {
printf("copymr: dsdir out of range\n");
pf->dsf_dir = 0;
}
NFSD_DEBUG(4, "copymr: buflen=%d\n", buflen);
error = nfsrv_copymr(vp, nvp,
ds->nfsdev_dsdir[pf->dsf_dir], ds, pf,
(struct pnfsdsfile *)buf,
buflen / sizeof(*pf), td->td_ucred, td);
vput(vp);
vput(nvp);
if (fds != NULL && error == 0) {
curdvp = fds->nfsdev_dsdir[pf->dsf_dir];
ret = vn_lock(curdvp, LK_EXCLUSIVE);
if (ret == 0) {
nfsrv_dsremove(curdvp, fname,
td->td_ucred, td);
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(curdvp);
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=%d\n", error);
}
free(cp, M_TEMP);
free(cp2, M_TEMP);
free(cp3, M_TEMP);
free(buf, 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
} else {
error = nfssvc_srvcall(td, uap, td->td_ucred);
}
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);
}
static int
nfssvc_srvcall(struct thread *p, struct nfssvc_args *uap, struct ucred *cred)
{
struct nfsex_args export;
Fix export_args ex_flags field so that is 64bits, the same as mnt_flags. Since mnt_flags was upgraded to 64bits there has been a quirk in "struct export_args", since it hold a copy of mnt_flags in ex_flags, which is an "int" (32bits). This happens to currently work, since all the flag bits used in ex_flags are defined in the low order 32bits. However, new export flags cannot be defined. Also, ex_anon is a "struct xucred", which limits it to 16 additional groups. This patch revises "struct export_args" to make ex_flags 64bits and replaces ex_anon with ex_uid, ex_ngroups and ex_groups (which points to a groups list, so it can be malloc'd up to NGROUPS in size. This requires that the VFS_CHECKEXP() arguments change, so I also modified the last "secflavors" argument to be an array pointer, so that the secflavors could be copied in VFS_CHECKEXP() while the export entry is locked. (Without this patch VFS_CHECKEXP() returns a pointer to the secflavors array and then it is used after being unlocked, which is potentially a problem if the exports entry is changed. In practice this does not occur when mountd is run with "-S", but I think it is worth fixing.) This patch also deleted the vfs_oexport_conv() function, since do_mount_update() does the conversion, as required by the old vfs_cmount() calls. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:10:18 +00:00
struct nfsex_oldargs oexp;
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 file *fp = NULL;
Fix export_args ex_flags field so that is 64bits, the same as mnt_flags. Since mnt_flags was upgraded to 64bits there has been a quirk in "struct export_args", since it hold a copy of mnt_flags in ex_flags, which is an "int" (32bits). This happens to currently work, since all the flag bits used in ex_flags are defined in the low order 32bits. However, new export flags cannot be defined. Also, ex_anon is a "struct xucred", which limits it to 16 additional groups. This patch revises "struct export_args" to make ex_flags 64bits and replaces ex_anon with ex_uid, ex_ngroups and ex_groups (which points to a groups list, so it can be malloc'd up to NGROUPS in size. This requires that the VFS_CHECKEXP() arguments change, so I also modified the last "secflavors" argument to be an array pointer, so that the secflavors could be copied in VFS_CHECKEXP() while the export entry is locked. (Without this patch VFS_CHECKEXP() returns a pointer to the secflavors array and then it is used after being unlocked, which is potentially a problem if the exports entry is changed. In practice this does not occur when mountd is run with "-S", but I think it is worth fixing.) This patch also deleted the vfs_oexport_conv() function, since do_mount_update() does the conversion, as required by the old vfs_cmount() calls. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:10:18 +00:00
int stablefd, i, 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
struct nfsd_clid adminrevoke;
struct nfsd_dumplist dumplist;
struct nfsd_dumpclients *dumpclients;
struct nfsd_dumplocklist dumplocklist;
struct nfsd_dumplocks *dumplocks;
struct nameidata nd;
vnode_t vp;
Add two new options to the nfssvc(2) syscall that allow processes running as root to suspend/resume execution of the kernel nfsd threads. An earlier version of this patch was tested by Vincent Hoffman (vince at unsane.co.uk) and John Hickey (jh at deterlab.net). Reviewed by: kib MFC after: 2 weeks
2012-10-14 22:33:17 +00:00
int error = EINVAL, igotlock;
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
struct proc *procp;
Fix export_args ex_flags field so that is 64bits, the same as mnt_flags. Since mnt_flags was upgraded to 64bits there has been a quirk in "struct export_args", since it hold a copy of mnt_flags in ex_flags, which is an "int" (32bits). This happens to currently work, since all the flag bits used in ex_flags are defined in the low order 32bits. However, new export flags cannot be defined. Also, ex_anon is a "struct xucred", which limits it to 16 additional groups. This patch revises "struct export_args" to make ex_flags 64bits and replaces ex_anon with ex_uid, ex_ngroups and ex_groups (which points to a groups list, so it can be malloc'd up to NGROUPS in size. This requires that the VFS_CHECKEXP() arguments change, so I also modified the last "secflavors" argument to be an array pointer, so that the secflavors could be copied in VFS_CHECKEXP() while the export entry is locked. (Without this patch VFS_CHECKEXP() returns a pointer to the secflavors array and then it is used after being unlocked, which is potentially a problem if the exports entry is changed. In practice this does not occur when mountd is run with "-S", but I think it is worth fixing.) This patch also deleted the vfs_oexport_conv() function, since do_mount_update() does the conversion, as required by the old vfs_cmount() calls. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:10:18 +00:00
gid_t *grps;
Add two new options to the nfssvc(2) syscall that allow processes running as root to suspend/resume execution of the kernel nfsd threads. An earlier version of this patch was tested by Vincent Hoffman (vince at unsane.co.uk) and John Hickey (jh at deterlab.net). Reviewed by: kib MFC after: 2 weeks
2012-10-14 22:33:17 +00:00
static int suspend_nfsd = 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 (uap->flag & NFSSVC_PUBLICFH) {
NFSBZERO((caddr_t)&nfs_pubfh.nfsrvfh_data,
sizeof (fhandle_t));
error = copyin(uap->argp,
&nfs_pubfh.nfsrvfh_data, sizeof (fhandle_t));
if (!error)
nfs_pubfhset = 1;
Fix export_args ex_flags field so that is 64bits, the same as mnt_flags. Since mnt_flags was upgraded to 64bits there has been a quirk in "struct export_args", since it hold a copy of mnt_flags in ex_flags, which is an "int" (32bits). This happens to currently work, since all the flag bits used in ex_flags are defined in the low order 32bits. However, new export flags cannot be defined. Also, ex_anon is a "struct xucred", which limits it to 16 additional groups. This patch revises "struct export_args" to make ex_flags 64bits and replaces ex_anon with ex_uid, ex_ngroups and ex_groups (which points to a groups list, so it can be malloc'd up to NGROUPS in size. This requires that the VFS_CHECKEXP() arguments change, so I also modified the last "secflavors" argument to be an array pointer, so that the secflavors could be copied in VFS_CHECKEXP() while the export entry is locked. (Without this patch VFS_CHECKEXP() returns a pointer to the secflavors array and then it is used after being unlocked, which is potentially a problem if the exports entry is changed. In practice this does not occur when mountd is run with "-S", but I think it is worth fixing.) This patch also deleted the vfs_oexport_conv() function, since do_mount_update() does the conversion, as required by the old vfs_cmount() calls. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:10:18 +00:00
} else if ((uap->flag & (NFSSVC_V4ROOTEXPORT | NFSSVC_NEWSTRUCT)) ==
(NFSSVC_V4ROOTEXPORT | NFSSVC_NEWSTRUCT)) {
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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 = copyin(uap->argp,(caddr_t)&export,
sizeof (struct nfsex_args));
Fix export_args ex_flags field so that is 64bits, the same as mnt_flags. Since mnt_flags was upgraded to 64bits there has been a quirk in "struct export_args", since it hold a copy of mnt_flags in ex_flags, which is an "int" (32bits). This happens to currently work, since all the flag bits used in ex_flags are defined in the low order 32bits. However, new export flags cannot be defined. Also, ex_anon is a "struct xucred", which limits it to 16 additional groups. This patch revises "struct export_args" to make ex_flags 64bits and replaces ex_anon with ex_uid, ex_ngroups and ex_groups (which points to a groups list, so it can be malloc'd up to NGROUPS in size. This requires that the VFS_CHECKEXP() arguments change, so I also modified the last "secflavors" argument to be an array pointer, so that the secflavors could be copied in VFS_CHECKEXP() while the export entry is locked. (Without this patch VFS_CHECKEXP() returns a pointer to the secflavors array and then it is used after being unlocked, which is potentially a problem if the exports entry is changed. In practice this does not occur when mountd is run with "-S", but I think it is worth fixing.) This patch also deleted the vfs_oexport_conv() function, since do_mount_update() does the conversion, as required by the old vfs_cmount() calls. Reviewed by: kib, freqlabs Relnotes: yes Differential Revision: https://reviews.freebsd.org/D25088
2020-06-14 00:10:18 +00:00
if (!error) {
grps = NULL;
if (export.export.ex_ngroups > NGROUPS_MAX ||
export.export.ex_ngroups < 0)
error = EINVAL;
else if (export.export.ex_ngroups > 0) {
grps = malloc(export.export.ex_ngroups *
sizeof(gid_t), M_TEMP, M_WAITOK);
error = copyin(export.export.ex_groups, grps,
export.export.ex_ngroups * sizeof(gid_t));
export.export.ex_groups = grps;
} else
export.export.ex_groups = NULL;
if (!error)
error = nfsrv_v4rootexport(&export, cred, p);
free(grps, M_TEMP);
}
} else if ((uap->flag & (NFSSVC_V4ROOTEXPORT | NFSSVC_NEWSTRUCT)) ==
NFSSVC_V4ROOTEXPORT) {
error = copyin(uap->argp,(caddr_t)&oexp,
sizeof (struct nfsex_oldargs));
if (!error) {
memset(&export.export, 0, sizeof(export.export));
export.export.ex_flags = (uint64_t)oexp.export.ex_flags;
export.export.ex_root = oexp.export.ex_root;
export.export.ex_uid = oexp.export.ex_anon.cr_uid;
export.export.ex_ngroups =
oexp.export.ex_anon.cr_ngroups;
export.export.ex_groups = NULL;
if (export.export.ex_ngroups > XU_NGROUPS ||
export.export.ex_ngroups < 0)
error = EINVAL;
else if (export.export.ex_ngroups > 0) {
export.export.ex_groups = malloc(
export.export.ex_ngroups * sizeof(gid_t),
M_TEMP, M_WAITOK);
for (i = 0; i < export.export.ex_ngroups; i++)
export.export.ex_groups[i] =
oexp.export.ex_anon.cr_groups[i];
}
export.export.ex_addr = oexp.export.ex_addr;
export.export.ex_addrlen = oexp.export.ex_addrlen;
export.export.ex_mask = oexp.export.ex_mask;
export.export.ex_masklen = oexp.export.ex_masklen;
export.export.ex_indexfile = oexp.export.ex_indexfile;
export.export.ex_numsecflavors =
oexp.export.ex_numsecflavors;
if (export.export.ex_numsecflavors >= MAXSECFLAVORS ||
export.export.ex_numsecflavors < 0)
error = EINVAL;
else {
for (i = 0; i < export.export.ex_numsecflavors;
i++)
export.export.ex_secflavors[i] =
oexp.export.ex_secflavors[i];
}
export.fspec = oexp.fspec;
if (error == 0)
error = nfsrv_v4rootexport(&export, cred, p);
free(export.export.ex_groups, 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
} else if (uap->flag & NFSSVC_NOPUBLICFH) {
nfs_pubfhset = 0;
error = 0;
} else if (uap->flag & NFSSVC_STABLERESTART) {
error = copyin(uap->argp, (caddr_t)&stablefd,
sizeof (int));
if (!error)
error = fp_getfvp(p, stablefd, &fp, &vp);
if (!error && (NFSFPFLAG(fp) & (FREAD | FWRITE)) != (FREAD | FWRITE))
error = EBADF;
if (!error && newnfs_numnfsd != 0)
error = EPERM;
if (!error) {
nfsrv_stablefirst.nsf_fp = fp;
nfsrv_setupstable(p);
}
} else if (uap->flag & NFSSVC_ADMINREVOKE) {
error = copyin(uap->argp, (caddr_t)&adminrevoke,
sizeof (struct nfsd_clid));
if (!error)
error = nfsrv_adminrevoke(&adminrevoke, p);
} else if (uap->flag & NFSSVC_DUMPCLIENTS) {
error = copyin(uap->argp, (caddr_t)&dumplist,
sizeof (struct nfsd_dumplist));
if (!error && (dumplist.ndl_size < 1 ||
dumplist.ndl_size > NFSRV_MAXDUMPLIST))
error = EPERM;
if (!error) {
len = sizeof (struct nfsd_dumpclients) * dumplist.ndl_size;
nfs: Zero the buffers exported by NFSSVC_DUMPCLIENTS and DUMPLOCKS. Note that these interfaces are available only to root. admbugs: 765 Reported by: Vlad Tsyrklevich <vlad@tsyrklevich.net> Reviewed by: rmacklem MFC after: 1 day Security: Kernel memory disclosure Sponsored by: The FreeBSD Foundation
2019-01-21 23:54:33 +00:00
dumpclients = malloc(len, 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
nfsrv_dumpclients(dumpclients, dumplist.ndl_size);
Remove Mac OS/X macros that did nothing for FreeBSD. The macros CAST_USER_ADDR_T() and CAST_DOWN() were used for the Mac OS/X port. The first of these macros was a no-op for FreeBSD and the second is no longer used. This patch gets rid of them. It also deletes the "mbuf_t" typedef which is no longer used in the FreeBSD code from nfskpiport.h This patch should not change semantics.
2020-04-25 02:18:59 +00:00
error = copyout(dumpclients, dumplist.ndl_list, len);
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(dumpclients, 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
}
} else if (uap->flag & NFSSVC_DUMPLOCKS) {
error = copyin(uap->argp, (caddr_t)&dumplocklist,
sizeof (struct nfsd_dumplocklist));
if (!error && (dumplocklist.ndllck_size < 1 ||
dumplocklist.ndllck_size > NFSRV_MAXDUMPLIST))
error = EPERM;
if (!error)
error = nfsrv_lookupfilename(&nd,
dumplocklist.ndllck_fname, p);
if (!error) {
len = sizeof (struct nfsd_dumplocks) *
dumplocklist.ndllck_size;
nfs: Zero the buffers exported by NFSSVC_DUMPCLIENTS and DUMPLOCKS. Note that these interfaces are available only to root. admbugs: 765 Reported by: Vlad Tsyrklevich <vlad@tsyrklevich.net> Reviewed by: rmacklem MFC after: 1 day Security: Kernel memory disclosure Sponsored by: The FreeBSD Foundation
2019-01-21 23:54:33 +00:00
dumplocks = malloc(len, 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
nfsrv_dumplocks(nd.ni_vp, dumplocks,
dumplocklist.ndllck_size, p);
vput(nd.ni_vp);
Remove Mac OS/X macros that did nothing for FreeBSD. The macros CAST_USER_ADDR_T() and CAST_DOWN() were used for the Mac OS/X port. The first of these macros was a no-op for FreeBSD and the second is no longer used. This patch gets rid of them. It also deletes the "mbuf_t" typedef which is no longer used in the FreeBSD code from nfskpiport.h This patch should not change semantics.
2020-04-25 02:18:59 +00:00
error = copyout(dumplocks, dumplocklist.ndllck_list,
len);
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(dumplocks, 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
} else if (uap->flag & NFSSVC_BACKUPSTABLE) {
procp = p->td_proc;
PROC_LOCK(procp);
nfsd_master_pid = procp->p_pid;
bcopy(procp->p_comm, nfsd_master_comm, MAXCOMLEN + 1);
nfsd_master_start = procp->p_stats->p_start;
nfsd_master_proc = procp;
PROC_UNLOCK(procp);
Add two new options to the nfssvc(2) syscall that allow processes running as root to suspend/resume execution of the kernel nfsd threads. An earlier version of this patch was tested by Vincent Hoffman (vince at unsane.co.uk) and John Hickey (jh at deterlab.net). Reviewed by: kib MFC after: 2 weeks
2012-10-14 22:33:17 +00:00
} else if ((uap->flag & NFSSVC_SUSPENDNFSD) != 0) {
NFSLOCKV4ROOTMUTEX();
if (suspend_nfsd == 0) {
/* Lock out all nfsd threads */
do {
igotlock = nfsv4_lock(&nfsd_suspend_lock, 1,
NULL, NFSV4ROOTLOCKMUTEXPTR, NULL);
} while (igotlock == 0 && suspend_nfsd == 0);
suspend_nfsd = 1;
}
NFSUNLOCKV4ROOTMUTEX();
error = 0;
} else if ((uap->flag & NFSSVC_RESUMENFSD) != 0) {
NFSLOCKV4ROOTMUTEX();
if (suspend_nfsd != 0) {
nfsv4_unlock(&nfsd_suspend_lock, 0);
suspend_nfsd = 0;
}
NFSUNLOCKV4ROOTMUTEX();
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
}
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);
}
Apply changes to the experimental nfs server so that it uses the security flavors as exported in FreeBSD-CURRENT. This allows it to use a slightly modified mountd.c instead of a different utility. Approved by: kib (mentor)
2009-05-14 21:39:08 +00:00
/*
* Check exports.
* Returns 0 if ok, 1 otherwise.
*/
int
nfsvno_testexp(struct nfsrv_descript *nd, struct nfsexstuff *exp)
{
int i;
nfsd: Fix when NFSERR_WRONGSEC may be replied to NFSv4 clients Commit d224f05fcfc1 pre-parsed the next operation number for the put file handle operations. This patch uses this next operation number, plus the type of the file handle being set by the put file handle operation, to implement the rules in RFC5661 Sec. 2.6 with respect to replying NFSERR_WRONGSEC. This patch also adds a check to see if NFSERR_WRONGSEC should be replied when about to perform Lookup, Lookupp or Open with a file name component, so that the NFSERR_WRONGSEC reply is done for these operations, as required by RFC5661 Sec. 2.6. This patch does not have any practical effect for the FreeBSD NFSv4 client and I believe that the same is true for the Linux client, since NFSERR_WRONGSEC is considered a fatal error at this time. MFC after: 2 weeks
2021-06-05 23:53:07 +00:00
/*
* Allow NFSv3 Fsinfo per RFC2623.
*/
if (((nd->nd_flag & ND_NFSV4) != 0 ||
nd->nd_procnum != NFSPROC_FSINFO) &&
((NFSVNO_EXTLS(exp) && (nd->nd_flag & ND_TLS) == 0) ||
(NFSVNO_EXTLSCERT(exp) &&
(nd->nd_flag & ND_TLSCERT) == 0) ||
(NFSVNO_EXTLSCERTUSER(exp) &&
(nd->nd_flag & ND_TLSCERTUSER) == 0))) {
if ((nd->nd_flag & ND_NFSV4) != 0)
return (NFSERR_WRONGSEC);
else if ((nd->nd_flag & ND_TLS) == 0)
return (NFSERR_AUTHERR | AUTH_NEEDS_TLS);
else
return (NFSERR_AUTHERR | AUTH_NEEDS_TLS_MUTUAL_HOST);
}
Apply changes to the experimental nfs server so that it uses the security flavors as exported in FreeBSD-CURRENT. This allows it to use a slightly modified mountd.c instead of a different utility. Approved by: kib (mentor)
2009-05-14 21:39:08 +00:00
/*
* This seems odd, but allow the case where the security flavor
* list is empty. This happens when NFSv4 is traversing non-exported
* file systems. Exported file systems should always have a non-empty
* security flavor list.
*/
if (exp->nes_numsecflavor == 0)
return (0);
for (i = 0; i < exp->nes_numsecflavor; i++) {
/*
* The tests for privacy and integrity must be first,
* since ND_GSS is set for everything but AUTH_SYS.
*/
if (exp->nes_secflavors[i] == RPCSEC_GSS_KRB5P &&
(nd->nd_flag & ND_GSSPRIVACY))
return (0);
if (exp->nes_secflavors[i] == RPCSEC_GSS_KRB5I &&
(nd->nd_flag & ND_GSSINTEGRITY))
return (0);
if (exp->nes_secflavors[i] == RPCSEC_GSS_KRB5 &&
(nd->nd_flag & ND_GSS))
return (0);
if (exp->nes_secflavors[i] == AUTH_SYS &&
(nd->nd_flag & ND_GSS) == 0)
return (0);
}
nfsd: Fix when NFSERR_WRONGSEC may be replied to NFSv4 clients Commit d224f05fcfc1 pre-parsed the next operation number for the put file handle operations. This patch uses this next operation number, plus the type of the file handle being set by the put file handle operation, to implement the rules in RFC5661 Sec. 2.6 with respect to replying NFSERR_WRONGSEC. This patch also adds a check to see if NFSERR_WRONGSEC should be replied when about to perform Lookup, Lookupp or Open with a file name component, so that the NFSERR_WRONGSEC reply is done for these operations, as required by RFC5661 Sec. 2.6. This patch does not have any practical effect for the FreeBSD NFSv4 client and I believe that the same is true for the Linux client, since NFSERR_WRONGSEC is considered a fatal error at this time. MFC after: 2 weeks
2021-06-05 23:53:07 +00:00
if ((nd->nd_flag & ND_NFSV4) != 0)
return (NFSERR_WRONGSEC);
return (NFSERR_AUTHERR | AUTH_TOOWEAK);
Apply changes to the experimental nfs server so that it uses the security flavors as exported in FreeBSD-CURRENT. This allows it to use a slightly modified mountd.c instead of a different utility. Approved by: kib (mentor)
2009-05-14 21:39:08 +00:00
}
Modify the file handle hash function in the experimental NFS server so that it will work better for non-UFS file systems. The new function simply sums the bytes of the fh_fid field of fhandle_t. MFC after: 10 days
2010-10-22 21:38:56 +00:00
/*
* Calculate a hash value for the fid in a file handle.
*/
Modify the experimental NFSv4 server's file handle hash function to use the generic hash32_buf() function. Although adding the bytes seemed sufficient for UFS and ZFS, since most of the bytes are the same for file handles on the same volume, this might not be sufficient for other file systems. Use of a generic function also seems preferable to one specific to NFSv4. Suggested by: gleb.kurtsou at gmail.com MFC after: 10 days
2010-10-23 22:28:29 +00:00
uint32_t
Modify the file handle hash function in the experimental NFS server so that it will work better for non-UFS file systems. The new function simply sums the bytes of the fh_fid field of fhandle_t. MFC after: 10 days
2010-10-22 21:38:56 +00:00
nfsrv_hashfh(fhandle_t *fhp)
{
Modify the experimental NFSv4 server's file handle hash function to use the generic hash32_buf() function. Although adding the bytes seemed sufficient for UFS and ZFS, since most of the bytes are the same for file handles on the same volume, this might not be sufficient for other file systems. Use of a generic function also seems preferable to one specific to NFSv4. Suggested by: gleb.kurtsou at gmail.com MFC after: 10 days
2010-10-23 22:28:29 +00:00
uint32_t hashval;
Modify the file handle hash function in the experimental NFS server so that it will work better for non-UFS file systems. The new function simply sums the bytes of the fh_fid field of fhandle_t. MFC after: 10 days
2010-10-22 21:38:56 +00:00
Modify the experimental NFSv4 server's file handle hash function to use the generic hash32_buf() function. Although adding the bytes seemed sufficient for UFS and ZFS, since most of the bytes are the same for file handles on the same volume, this might not be sufficient for other file systems. Use of a generic function also seems preferable to one specific to NFSv4. Suggested by: gleb.kurtsou at gmail.com MFC after: 10 days
2010-10-23 22:28:29 +00:00
hashval = hash32_buf(&fhp->fh_fid, sizeof(struct fid), 0);
Modify the file handle hash function in the experimental NFS server so that it will work better for non-UFS file systems. The new function simply sums the bytes of the fh_fid field of fhandle_t. MFC after: 10 days
2010-10-22 21:38:56 +00:00
return (hashval);
}
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
/*
* Calculate a hash value for the sessionid.
*/
uint32_t
nfsrv_hashsessionid(uint8_t *sessionid)
{
uint32_t hashval;
hashval = hash32_buf(sessionid, NFSX_V4SESSIONID, 0);
return (hashval);
}
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
/*
* Signal the userland master nfsd to backup the stable restart file.
*/
void
nfsrv_backupstable(void)
{
struct proc *procp;
if (nfsd_master_proc != NULL) {
procp = pfind(nfsd_master_pid);
/* Try to make sure it is the correct process. */
if (procp == nfsd_master_proc &&
procp->p_stats->p_start.tv_sec ==
nfsd_master_start.tv_sec &&
procp->p_stats->p_start.tv_usec ==
nfsd_master_start.tv_usec &&
strcmp(procp->p_comm, nfsd_master_comm) == 0)
In order to maximize the re-usability of kernel code in user space this patch modifies makesyscalls.sh to prefix all of the non-compatibility calls (e.g. not linux_, freebsd32_) with sys_ and updates the kernel entry points and all places in the code that use them. It also fixes an additional name space collision between the kernel function psignal and the libc function of the same name by renaming the kernel psignal kern_psignal(). By introducing this change now we will ease future MFCs that change syscalls. Reviewed by: rwatson Approved by: re (bz)
2011-09-16 13:58:51 +00:00
kern_psignal(procp, SIGUSR2);
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
else
nfsd_master_proc = NULL;
if (procp != NULL)
PROC_UNLOCK(procp);
}
}
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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
* Create a DS data file for nfsrv_pnfscreate(). Called for each mirror.
* The arguments are in a structure, so that they can be passed through
* taskqueue for a kernel process to execute this function.
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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
struct nfsrvdscreate {
int done;
int inprog;
struct task tsk;
struct ucred *tcred;
struct vnode *dvp;
NFSPROC_T *p;
struct pnfsdsfile *pf;
int err;
fhandle_t fh;
struct vattr va;
struct vattr createva;
};
int
nfsrv_dscreate(struct vnode *dvp, struct vattr *vap, struct vattr *nvap,
fhandle_t *fhp, struct pnfsdsfile *pf, struct pnfsdsattr *dsa,
char *fnamep, struct ucred *tcred, NFSPROC_T *p, struct vnode **nvpp)
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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
struct vnode *nvp;
struct nameidata named;
struct vattr va;
char *bufp;
u_long *hashp;
struct nfsnode *np;
struct nfsmount *nmp;
int 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
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
NFSNAMEICNDSET(&named.ni_cnd, tcred, CREATE,
LOCKPARENT | LOCKLEAF | SAVESTART | NOCACHE);
nfsvno_setpathbuf(&named, &bufp, &hashp);
named.ni_cnd.cn_lkflags = LK_EXCLUSIVE;
named.ni_cnd.cn_thread = p;
named.ni_cnd.cn_nameptr = bufp;
if (fnamep != NULL) {
strlcpy(bufp, fnamep, PNFS_FILENAME_LEN + 1);
named.ni_cnd.cn_namelen = strlen(bufp);
} else
named.ni_cnd.cn_namelen = nfsrv_putfhname(fhp, bufp);
NFSD_DEBUG(4, "nfsrv_dscreate: dvp=%p fname=%s\n", dvp, bufp);
/* Create the date file in the DS mount. */
error = NFSVOPLOCK(dvp, LK_EXCLUSIVE);
if (error == 0) {
error = VOP_CREATE(dvp, &nvp, &named.ni_cnd, vap);
nfsserver: use VOP_VPUT_PAIR(). Apply VOP_VPUT_PAIR() to the end of vnode operations after the VOP_MKNOD(), VOP_MKDIR(), VOP_LINK(), VOP_SYMLINK(), VOP_CREATE(). Reviewed by: chs, mckusick Tested by: pho MFC after: 2 weeks Sponsored by: The FreeBSD Foundation
2021-01-24 13:02:27 +00:00
vref(dvp);
VOP_VPUT_PAIR(dvp, error == 0 ? &nvp : NULL, 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
if (error == 0) {
/* Set the ownership of the file. */
error = VOP_SETATTR(nvp, nvap, tcred);
NFSD_DEBUG(4, "nfsrv_dscreate:"
" setattr-uid=%d\n", error);
if (error != 0)
vput(nvp);
}
if (error != 0)
printf("pNFS: pnfscreate failed=%d\n", error);
} else
printf("pNFS: pnfscreate vnlock=%d\n", error);
if (error == 0) {
np = VTONFS(nvp);
nmp = VFSTONFS(nvp->v_mount);
if (strcmp(nvp->v_mount->mnt_vfc->vfc_name, "nfs")
!= 0 || nmp->nm_nam->sa_len > sizeof(
struct sockaddr_in6) ||
np->n_fhp->nfh_len != NFSX_MYFH) {
printf("Bad DS file: fstype=%s salen=%d"
" fhlen=%d\n",
nvp->v_mount->mnt_vfc->vfc_name,
nmp->nm_nam->sa_len, np->n_fhp->nfh_len);
error = ENOENT;
Fix several performance related issues in the new NFS server's DRC for NFS over TCP. - Increase the size of the hash tables. - Create a separate mutex for each hash list of the TCP hash table. - Single thread the code that deletes stale cache entries. - Add a tunable called vfs.nfsd.tcphighwater, which can be increased to allow the cache to grow larger, avoiding the overhead of frequent scans to delete stale cache entries. (The default value will result in frequent scans to delete stale cache entries, analagous to what the pre-patched code does.) - Add a tunable called vfs.nfsd.cachetcp that can be used to disable DRC caching for NFS over TCP, since the old NFS server didn't DRC cache TCP. It also adjusts the size of nfsrc_floodlevel dynamically, so that it is always greater than vfs.nfsd.tcphighwater. For UDP the algorithm remains the same as the pre-patched code, but the tunable vfs.nfsd.udphighwater can be used to allow the cache to grow larger and reduce the overhead caused by frequent scans for stale entries. UDP also uses a larger hash table size than the pre-patched code. Reported by: wollman Tested by: wollman (earlier version of patch) Submitted by: ivoras (earlier patch) Reviewed by: jhb (earlier version of patch) MFC after: 1 month
2013-08-14 21:11:26 +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
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
/* Set extattrs for the DS on the MDS file. */
if (error == 0) {
if (dsa != NULL) {
error = VOP_GETATTR(nvp, &va, tcred);
if (error == 0) {
dsa->dsa_filerev = va.va_filerev;
dsa->dsa_size = va.va_size;
dsa->dsa_atime = va.va_atime;
dsa->dsa_mtime = va.va_mtime;
Fix the pNFS server's reporting of SpaceUsed (va_bytes). The pNFS server currently reports SpaceUsed (va_bytes) for the metadata file. This in not correct, since the metadata file is always empty and, as such, va_bytes is just the allocation for the empty file. This patch adds va_bytes to the list of attributes acquired from the DS for a file, so that it includes the allocated data size and is updated when the file is written. For files created on a pNFS server before this patch is applied, the va_bytes value is estimated by rounding va_size up to a multiple of BLKDEV_IOSIZE. Once the file is written after this patch has been applied to the metadata server, the va_bytes returned for the file will be correct. This patch only affects a pNFS metadata server. Found during testing of the NFSv4.2 pNFS server for the Allocate operation. (Not yet in head/current.) MFC after: 2 weeks
2019-11-22 00:22:55 +00:00
dsa->dsa_bytes = va.va_bytes;
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) {
NFSBCOPY(np->n_fhp->nfh_fh, &pf->dsf_fh,
NFSX_MYFH);
NFSBCOPY(nmp->nm_nam, &pf->dsf_sin,
nmp->nm_nam->sa_len);
NFSBCOPY(named.ni_cnd.cn_nameptr,
pf->dsf_filename,
sizeof(pf->dsf_filename));
}
} else
printf("pNFS: pnfscreate can't get DS"
" attr=%d\n", error);
if (nvpp != NULL && error == 0)
*nvpp = nvp;
else
vput(nvp);
}
nfsvno_relpathbuf(&named);
return (error);
}
/*
* Start up the thread that will execute nfsrv_dscreate().
*/
static void
start_dscreate(void *arg, int pending)
{
struct nfsrvdscreate *dsc;
dsc = (struct nfsrvdscreate *)arg;
dsc->err = nfsrv_dscreate(dsc->dvp, &dsc->createva, &dsc->va, &dsc->fh,
dsc->pf, NULL, NULL, dsc->tcred, dsc->p, NULL);
dsc->done = 1;
NFSD_DEBUG(4, "start_dscreate: err=%d\n", dsc->err);
}
/*
* Create a pNFS data file on the Data Server(s).
*/
static void
nfsrv_pnfscreate(struct vnode *vp, struct vattr *vap, struct ucred *cred,
NFSPROC_T *p)
{
Fix the build for MAC not defined and a couple of might not be initialized. r355677 broke the build for the not MAC defined case and a couple of might not be initialized warnings were generated for riscv. Others seem to be erroneous. Hopefully there won't be too many more build errors. Pointy hat goes on me.
2019-12-13 00:45:14 +00:00
struct nfsrvdscreate *dsc, *tdsc = NULL;
Fix handling of the hybrid DS case for a pNFS server. After the addition of the "#mds_path" suffix for a DS specification on the "-p" nfsd option, it is possible to have a mix of DSs assigned to an MDS file system and DSs that store files for all DSs. This is what I referred to as "hybrid" above. At first, I didn't think this hybrid case would be useful, but I now believe that some system administrators may fine it useful. This patch modifies the file storage assignment algorithm so that it makes the "#mds_path" DSs take priority and the all file systems DSs are now only used for MDS file systems with no "#mds_path" DS servers. This only affects the pNFS server for this "hybrid" case.
2018-07-07 19:27:49 +00:00
struct nfsdevice *ds, *tds, *fds;
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 pnfsdsfile *pf, *tpf;
struct pnfsdsattr dsattr;
struct vattr va;
struct vnode *dvp[NFSDEV_MAXMIRRORS];
struct nfsmount *nmp;
fhandle_t fh;
uid_t vauid;
gid_t vagid;
u_short vamode;
struct ucred *tcred;
int dsdir[NFSDEV_MAXMIRRORS], error, i, mirrorcnt, ret;
int failpos, timo;
/* Get a DS server directory in a round-robin order. */
mirrorcnt = 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
mp = vp->v_mount;
Fix handling of the hybrid DS case for a pNFS server. After the addition of the "#mds_path" suffix for a DS specification on the "-p" nfsd option, it is possible to have a mix of DSs assigned to an MDS file system and DSs that store files for all DSs. This is what I referred to as "hybrid" above. At first, I didn't think this hybrid case would be useful, but I now believe that some system administrators may fine it useful. This patch modifies the file storage assignment algorithm so that it makes the "#mds_path" DSs take priority and the all file systems DSs are now only used for MDS file systems with no "#mds_path" DS servers. This only affects the pNFS server for this "hybrid" case.
2018-07-07 19:27:49 +00:00
ds = fds = 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
NFSDDSLOCK();
Fix handling of the hybrid DS case for a pNFS server. After the addition of the "#mds_path" suffix for a DS specification on the "-p" nfsd option, it is possible to have a mix of DSs assigned to an MDS file system and DSs that store files for all DSs. This is what I referred to as "hybrid" above. At first, I didn't think this hybrid case would be useful, but I now believe that some system administrators may fine it useful. This patch modifies the file storage assignment algorithm so that it makes the "#mds_path" DSs take priority and the all file systems DSs are now only used for MDS file systems with no "#mds_path" DS servers. This only affects the pNFS server for this "hybrid" case.
2018-07-07 19:27:49 +00:00
/*
* Search for the first entry that handles this MDS fs, but use the
* first entry for all MDS fs's otherwise.
*/
TAILQ_FOREACH(tds, &nfsrv_devidhead, nfsdev_list) {
if (tds->nfsdev_nmp != NULL) {
if (tds->nfsdev_mdsisset == 0 && ds == NULL)
ds = tds;
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
else if (tds->nfsdev_mdsisset != 0 && fsidcmp(
&mp->mnt_stat.f_fsid, &tds->nfsdev_mdsfsid) == 0) {
Fix handling of the hybrid DS case for a pNFS server. After the addition of the "#mds_path" suffix for a DS specification on the "-p" nfsd option, it is possible to have a mix of DSs assigned to an MDS file system and DSs that store files for all DSs. This is what I referred to as "hybrid" above. At first, I didn't think this hybrid case would be useful, but I now believe that some system administrators may fine it useful. This patch modifies the file storage assignment algorithm so that it makes the "#mds_path" DSs take priority and the all file systems DSs are now only used for MDS file systems with no "#mds_path" DS servers. This only affects the pNFS server for this "hybrid" case.
2018-07-07 19:27:49 +00:00
ds = fds = tds;
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
}
if (ds == NULL) {
NFSDDSUNLOCK();
NFSD_DEBUG(4, "nfsrv_pnfscreate: no srv\n");
return;
}
i = dsdir[0] = ds->nfsdev_nextdir;
ds->nfsdev_nextdir = (ds->nfsdev_nextdir + 1) % nfsrv_dsdirsize;
dvp[0] = ds->nfsdev_dsdir[i];
Fix handling of the hybrid DS case for a pNFS server. After the addition of the "#mds_path" suffix for a DS specification on the "-p" nfsd option, it is possible to have a mix of DSs assigned to an MDS file system and DSs that store files for all DSs. This is what I referred to as "hybrid" above. At first, I didn't think this hybrid case would be useful, but I now believe that some system administrators may fine it useful. This patch modifies the file storage assignment algorithm so that it makes the "#mds_path" DSs take priority and the all file systems DSs are now only used for MDS file systems with no "#mds_path" DS servers. This only affects the pNFS server for this "hybrid" case.
2018-07-07 19:27:49 +00:00
tds = TAILQ_NEXT(ds, nfsdev_list);
if (nfsrv_maxpnfsmirror > 1 && tds != NULL) {
TAILQ_FOREACH_FROM(tds, &nfsrv_devidhead, nfsdev_list) {
if (tds->nfsdev_nmp != NULL &&
((tds->nfsdev_mdsisset == 0 && fds == NULL) ||
(tds->nfsdev_mdsisset != 0 && fds != NULL &&
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(&mp->mnt_stat.f_fsid,
&tds->nfsdev_mdsfsid) == 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
dsdir[mirrorcnt] = i;
Fix handling of the hybrid DS case for a pNFS server. After the addition of the "#mds_path" suffix for a DS specification on the "-p" nfsd option, it is possible to have a mix of DSs assigned to an MDS file system and DSs that store files for all DSs. This is what I referred to as "hybrid" above. At first, I didn't think this hybrid case would be useful, but I now believe that some system administrators may fine it useful. This patch modifies the file storage assignment algorithm so that it makes the "#mds_path" DSs take priority and the all file systems DSs are now only used for MDS file systems with no "#mds_path" DS servers. This only affects the pNFS server for this "hybrid" case.
2018-07-07 19:27:49 +00:00
dvp[mirrorcnt] = tds->nfsdev_dsdir[i];
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
mirrorcnt++;
if (mirrorcnt >= nfsrv_maxpnfsmirror)
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 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
}
/* Put at end of list to implement round-robin usage. */
TAILQ_REMOVE(&nfsrv_devidhead, ds, nfsdev_list);
TAILQ_INSERT_TAIL(&nfsrv_devidhead, ds, nfsdev_list);
NFSDDSUNLOCK();
dsc = NULL;
if (mirrorcnt > 1)
tdsc = dsc = malloc(sizeof(*dsc) * (mirrorcnt - 1), M_TEMP,
M_WAITOK | M_ZERO);
Fix the pNFS server for a case where mirror level equals number of DSs. If a pNFS service was set up where the number of DSs equals the mirror level and then a DS was disabled, the service would create files with duplicate entries for the same DS. This bug occurred because I didn't realize that TAILQ_FOREACH_FROM() would start at the beginning of the list when the inital value of the variable was NULL. This patch also changes the pNFS server DS file creation code so that it creates entrie(s) with 0.0.0.0 IP address when it cannot create mirror level files due to lack of DSs. The patch only affects the pNFS service and only when it was created with a number of DSs equal to the mirror level and mirroring is enabled.
2018-06-29 12:41:36 +00:00
tpf = pf = malloc(sizeof(*pf) * nfsrv_maxpnfsmirror, 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
error = nfsvno_getfh(vp, &fh, p);
if (error == 0)
error = VOP_GETATTR(vp, &va, cred);
if (error == 0) {
/* Set the attributes for "vp" to Setattr the DS vp. */
vauid = va.va_uid;
vagid = va.va_gid;
vamode = va.va_mode;
VATTR_NULL(&va);
va.va_uid = vauid;
va.va_gid = vagid;
va.va_mode = vamode;
va.va_size = 0;
} else
printf("pNFS: pnfscreate getfh+attr=%d\n", error);
NFSD_DEBUG(4, "nfsrv_pnfscreate: cruid=%d crgid=%d\n", cred->cr_uid,
cred->cr_gid);
/* Make data file name based on FH. */
tcred = newnfs_getcred();
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, 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
/*
* Create the file on each DS mirror, using kernel process(es) for the
* additional mirrors.
*/
failpos = -1;
for (i = 0; i < mirrorcnt - 1 && error == 0; i++, tpf++, tdsc++) {
tpf->dsf_dir = dsdir[i];
tdsc->tcred = tcred;
tdsc->p = p;
tdsc->pf = tpf;
tdsc->createva = *vap;
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(&fh, &tdsc->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
tdsc->va = va;
tdsc->dvp = dvp[i];
tdsc->done = 0;
tdsc->inprog = 0;
tdsc->err = 0;
ret = EIO;
if (nfs_pnfsiothreads != 0) {
ret = nfs_pnfsio(start_dscreate, tdsc);
NFSD_DEBUG(4, "nfsrv_pnfscreate: nfs_pnfsio=%d\n", ret);
}
if (ret != 0) {
ret = nfsrv_dscreate(dvp[i], vap, &va, &fh, tpf, NULL,
NULL, tcred, p, NULL);
if (ret != 0) {
KASSERT(error == 0, ("nfsrv_dscreate err=%d",
error));
if (failpos == -1 && nfsds_failerr(ret))
failpos = i;
else
error = ret;
}
}
}
if (error == 0) {
tpf->dsf_dir = dsdir[mirrorcnt - 1];
error = nfsrv_dscreate(dvp[mirrorcnt - 1], vap, &va, &fh, tpf,
&dsattr, NULL, tcred, p, NULL);
if (failpos == -1 && mirrorcnt > 1 && nfsds_failerr(error)) {
failpos = mirrorcnt - 1;
error = 0;
}
}
timo = hz / 50; /* Wait for 20msec. */
if (timo < 1)
timo = 1;
/* Wait for kernel task(s) to complete. */
for (tdsc = dsc, i = 0; i < mirrorcnt - 1; i++, tdsc++) {
while (tdsc->inprog != 0 && tdsc->done == 0)
tsleep(&tdsc->tsk, PVFS, "srvdcr", timo);
if (tdsc->err != 0) {
if (failpos == -1 && nfsds_failerr(tdsc->err))
failpos = i;
else if (error == 0)
error = tdsc->err;
}
}
Clean up the experimental NFS server replay cache when the module is unloaded. Reviewed by: rmacklem Approved by: zml (mentor)
2011-01-12 23:34:09 +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
/*
* If failpos has been set, that mirror has failed, so it needs
* to be disabled.
*/
if (failpos >= 0) {
nmp = VFSTONFS(dvp[failpos]->v_mount);
NFSLOCKMNT(nmp);
if ((nmp->nm_privflag & (NFSMNTP_FORCEDISM |
NFSMNTP_CANCELRPCS)) == 0) {
nmp->nm_privflag |= NFSMNTP_CANCELRPCS;
NFSUNLOCKMNT(nmp);
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(PNFSDOP_DELDSSERVER, 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, "dscreatfail fail=%d ds=%p\n", failpos,
ds);
if (ds != NULL)
nfsrv_killrpcs(nmp);
NFSLOCKMNT(nmp);
nmp->nm_privflag &= ~NFSMNTP_CANCELRPCS;
wakeup(nmp);
}
NFSUNLOCKMNT(nmp);
}
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
NFSFREECRED(tcred);
if (error == 0) {
ASSERT_VOP_ELOCKED(vp, "nfsrv_pnfscreate vp");
Fix the pNFS server for a case where mirror level equals number of DSs. If a pNFS service was set up where the number of DSs equals the mirror level and then a DS was disabled, the service would create files with duplicate entries for the same DS. This bug occurred because I didn't realize that TAILQ_FOREACH_FROM() would start at the beginning of the list when the inital value of the variable was NULL. This patch also changes the pNFS server DS file creation code so that it creates entrie(s) with 0.0.0.0 IP address when it cannot create mirror level files due to lack of DSs. The patch only affects the pNFS service and only when it was created with a number of DSs equal to the mirror level and mirroring is enabled.
2018-06-29 12:41:36 +00:00
NFSD_DEBUG(4, "nfsrv_pnfscreate: mirrorcnt=%d maxmirror=%d\n",
mirrorcnt, nfsrv_maxpnfsmirror);
/*
* For all mirrors that couldn't be created, fill in the
* *pf structure, but with an IP address == 0.0.0.0.
*/
tpf = pf + mirrorcnt;
for (i = mirrorcnt; i < nfsrv_maxpnfsmirror; i++, tpf++) {
*tpf = *pf;
tpf->dsf_sin.sin_family = AF_INET;
tpf->dsf_sin.sin_len = sizeof(struct sockaddr_in);
tpf->dsf_sin.sin_addr.s_addr = 0;
tpf->dsf_sin.sin_port = 0;
}
Fix LORs between vn_start_write() and vn_lock() in the pNFS server. When coding the pNFS server, I added several vn_start_write() calls done while the vnode was locked, not realizing I had introduced LORs and possible deadlock when an exported file system on the MDS is suspended. This patch fixes this by removing the added vn_start_write() calls and modifying the code so that the extant vn_start_write() call before the NFS RPC/operation is done when needed by the pNFS server. Flags are changed so that LayoutCommit and LayoutReturn now get a vn_start_write() done for them. When the pNFS server is enabled, the code now also changes the flags for Getattr, so that the vn_start_write() is done for Getattr, since it may need to do a vn_set_extattr(). The nfs_writerpc flag array was made global to the NFS server and renamed nfsrv_writerpc, which is consistent naming for globals in the NFS server. Thanks go to kib@ for reporting that doing vn_start_write() while the vnode is locked results in a LOR. This patch only affects the behaviour of the pNFS server.
2018-08-17 21:12:16 +00:00
error = vn_extattr_set(vp, IO_NODELOCKED,
EXTATTR_NAMESPACE_SYSTEM, "pnfsd.dsfile",
sizeof(*pf) * nfsrv_maxpnfsmirror, (char *)pf, p);
if (error == 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
error = vn_extattr_set(vp, IO_NODELOCKED,
Fix LORs between vn_start_write() and vn_lock() in the pNFS server. When coding the pNFS server, I added several vn_start_write() calls done while the vnode was locked, not realizing I had introduced LORs and possible deadlock when an exported file system on the MDS is suspended. This patch fixes this by removing the added vn_start_write() calls and modifying the code so that the extant vn_start_write() call before the NFS RPC/operation is done when needed by the pNFS server. Flags are changed so that LayoutCommit and LayoutReturn now get a vn_start_write() done for them. When the pNFS server is enabled, the code now also changes the flags for Getattr, so that the vn_start_write() is done for Getattr, since it may need to do a vn_set_extattr(). The nfs_writerpc flag array was made global to the NFS server and renamed nfsrv_writerpc, which is consistent naming for globals in the NFS server. Thanks go to kib@ for reporting that doing vn_start_write() while the vnode is locked results in a LOR. This patch only affects the behaviour of the pNFS server.
2018-08-17 21:12:16 +00:00
EXTATTR_NAMESPACE_SYSTEM, "pnfsd.dsattr",
sizeof(dsattr), (char *)&dsattr, p);
if (error != 0)
printf("pNFS: pnfscreate setextattr=%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("pNFS: pnfscreate=%d\n", error);
free(pf, M_TEMP);
free(dsc, M_TEMP);
}
Clean up the experimental NFS server replay cache when the module is unloaded. Reviewed by: rmacklem Approved by: zml (mentor)
2011-01-12 23:34:09 +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
/*
* Get the information needed to remove the pNFS Data Server file from the
* Metadata file. Upon success, ddvp is set non-NULL to the locked
* DS directory vnode. The caller must unlock *ddvp when done with it.
*/
static void
nfsrv_pnfsremovesetup(struct vnode *vp, NFSPROC_T *p, struct vnode **dvpp,
int *mirrorcntp, char *fname, fhandle_t *fhp)
{
struct vattr va;
struct ucred *tcred;
char *buf;
int buflen, error;
dvpp[0] = NULL;
/* If not an exported regular file or not a pNFS server, just return. */
if (vp->v_type != VREG || (vp->v_mount->mnt_flag & MNT_EXPORTED) == 0 ||
nfsrv_devidcnt == 0)
return;
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
/* Check to see if this is the last hard link. */
tcred = newnfs_getcred();
error = VOP_GETATTR(vp, &va, tcred);
NFSFREECRED(tcred);
if (error != 0) {
printf("pNFS: nfsrv_pnfsremovesetup getattr=%d\n", error);
return;
}
if (va.va_nlink > 1)
return;
error = nfsvno_getfh(vp, fhp, p);
if (error != 0) {
printf("pNFS: nfsrv_pnfsremovesetup getfh=%d\n", error);
return;
}
buflen = 1024;
buf = malloc(buflen, M_TEMP, M_WAITOK);
/* Get the directory vnode for the DS mount and the file handle. */
error = nfsrv_dsgetsockmnt(vp, 0, buf, &buflen, mirrorcntp, p, dvpp,
NULL, NULL, fname, NULL, NULL, NULL, NULL, NULL);
free(buf, M_TEMP);
if (error != 0)
printf("pNFS: nfsrv_pnfsremovesetup getsockmnt=%d\n", error);
}
/*
* Remove a DS data file for nfsrv_pnfsremove(). Called for each mirror.
* The arguments are in a structure, so that they can be passed through
* taskqueue for a kernel process to execute this function.
*/
struct nfsrvdsremove {
int done;
int inprog;
struct task tsk;
struct ucred *tcred;
struct vnode *dvp;
NFSPROC_T *p;
int err;
char fname[PNFS_FILENAME_LEN + 1];
};
static int
nfsrv_dsremove(struct vnode *dvp, char *fname, struct ucred *tcred,
NFSPROC_T *p)
{
struct nameidata named;
struct vnode *nvp;
char *bufp;
u_long *hashp;
int error;
error = NFSVOPLOCK(dvp, LK_EXCLUSIVE);
if (error != 0)
return (error);
named.ni_cnd.cn_nameiop = DELETE;
named.ni_cnd.cn_lkflags = LK_EXCLUSIVE | LK_RETRY;
named.ni_cnd.cn_cred = tcred;
named.ni_cnd.cn_thread = p;
named.ni_cnd.cn_flags = ISLASTCN | LOCKPARENT | LOCKLEAF | SAVENAME;
nfsvno_setpathbuf(&named, &bufp, &hashp);
named.ni_cnd.cn_nameptr = bufp;
named.ni_cnd.cn_namelen = strlen(fname);
strlcpy(bufp, fname, NAME_MAX);
NFSD_DEBUG(4, "nfsrv_pnfsremove: filename=%s\n", bufp);
error = VOP_LOOKUP(dvp, &nvp, &named.ni_cnd);
NFSD_DEBUG(4, "nfsrv_pnfsremove: aft LOOKUP=%d\n", error);
if (error == 0) {
error = VOP_REMOVE(dvp, nvp, &named.ni_cnd);
vput(nvp);
}
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(dvp);
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
nfsvno_relpathbuf(&named);
if (error != 0)
printf("pNFS: nfsrv_pnfsremove failed=%d\n", error);
return (error);
}
/*
* Start up the thread that will execute nfsrv_dsremove().
*/
static void
start_dsremove(void *arg, int pending)
{
struct nfsrvdsremove *dsrm;
dsrm = (struct nfsrvdsremove *)arg;
dsrm->err = nfsrv_dsremove(dsrm->dvp, dsrm->fname, dsrm->tcred,
dsrm->p);
dsrm->done = 1;
NFSD_DEBUG(4, "start_dsremove: err=%d\n", dsrm->err);
}
/*
* Remove a pNFS data file from a Data Server.
* nfsrv_pnfsremovesetup() must have been called before the MDS file was
* removed to set up the dvp and fill in the FH.
*/
static void
nfsrv_pnfsremove(struct vnode **dvp, int mirrorcnt, char *fname, fhandle_t *fhp,
NFSPROC_T *p)
{
struct ucred *tcred;
struct nfsrvdsremove *dsrm, *tdsrm;
struct nfsdevice *ds;
struct nfsmount *nmp;
int failpos, i, ret, timo;
tcred = newnfs_getcred();
dsrm = NULL;
if (mirrorcnt > 1)
dsrm = malloc(sizeof(*dsrm) * mirrorcnt - 1, M_TEMP, M_WAITOK);
/*
* Remove the file on each DS mirror, using kernel process(es) for the
* additional mirrors.
*/
failpos = -1;
for (tdsrm = dsrm, i = 0; i < mirrorcnt - 1; i++, tdsrm++) {
tdsrm->tcred = tcred;
tdsrm->p = p;
tdsrm->dvp = dvp[i];
strlcpy(tdsrm->fname, fname, PNFS_FILENAME_LEN + 1);
tdsrm->inprog = 0;
tdsrm->done = 0;
tdsrm->err = 0;
ret = EIO;
if (nfs_pnfsiothreads != 0) {
ret = nfs_pnfsio(start_dsremove, tdsrm);
NFSD_DEBUG(4, "nfsrv_pnfsremove: nfs_pnfsio=%d\n", ret);
}
if (ret != 0) {
ret = nfsrv_dsremove(dvp[i], fname, tcred, p);
if (failpos == -1 && nfsds_failerr(ret))
failpos = i;
}
}
ret = nfsrv_dsremove(dvp[mirrorcnt - 1], fname, tcred, p);
if (failpos == -1 && mirrorcnt > 1 && nfsds_failerr(ret))
failpos = mirrorcnt - 1;
timo = hz / 50; /* Wait for 20msec. */
if (timo < 1)
timo = 1;
/* Wait for kernel task(s) to complete. */
for (tdsrm = dsrm, i = 0; i < mirrorcnt - 1; i++, tdsrm++) {
while (tdsrm->inprog != 0 && tdsrm->done == 0)
tsleep(&tdsrm->tsk, PVFS, "srvdsrm", timo);
if (failpos == -1 && nfsds_failerr(tdsrm->err))
failpos = i;
}
/*
* If failpos has been set, that mirror has failed, so it needs
* to be disabled.
*/
if (failpos >= 0) {
nmp = VFSTONFS(dvp[failpos]->v_mount);
NFSLOCKMNT(nmp);
if ((nmp->nm_privflag & (NFSMNTP_FORCEDISM |
NFSMNTP_CANCELRPCS)) == 0) {
nmp->nm_privflag |= NFSMNTP_CANCELRPCS;
NFSUNLOCKMNT(nmp);
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(PNFSDOP_DELDSSERVER, 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, "dsremovefail fail=%d ds=%p\n", failpos,
ds);
if (ds != NULL)
nfsrv_killrpcs(nmp);
NFSLOCKMNT(nmp);
nmp->nm_privflag &= ~NFSMNTP_CANCELRPCS;
wakeup(nmp);
}
NFSUNLOCKMNT(nmp);
}
/* Get rid all layouts for the file. */
nfsrv_freefilelayouts(fhp);
NFSFREECRED(tcred);
free(dsrm, M_TEMP);
}
/*
* Generate a file name based on the file handle and put it in *bufp.
* Return the number of bytes generated.
*/
static int
nfsrv_putfhname(fhandle_t *fhp, char *bufp)
{
int i;
uint8_t *cp;
const uint8_t *hexdigits = "0123456789abcdef";
cp = (uint8_t *)fhp;
for (i = 0; i < sizeof(*fhp); i++) {
bufp[2 * i] = hexdigits[(*cp >> 4) & 0xf];
bufp[2 * i + 1] = hexdigits[*cp++ & 0xf];
}
bufp[2 * i] = '\0';
return (2 * i);
}
/*
* Update the Metadata file's attributes from the DS file when a Read/Write
* layout is returned.
* Basically just call nfsrv_proxyds() with procedure == NFSPROC_LAYOUTRETURN
* so that it does a nfsrv_getattrdsrpc() and nfsrv_setextattr() on the DS file.
*/
int
nfsrv_updatemdsattr(struct vnode *vp, struct nfsvattr *nap, NFSPROC_T *p)
{
struct ucred *tcred;
int error;
/* Do this as root so that it won't fail with EACCES. */
tcred = newnfs_getcred();
Delete the unused "nd" argument for nfsrv_proxyds(). The "nd" argument for nfsrv_proxyds() is no longer used by the function. This patch deletes it. This allows a subsequent patch to delete the "nd" argument from nfsvno_getattr(), since it's only use of "nd" was to pass it 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.
2019-09-05 22:25:19 +00:00
error = nfsrv_proxyds(vp, 0, 0, tcred, p, NFSPROC_LAYOUTRETURN,
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
NULL, NULL, NULL, nap, NULL, NULL, 0, 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
NFSFREECRED(tcred);
return (error);
}
/*
* Set the NFSv4 ACL on the DS file to the same ACL as the MDS file.
*/
static int
nfsrv_dssetacl(struct vnode *vp, struct acl *aclp, struct ucred *cred,
NFSPROC_T *p)
{
int error;
Delete the unused "nd" argument for nfsrv_proxyds(). The "nd" argument for nfsrv_proxyds() is no longer used by the function. This patch deletes it. This allows a subsequent patch to delete the "nd" argument from nfsvno_getattr(), since it's only use of "nd" was to pass it 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.
2019-09-05 22:25:19 +00:00
error = nfsrv_proxyds(vp, 0, 0, cred, p, NFSPROC_SETACL,
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
NULL, NULL, NULL, NULL, aclp, NULL, 0, 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 (error);
}
static int
Delete the unused "nd" argument for nfsrv_proxyds(). The "nd" argument for nfsrv_proxyds() is no longer used by the function. This patch deletes it. This allows a subsequent patch to delete the "nd" argument from nfsvno_getattr(), since it's only use of "nd" was to pass it 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.
2019-09-05 22:25:19 +00:00
nfsrv_proxyds(struct vnode *vp, off_t off, int cnt, struct ucred *cred,
struct thread *p, int ioproc, struct mbuf **mpp, char *cp,
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 mbuf **mpp2, struct nfsvattr *nap, struct acl *aclp,
off_t *offp, int content, bool *eofp)
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 nfsmount *nmp[NFSDEV_MAXMIRRORS], *failnmp;
fhandle_t fh[NFSDEV_MAXMIRRORS];
struct vnode *dvp[NFSDEV_MAXMIRRORS];
struct nfsdevice *ds;
struct pnfsdsattr dsattr;
Fix the pNFS server's reporting of SpaceUsed (va_bytes). The pNFS server currently reports SpaceUsed (va_bytes) for the metadata file. This in not correct, since the metadata file is always empty and, as such, va_bytes is just the allocation for the empty file. This patch adds va_bytes to the list of attributes acquired from the DS for a file, so that it includes the allocated data size and is updated when the file is written. For files created on a pNFS server before this patch is applied, the va_bytes value is estimated by rounding va_size up to a multiple of BLKDEV_IOSIZE. Once the file is written after this patch has been applied to the metadata server, the va_bytes returned for the file will be correct. This patch only affects a pNFS metadata server. Found during testing of the NFSv4.2 pNFS server for the Allocate operation. (Not yet in head/current.) MFC after: 2 weeks
2019-11-22 00:22:55 +00:00
struct opnfsdsattr odsattr;
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
char *buf;
int buflen, error, failpos, i, mirrorcnt, origmircnt, trycnt;
NFSD_DEBUG(4, "in nfsrv_proxyds\n");
/*
* If not a regular file, not exported or not a pNFS server,
* just return ENOENT.
*/
if (vp->v_type != VREG || (vp->v_mount->mnt_flag & MNT_EXPORTED) == 0 ||
nfsrv_devidcnt == 0)
return (ENOENT);
buflen = 1024;
buf = malloc(buflen, M_TEMP, M_WAITOK);
error = 0;
/*
* For Getattr, get the Change attribute (va_filerev) and size (va_size)
* from the MetaData file's extended attribute.
*/
if (ioproc == NFSPROC_GETATTR) {
error = vn_extattr_get(vp, IO_NODELOCKED,
EXTATTR_NAMESPACE_SYSTEM, "pnfsd.dsattr", &buflen, buf,
p);
if (error == 0) {
Fix the pNFS server's reporting of SpaceUsed (va_bytes). The pNFS server currently reports SpaceUsed (va_bytes) for the metadata file. This in not correct, since the metadata file is always empty and, as such, va_bytes is just the allocation for the empty file. This patch adds va_bytes to the list of attributes acquired from the DS for a file, so that it includes the allocated data size and is updated when the file is written. For files created on a pNFS server before this patch is applied, the va_bytes value is estimated by rounding va_size up to a multiple of BLKDEV_IOSIZE. Once the file is written after this patch has been applied to the metadata server, the va_bytes returned for the file will be correct. This patch only affects a pNFS metadata server. Found during testing of the NFSv4.2 pNFS server for the Allocate operation. (Not yet in head/current.) MFC after: 2 weeks
2019-11-22 00:22:55 +00:00
if (buflen == sizeof(odsattr)) {
NFSBCOPY(buf, &odsattr, buflen);
nap->na_filerev = odsattr.dsa_filerev;
nap->na_size = odsattr.dsa_size;
nap->na_atime = odsattr.dsa_atime;
nap->na_mtime = odsattr.dsa_mtime;
/*
* Fake na_bytes by rounding up na_size.
* Since we don't know the block size, just
* use BLKDEV_IOSIZE.
*/
nap->na_bytes = (odsattr.dsa_size +
BLKDEV_IOSIZE - 1) & ~(BLKDEV_IOSIZE - 1);
} else if (buflen == sizeof(dsattr)) {
NFSBCOPY(buf, &dsattr, buflen);
nap->na_filerev = dsattr.dsa_filerev;
nap->na_size = dsattr.dsa_size;
nap->na_atime = dsattr.dsa_atime;
nap->na_mtime = dsattr.dsa_mtime;
nap->na_bytes = dsattr.dsa_bytes;
} else
error = ENXIO;
}
if (error == 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
/*
* If nfsrv_pnfsgetdsattr is 0 or nfsrv_checkdsattr()
* returns 0, just return now. nfsrv_checkdsattr()
* returns 0 if there is no Read/Write layout
* plus either an Open/Write_access or Write
* delegation issued to a client for the file.
*/
if (nfsrv_pnfsgetdsattr == 0 ||
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(vp, p) == 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
free(buf, M_TEMP);
return (error);
}
}
/*
* Clear ENOATTR so the code below will attempt to do a
* nfsrv_getattrdsrpc() to get the attributes and (re)create
* the extended attribute.
*/
if (error == ENOATTR)
error = 0;
}
origmircnt = -1;
trycnt = 0;
tryagain:
if (error == 0) {
buflen = 1024;
Fix LORs between vn_start_write() and vn_lock() in the pNFS server. When coding the pNFS server, I added several vn_start_write() calls done while the vnode was locked, not realizing I had introduced LORs and possible deadlock when an exported file system on the MDS is suspended. This patch fixes this by removing the added vn_start_write() calls and modifying the code so that the extant vn_start_write() call before the NFS RPC/operation is done when needed by the pNFS server. Flags are changed so that LayoutCommit and LayoutReturn now get a vn_start_write() done for them. When the pNFS server is enabled, the code now also changes the flags for Getattr, so that the vn_start_write() is done for Getattr, since it may need to do a vn_set_extattr(). The nfs_writerpc flag array was made global to the NFS server and renamed nfsrv_writerpc, which is consistent naming for globals in the NFS server. Thanks go to kib@ for reporting that doing vn_start_write() while the vnode is locked results in a LOR. This patch only affects the behaviour of the pNFS server.
2018-08-17 21:12:16 +00:00
if (ioproc == NFSPROC_READDS && NFSVOPISLOCKED(vp) ==
LK_EXCLUSIVE)
printf("nfsrv_proxyds: Readds vp exclusively locked\n");
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 = nfsrv_dsgetsockmnt(vp, LK_SHARED, buf, &buflen,
&mirrorcnt, p, dvp, fh, NULL, NULL, NULL, NULL, NULL,
NULL, NULL);
if (error == 0) {
for (i = 0; i < mirrorcnt; i++)
nmp[i] = VFSTONFS(dvp[i]->v_mount);
} else
printf("pNFS: proxy getextattr sockaddr=%d\n", error);
} else
printf("pNFS: nfsrv_dsgetsockmnt=%d\n", error);
if (error == 0) {
failpos = -1;
if (origmircnt == -1)
origmircnt = mirrorcnt;
/*
* If failpos is set to a mirror#, then that mirror has
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
* failed and will be disabled. For Read, Getattr and Seek, the
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
* function only tries one mirror, so if that mirror has
* failed, it will need to be retried. As such, increment
* tryitagain for these cases.
* For Write, Setattr and Setacl, the function tries all
* mirrors and will not return an error for the case where
* one mirror has failed. For these cases, the functioning
* mirror(s) will have been modified, so a retry isn't
* necessary. These functions will set failpos for the
* failed mirror#.
*/
if (ioproc == NFSPROC_READDS) {
error = nfsrv_readdsrpc(fh, off, cnt, cred, p, nmp[0],
mpp, mpp2);
if (nfsds_failerr(error) && mirrorcnt > 1) {
/*
* Setting failpos will cause the mirror
* to be disabled and then a retry of this
* read is required.
*/
failpos = 0;
error = 0;
trycnt++;
}
} else if (ioproc == NFSPROC_WRITEDS)
error = nfsrv_writedsrpc(fh, off, cnt, cred, p, vp,
&nmp[0], mirrorcnt, mpp, cp, &failpos);
else if (ioproc == NFSPROC_SETATTR)
error = nfsrv_setattrdsrpc(fh, cred, p, vp, &nmp[0],
mirrorcnt, nap, &failpos);
else if (ioproc == NFSPROC_SETACL)
error = nfsrv_setacldsrpc(fh, cred, p, vp, &nmp[0],
mirrorcnt, aclp, &failpos);
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
else if (ioproc == NFSPROC_SEEKDS) {
error = nfsrv_seekdsrpc(fh, offp, content, eofp, cred,
p, nmp[0]);
if (nfsds_failerr(error) && mirrorcnt > 1) {
/*
* Setting failpos will cause the mirror
* to be disabled and then a retry of this
* read is required.
*/
failpos = 0;
error = 0;
trycnt++;
}
} else if (ioproc == NFSPROC_ALLOCATE)
error = nfsrv_allocatedsrpc(fh, off, *offp, cred, p, vp,
&nmp[0], mirrorcnt, &failpos);
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 {
error = nfsrv_getattrdsrpc(&fh[mirrorcnt - 1], cred, p,
vp, nmp[mirrorcnt - 1], nap);
if (nfsds_failerr(error) && mirrorcnt > 1) {
/*
* Setting failpos will cause the mirror
* to be disabled and then a retry of this
* getattr is required.
*/
failpos = mirrorcnt - 1;
error = 0;
trycnt++;
}
}
ds = NULL;
if (failpos >= 0) {
failnmp = nmp[failpos];
NFSLOCKMNT(failnmp);
if ((failnmp->nm_privflag & (NFSMNTP_FORCEDISM |
NFSMNTP_CANCELRPCS)) == 0) {
failnmp->nm_privflag |= NFSMNTP_CANCELRPCS;
NFSUNLOCKMNT(failnmp);
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(PNFSDOP_DELDSSERVER,
failnmp, 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, "dsldsnmp fail=%d ds=%p\n",
failpos, ds);
if (ds != NULL)
nfsrv_killrpcs(failnmp);
NFSLOCKMNT(failnmp);
failnmp->nm_privflag &= ~NFSMNTP_CANCELRPCS;
wakeup(failnmp);
}
NFSUNLOCKMNT(failnmp);
}
for (i = 0; i < mirrorcnt; i++)
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(dvp[i]);
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_proxyds: aft RPC=%d trya=%d\n", error,
trycnt);
/* Try the Read/Getattr again if a mirror was deleted. */
if (ds != NULL && trycnt > 0 && trycnt < origmircnt)
goto tryagain;
} else {
/* Return ENOENT for any Extended Attribute error. */
error = ENOENT;
}
free(buf, M_TEMP);
NFSD_DEBUG(4, "nfsrv_proxyds: error=%d\n", error);
return (error);
}
/*
* Get the DS mount point, fh and directory from the "pnfsd.dsfile" extended
* attribute.
* newnmpp - If it points to a non-NULL nmp, that is the destination and needs
* to be checked. If it points to a NULL nmp, then it returns
* a suitable destination.
* curnmp - If non-NULL, it is the source mount for the copy.
*/
int
nfsrv_dsgetsockmnt(struct vnode *vp, int lktype, char *buf, int *buflenp,
int *mirrorcntp, NFSPROC_T *p, struct vnode **dvpp, fhandle_t *fhp,
char *devid, char *fnamep, struct vnode **nvpp, struct nfsmount **newnmpp,
struct nfsmount *curnmp, int *ippos, int *dsdirp)
{
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 vnode *dvp, *nvp = NULL, **tdvpp;
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
struct nfsmount *nmp, *newnmp;
struct sockaddr *sad;
struct sockaddr_in *sin;
Fix handling of the hybrid DS case for a pNFS server. After the addition of the "#mds_path" suffix for a DS specification on the "-p" nfsd option, it is possible to have a mix of DSs assigned to an MDS file system and DSs that store files for all DSs. This is what I referred to as "hybrid" above. At first, I didn't think this hybrid case would be useful, but I now believe that some system administrators may fine it useful. This patch modifies the file storage assignment algorithm so that it makes the "#mds_path" DSs take priority and the all file systems DSs are now only used for MDS file systems with no "#mds_path" DS servers. This only affects the pNFS server for this "hybrid" case.
2018-07-07 19:27:49 +00:00
struct nfsdevice *ds, *tds, *fndds;
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 pnfsdsfile *pf;
uint32_t dsdir;
int error, fhiszero, fnd, gotone, i, mirrorcnt;
ASSERT_VOP_LOCKED(vp, "nfsrv_dsgetsockmnt vp");
*mirrorcntp = 1;
tdvpp = dvpp;
if (nvpp != NULL)
*nvpp = NULL;
if (dvpp != NULL)
*dvpp = NULL;
if (ippos != NULL)
*ippos = -1;
if (newnmpp != NULL)
newnmp = *newnmpp;
else
newnmp = 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
mp = vp->v_mount;
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 = vn_extattr_get(vp, IO_NODELOCKED, EXTATTR_NAMESPACE_SYSTEM,
"pnfsd.dsfile", buflenp, buf, p);
mirrorcnt = *buflenp / sizeof(*pf);
if (error == 0 && (mirrorcnt < 1 || mirrorcnt > NFSDEV_MAXMIRRORS ||
*buflenp != sizeof(*pf) * mirrorcnt))
error = ENOATTR;
pf = (struct pnfsdsfile *)buf;
/* If curnmp != NULL, check for a match in the mirror list. */
if (curnmp != NULL && error == 0) {
fnd = 0;
for (i = 0; i < mirrorcnt; i++, pf++) {
sad = (struct sockaddr *)&pf->dsf_sin;
if (nfsaddr2_match(sad, curnmp->nm_nam)) {
if (ippos != NULL)
*ippos = i;
fnd = 1;
break;
}
}
if (fnd == 0)
error = ENXIO;
}
gotone = 0;
pf = (struct pnfsdsfile *)buf;
NFSD_DEBUG(4, "nfsrv_dsgetsockmnt: mirrorcnt=%d err=%d\n", mirrorcnt,
error);
for (i = 0; i < mirrorcnt && error == 0; i++, pf++) {
fhiszero = 0;
sad = (struct sockaddr *)&pf->dsf_sin;
sin = &pf->dsf_sin;
dsdir = pf->dsf_dir;
if (dsdir >= nfsrv_dsdirsize) {
printf("nfsrv_dsgetsockmnt: dsdir=%d\n", dsdir);
error = ENOATTR;
} else if (nvpp != NULL && newnmp != NULL &&
nfsaddr2_match(sad, newnmp->nm_nam))
error = EEXIST;
if (error == 0) {
if (ippos != NULL && curnmp == NULL &&
sad->sa_family == AF_INET &&
sin->sin_addr.s_addr == 0)
*ippos = i;
if (NFSBCMP(&zerofh, &pf->dsf_fh, sizeof(zerofh)) == 0)
fhiszero = 1;
/* Use the socket address to find the mount point. */
fndds = NULL;
NFSDDSLOCK();
Fix handling of the hybrid DS case for a pNFS server. After the addition of the "#mds_path" suffix for a DS specification on the "-p" nfsd option, it is possible to have a mix of DSs assigned to an MDS file system and DSs that store files for all DSs. This is what I referred to as "hybrid" above. At first, I didn't think this hybrid case would be useful, but I now believe that some system administrators may fine it useful. This patch modifies the file storage assignment algorithm so that it makes the "#mds_path" DSs take priority and the all file systems DSs are now only used for MDS file systems with no "#mds_path" DS servers. This only affects the pNFS server for this "hybrid" case.
2018-07-07 19:27:49 +00:00
/* Find a match for the IP address. */
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_FOREACH(ds, &nfsrv_devidhead, nfsdev_list) {
if (ds->nfsdev_nmp != NULL) {
dvp = ds->nfsdev_dvp;
nmp = VFSTONFS(dvp->v_mount);
if (nmp != ds->nfsdev_nmp)
printf("different2 nmp %p %p\n",
nmp, ds->nfsdev_nmp);
Fix handling of the hybrid DS case for a pNFS server. After the addition of the "#mds_path" suffix for a DS specification on the "-p" nfsd option, it is possible to have a mix of DSs assigned to an MDS file system and DSs that store files for all DSs. This is what I referred to as "hybrid" above. At first, I didn't think this hybrid case would be useful, but I now believe that some system administrators may fine it useful. This patch modifies the file storage assignment algorithm so that it makes the "#mds_path" DSs take priority and the all file systems DSs are now only used for MDS file systems with no "#mds_path" DS servers. This only affects the pNFS server for this "hybrid" case.
2018-07-07 19:27:49 +00:00
if (nfsaddr2_match(sad, nmp->nm_nam)) {
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
fndds = ds;
Fix handling of the hybrid DS case for a pNFS server. After the addition of the "#mds_path" suffix for a DS specification on the "-p" nfsd option, it is possible to have a mix of DSs assigned to an MDS file system and DSs that store files for all DSs. This is what I referred to as "hybrid" above. At first, I didn't think this hybrid case would be useful, but I now believe that some system administrators may fine it useful. This patch modifies the file storage assignment algorithm so that it makes the "#mds_path" DSs take priority and the all file systems DSs are now only used for MDS file systems with no "#mds_path" DS servers. This only affects the pNFS server for this "hybrid" case.
2018-07-07 19:27:49 +00:00
break;
}
}
}
if (fndds != NULL && newnmpp != NULL &&
newnmp == NULL) {
/* Search for a place to make a mirror copy. */
TAILQ_FOREACH(tds, &nfsrv_devidhead,
nfsdev_list) {
if (tds->nfsdev_nmp != NULL &&
fndds != tds &&
((tds->nfsdev_mdsisset == 0 &&
fndds->nfsdev_mdsisset == 0) ||
(tds->nfsdev_mdsisset != 0 &&
fndds->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(&tds->nfsdev_mdsfsid,
&mp->mnt_stat.f_fsid) == 0))) {
Fix handling of the hybrid DS case for a pNFS server. After the addition of the "#mds_path" suffix for a DS specification on the "-p" nfsd option, it is possible to have a mix of DSs assigned to an MDS file system and DSs that store files for all DSs. This is what I referred to as "hybrid" above. At first, I didn't think this hybrid case would be useful, but I now believe that some system administrators may fine it useful. This patch modifies the file storage assignment algorithm so that it makes the "#mds_path" DSs take priority and the all file systems DSs are now only used for MDS file systems with no "#mds_path" DS servers. This only affects the pNFS server for this "hybrid" case.
2018-07-07 19:27:49 +00:00
*newnmpp = tds->nfsdev_nmp;
break;
}
}
if (tds != NULL) {
/*
* Move this entry to the end of the
* list, so it won't be selected as
* easily the next time.
*/
TAILQ_REMOVE(&nfsrv_devidhead, tds,
nfsdev_list);
TAILQ_INSERT_TAIL(&nfsrv_devidhead, tds,
nfsdev_list);
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 (fndds != NULL) {
dvp = fndds->nfsdev_dsdir[dsdir];
if (lktype != 0 || fhiszero != 0 ||
(nvpp != NULL && *nvpp == NULL)) {
if (fhiszero != 0)
error = vn_lock(dvp,
LK_EXCLUSIVE);
else if (lktype != 0)
error = vn_lock(dvp, lktype);
else
error = vn_lock(dvp, LK_SHARED);
/*
* If the file handle is all 0's, try to
* do a Lookup against the DS to acquire
* it.
* If dvpp == NULL or the Lookup fails,
* unlock dvp after the call.
*/
if (error == 0 && (fhiszero != 0 ||
(nvpp != NULL && *nvpp == NULL))) {
error = nfsrv_pnfslookupds(vp,
dvp, pf, &nvp, p);
if (error == 0) {
if (fhiszero != 0)
nfsrv_pnfssetfh(
vp, pf,
Fix LORs between vn_start_write() and vn_lock() in the pNFS server. When coding the pNFS server, I added several vn_start_write() calls done while the vnode was locked, not realizing I had introduced LORs and possible deadlock when an exported file system on the MDS is suspended. This patch fixes this by removing the added vn_start_write() calls and modifying the code so that the extant vn_start_write() call before the NFS RPC/operation is done when needed by the pNFS server. Flags are changed so that LayoutCommit and LayoutReturn now get a vn_start_write() done for them. When the pNFS server is enabled, the code now also changes the flags for Getattr, so that the vn_start_write() is done for Getattr, since it may need to do a vn_set_extattr(). The nfs_writerpc flag array was made global to the NFS server and renamed nfsrv_writerpc, which is consistent naming for globals in the NFS server. Thanks go to kib@ for reporting that doing vn_start_write() while the vnode is locked results in a LOR. This patch only affects the behaviour of the pNFS server.
2018-08-17 21:12:16 +00:00
devid,
fnamep,
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
nvp, p);
if (nvpp != NULL &&
*nvpp == NULL) {
*nvpp = nvp;
*dsdirp = dsdir;
} else
vput(nvp);
}
if (error != 0 || lktype == 0)
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(dvp);
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) {
gotone++;
NFSD_DEBUG(4, "gotone=%d\n", gotone);
if (devid != NULL) {
NFSBCOPY(fndds->nfsdev_deviceid,
devid, NFSX_V4DEVICEID);
devid += NFSX_V4DEVICEID;
}
if (dvpp != NULL)
*tdvpp++ = dvp;
if (fhp != NULL)
NFSBCOPY(&pf->dsf_fh, fhp++,
NFSX_MYFH);
if (fnamep != NULL && gotone == 1)
strlcpy(fnamep,
pf->dsf_filename,
sizeof(pf->dsf_filename));
} else
NFSD_DEBUG(4, "nfsrv_dsgetsockmnt "
"err=%d\n", error);
}
}
}
if (error == 0 && gotone == 0)
error = ENOENT;
NFSD_DEBUG(4, "eo nfsrv_dsgetsockmnt: gotone=%d err=%d\n", gotone,
error);
if (error == 0)
*mirrorcntp = gotone;
else {
if (gotone > 0 && dvpp != NULL) {
/*
* If the error didn't occur on the first one and
* dvpp != NULL, the one(s) prior to the failure will
* have locked dvp's that need to be unlocked.
*/
for (i = 0; i < gotone; i++) {
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(*dvpp);
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
*dvpp++ = NULL;
}
}
/*
* If it found the vnode to be copied from before a failure,
* it needs to be vput()'d.
*/
if (nvpp != NULL && *nvpp != NULL) {
vput(*nvpp);
*nvpp = NULL;
}
}
return (error);
}
/*
* Set the extended attribute for the Change attribute.
*/
static int
nfsrv_setextattr(struct vnode *vp, struct nfsvattr *nap, NFSPROC_T *p)
{
struct pnfsdsattr dsattr;
int error;
ASSERT_VOP_ELOCKED(vp, "nfsrv_setextattr vp");
Fix LORs between vn_start_write() and vn_lock() in the pNFS server. When coding the pNFS server, I added several vn_start_write() calls done while the vnode was locked, not realizing I had introduced LORs and possible deadlock when an exported file system on the MDS is suspended. This patch fixes this by removing the added vn_start_write() calls and modifying the code so that the extant vn_start_write() call before the NFS RPC/operation is done when needed by the pNFS server. Flags are changed so that LayoutCommit and LayoutReturn now get a vn_start_write() done for them. When the pNFS server is enabled, the code now also changes the flags for Getattr, so that the vn_start_write() is done for Getattr, since it may need to do a vn_set_extattr(). The nfs_writerpc flag array was made global to the NFS server and renamed nfsrv_writerpc, which is consistent naming for globals in the NFS server. Thanks go to kib@ for reporting that doing vn_start_write() while the vnode is locked results in a LOR. This patch only affects the behaviour of the pNFS server.
2018-08-17 21:12:16 +00:00
dsattr.dsa_filerev = nap->na_filerev;
dsattr.dsa_size = nap->na_size;
dsattr.dsa_atime = nap->na_atime;
dsattr.dsa_mtime = nap->na_mtime;
Fix the pNFS server's reporting of SpaceUsed (va_bytes). The pNFS server currently reports SpaceUsed (va_bytes) for the metadata file. This in not correct, since the metadata file is always empty and, as such, va_bytes is just the allocation for the empty file. This patch adds va_bytes to the list of attributes acquired from the DS for a file, so that it includes the allocated data size and is updated when the file is written. For files created on a pNFS server before this patch is applied, the va_bytes value is estimated by rounding va_size up to a multiple of BLKDEV_IOSIZE. Once the file is written after this patch has been applied to the metadata server, the va_bytes returned for the file will be correct. This patch only affects a pNFS metadata server. Found during testing of the NFSv4.2 pNFS server for the Allocate operation. (Not yet in head/current.) MFC after: 2 weeks
2019-11-22 00:22:55 +00:00
dsattr.dsa_bytes = nap->na_bytes;
Fix LORs between vn_start_write() and vn_lock() in the pNFS server. When coding the pNFS server, I added several vn_start_write() calls done while the vnode was locked, not realizing I had introduced LORs and possible deadlock when an exported file system on the MDS is suspended. This patch fixes this by removing the added vn_start_write() calls and modifying the code so that the extant vn_start_write() call before the NFS RPC/operation is done when needed by the pNFS server. Flags are changed so that LayoutCommit and LayoutReturn now get a vn_start_write() done for them. When the pNFS server is enabled, the code now also changes the flags for Getattr, so that the vn_start_write() is done for Getattr, since it may need to do a vn_set_extattr(). The nfs_writerpc flag array was made global to the NFS server and renamed nfsrv_writerpc, which is consistent naming for globals in the NFS server. Thanks go to kib@ for reporting that doing vn_start_write() while the vnode is locked results in a LOR. This patch only affects the behaviour of the pNFS server.
2018-08-17 21:12:16 +00:00
error = vn_extattr_set(vp, IO_NODELOCKED, EXTATTR_NAMESPACE_SYSTEM,
"pnfsd.dsattr", sizeof(dsattr), (char *)&dsattr, 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 (error != 0)
printf("pNFS: setextattr=%d\n", error);
return (error);
}
static int
nfsrv_readdsrpc(fhandle_t *fhp, off_t off, int len, struct ucred *cred,
NFSPROC_T *p, struct nfsmount *nmp, struct mbuf **mpp, struct mbuf **mpendp)
{
uint32_t *tl;
struct nfsrv_descript *nd;
nfsv4stateid_t st;
struct mbuf *m, *m2;
int error = 0, retlen, tlen, trimlen;
NFSD_DEBUG(4, "in nfsrv_readdsrpc\n");
nd = malloc(sizeof(*nd), M_TEMP, M_WAITOK | M_ZERO);
*mpp = NULL;
/*
* Use a stateid where other is an alternating 01010 pattern and
* seqid is 0xffffffff. This value is not defined as special by
* the RFC and is used by the FreeBSD NFS server to indicate an
* MDS->DS proxy operation.
*/
st.other[0] = 0x55555555;
st.other[1] = 0x55555555;
st.other[2] = 0x55555555;
st.seqid = 0xffffffff;
nfscl_reqstart(nd, NFSPROC_READDS, nmp, (u_int8_t *)fhp, sizeof(*fhp),
Delete the unused "use_ext" argument to nfscl_reqstart(). This is a partial revert of r363210, since the "use_ext" argument added by that commit is not actually useful. This patch should not result in any semantics change.
2020-08-18 01:41:12 +00:00
NULL, NULL, 0, 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_stateidtom(nd, &st, NFSSTATEID_PUTSTATEID);
NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED * 3);
txdr_hyper(off, tl);
*(tl + 2) = txdr_unsigned(len);
error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
if (error != 0) {
free(nd, M_TEMP);
return (error);
}
if (nd->nd_repstat == 0) {
NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
NFSM_STRSIZ(retlen, len);
if (retlen > 0) {
/* Trim off the pre-data XDR from the mbuf chain. */
m = nd->nd_mrep;
while (m != NULL && m != nd->nd_md) {
if (m->m_next == nd->nd_md) {
m->m_next = NULL;
m_freem(nd->nd_mrep);
nd->nd_mrep = m = nd->nd_md;
} else
m = m->m_next;
}
if (m == NULL) {
printf("nfsrv_readdsrpc: busted mbuf list\n");
error = ENOENT;
goto nfsmout;
}
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
/*
* Now, adjust first mbuf so that any XDR before the
* read data is skipped over.
*/
trimlen = nd->nd_dpos - mtod(m, char *);
if (trimlen > 0) {
m->m_len -= trimlen;
NFSM_DATAP(m, trimlen);
}
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
/*
* Truncate the mbuf chain at retlen bytes of data,
* plus XDR padding that brings the length up to a
* multiple of 4.
*/
tlen = NFSM_RNDUP(retlen);
do {
if (m->m_len >= tlen) {
m->m_len = tlen;
tlen = 0;
m2 = m->m_next;
m->m_next = NULL;
m_freem(m2);
break;
}
tlen -= m->m_len;
m = m->m_next;
} while (m != NULL);
if (tlen > 0) {
printf("nfsrv_readdsrpc: busted mbuf list\n");
error = ENOENT;
goto nfsmout;
}
*mpp = nd->nd_mrep;
*mpendp = m;
nd->nd_mrep = NULL;
}
} else
error = nd->nd_repstat;
nfsmout:
/* If nd->nd_mrep is already NULL, this is a no-op. */
m_freem(nd->nd_mrep);
free(nd, M_TEMP);
NFSD_DEBUG(4, "nfsrv_readdsrpc error=%d\n", error);
return (error);
}
/*
* Do a write RPC on a DS data file, using this structure for the arguments,
* so that this function can be executed by a separate kernel process.
*/
struct nfsrvwritedsdorpc {
int done;
int inprog;
struct task tsk;
fhandle_t fh;
off_t off;
int len;
struct nfsmount *nmp;
struct ucred *cred;
NFSPROC_T *p;
struct mbuf *m;
int err;
};
static int
nfsrv_writedsdorpc(struct nfsmount *nmp, fhandle_t *fhp, off_t off, int len,
struct nfsvattr *nap, struct mbuf *m, struct ucred *cred, NFSPROC_T *p)
{
uint32_t *tl;
struct nfsrv_descript *nd;
nfsattrbit_t attrbits;
nfsv4stateid_t st;
int commit, error, retlen;
nd = malloc(sizeof(*nd), M_TEMP, M_WAITOK | M_ZERO);
nfscl_reqstart(nd, NFSPROC_WRITE, nmp, (u_int8_t *)fhp,
Delete the unused "use_ext" argument to nfscl_reqstart(). This is a partial revert of r363210, since the "use_ext" argument added by that commit is not actually useful. This patch should not result in any semantics change.
2020-08-18 01:41:12 +00:00
sizeof(fhandle_t), NULL, NULL, 0, 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
/*
* Use a stateid where other is an alternating 01010 pattern and
* seqid is 0xffffffff. This value is not defined as special by
* the RFC and is used by the FreeBSD NFS server to indicate an
* MDS->DS proxy operation.
*/
st.other[0] = 0x55555555;
st.other[1] = 0x55555555;
st.other[2] = 0x55555555;
st.seqid = 0xffffffff;
nfsm_stateidtom(nd, &st, NFSSTATEID_PUTSTATEID);
NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
txdr_hyper(off, tl);
tl += 2;
/*
* Do all writes FileSync, since the server doesn't hold onto dirty
* buffers. Since clients should be accessing the DS servers directly
* using the pNFS layouts, this just needs to work correctly as a
* fallback.
*/
*tl++ = txdr_unsigned(NFSWRITE_FILESYNC);
*tl = txdr_unsigned(len);
NFSD_DEBUG(4, "nfsrv_writedsdorpc: len=%d\n", len);
/* Put data in mbuf chain. */
nd->nd_mb->m_next = m;
/* Set nd_mb and nd_bpos to end of data. */
while (m->m_next != NULL)
m = m->m_next;
nd->nd_mb = m;
Add a function nfsm_set() to initialize "struct nfsrv_descript" for building mbuf lists. This function is currently trivial, but will that will change when support for building NFS messages in ext_pgs mbufs is added. Adding support for ext_pgs mbufs is needed for KERN_TLS, which will be used to implement nfs-over-tls.
2020-05-18 00:07:45 +00:00
nfsm_set(nd, m->m_len);
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_writedsdorpc: lastmb len=%d\n", m->m_len);
Fix the pNFS server's reporting of SpaceUsed (va_bytes). The pNFS server currently reports SpaceUsed (va_bytes) for the metadata file. This in not correct, since the metadata file is always empty and, as such, va_bytes is just the allocation for the empty file. This patch adds va_bytes to the list of attributes acquired from the DS for a file, so that it includes the allocated data size and is updated when the file is written. For files created on a pNFS server before this patch is applied, the va_bytes value is estimated by rounding va_size up to a multiple of BLKDEV_IOSIZE. Once the file is written after this patch has been applied to the metadata server, the va_bytes returned for the file will be correct. This patch only affects a pNFS metadata server. Found during testing of the NFSv4.2 pNFS server for the Allocate operation. (Not yet in head/current.) MFC after: 2 weeks
2019-11-22 00:22:55 +00:00
/* Do a Getattr for the attributes that change upon writing. */
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
NFSZERO_ATTRBIT(&attrbits);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_SIZE);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEACCESS);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
Fix the pNFS server's reporting of SpaceUsed (va_bytes). The pNFS server currently reports SpaceUsed (va_bytes) for the metadata file. This in not correct, since the metadata file is always empty and, as such, va_bytes is just the allocation for the empty file. This patch adds va_bytes to the list of attributes acquired from the DS for a file, so that it includes the allocated data size and is updated when the file is written. For files created on a pNFS server before this patch is applied, the va_bytes value is estimated by rounding va_size up to a multiple of BLKDEV_IOSIZE. Once the file is written after this patch has been applied to the metadata server, the va_bytes returned for the file will be correct. This patch only affects a pNFS metadata server. Found during testing of the NFSv4.2 pNFS server for the Allocate operation. (Not yet in head/current.) MFC after: 2 weeks
2019-11-22 00:22:55 +00:00
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_SPACEUSED);
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 *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
(void) nfsrv_putattrbit(nd, &attrbits);
error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
if (error != 0) {
free(nd, M_TEMP);
return (error);
}
NFSD_DEBUG(4, "nfsrv_writedsdorpc: aft writerpc=%d\n", nd->nd_repstat);
/* Get rid of weak cache consistency data for now. */
if ((nd->nd_flag & (ND_NOMOREDATA | ND_NFSV4 | ND_V4WCCATTR)) ==
(ND_NFSV4 | ND_V4WCCATTR)) {
error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0, NULL, NULL,
NULL, NULL, NULL, 0, NULL, NULL, NULL, NULL, NULL);
NFSD_DEBUG(4, "nfsrv_writedsdorpc: wcc attr=%d\n", error);
if (error != 0)
goto nfsmout;
/*
* Get rid of Op# and status for next op.
*/
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
if (*++tl != 0)
nd->nd_flag |= ND_NOMOREDATA;
}
if (nd->nd_repstat == 0) {
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
retlen = fxdr_unsigned(int, *tl++);
commit = fxdr_unsigned(int, *tl);
if (commit != NFSWRITE_FILESYNC)
error = NFSERR_IO;
NFSD_DEBUG(4, "nfsrv_writedsdorpc:retlen=%d commit=%d err=%d\n",
retlen, commit, error);
} else
error = nd->nd_repstat;
/* We have no use for the Write Verifier since we use FileSync. */
/*
* Get the Change, Size, Access Time and Modify Time attributes and set
* on the Metadata file, so its attributes will be what the file's
* would be if it had been written.
*/
if (error == 0) {
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0, NULL, NULL,
NULL, NULL, NULL, 0, NULL, NULL, NULL, NULL, NULL);
}
NFSD_DEBUG(4, "nfsrv_writedsdorpc: aft loadattr=%d\n", error);
nfsmout:
m_freem(nd->nd_mrep);
free(nd, M_TEMP);
NFSD_DEBUG(4, "nfsrv_writedsdorpc error=%d\n", error);
return (error);
}
/*
* Start up the thread that will execute nfsrv_writedsdorpc().
*/
static void
start_writedsdorpc(void *arg, int pending)
{
struct nfsrvwritedsdorpc *drpc;
drpc = (struct nfsrvwritedsdorpc *)arg;
drpc->err = nfsrv_writedsdorpc(drpc->nmp, &drpc->fh, drpc->off,
drpc->len, NULL, drpc->m, drpc->cred, drpc->p);
drpc->done = 1;
NFSD_DEBUG(4, "start_writedsdorpc: err=%d\n", drpc->err);
}
static int
nfsrv_writedsrpc(fhandle_t *fhp, off_t off, int len, struct ucred *cred,
NFSPROC_T *p, struct vnode *vp, struct nfsmount **nmpp, int mirrorcnt,
struct mbuf **mpp, char *cp, int *failposp)
{
Add some more initializations to quiet riscv build. The one case in nfs_copy_file_range() was a legitimate case, although it would probably never occur in practice.
2019-12-13 01:34:25 +00:00
struct nfsrvwritedsdorpc *drpc, *tdrpc = 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
struct nfsvattr na;
struct mbuf *m;
int error, i, offs, ret, timo;
NFSD_DEBUG(4, "in nfsrv_writedsrpc\n");
KASSERT(*mpp != NULL, ("nfsrv_writedsrpc: NULL mbuf chain"));
drpc = NULL;
if (mirrorcnt > 1)
tdrpc = drpc = malloc(sizeof(*drpc) * (mirrorcnt - 1), M_TEMP,
M_WAITOK);
/* Calculate offset in mbuf chain that data starts. */
offs = cp - mtod(*mpp, char *);
NFSD_DEBUG(4, "nfsrv_writedsrpc: mcopy offs=%d len=%d\n", offs, len);
/*
* Do the write RPC for every DS, using a separate kernel process
* for every DS except the last one.
*/
error = 0;
for (i = 0; i < mirrorcnt - 1; i++, tdrpc++) {
tdrpc->done = 0;
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(fhp, &tdrpc->fh, sizeof(*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
tdrpc->off = off;
tdrpc->len = len;
tdrpc->nmp = *nmpp;
tdrpc->cred = cred;
tdrpc->p = p;
tdrpc->inprog = 0;
tdrpc->err = 0;
tdrpc->m = m_copym(*mpp, offs, NFSM_RNDUP(len), M_WAITOK);
ret = EIO;
if (nfs_pnfsiothreads != 0) {
ret = nfs_pnfsio(start_writedsdorpc, tdrpc);
NFSD_DEBUG(4, "nfsrv_writedsrpc: nfs_pnfsio=%d\n",
ret);
}
if (ret != 0) {
ret = nfsrv_writedsdorpc(*nmpp, fhp, off, len, NULL,
tdrpc->m, cred, p);
if (nfsds_failerr(ret) && *failposp == -1)
*failposp = i;
else if (error == 0 && ret != 0)
error = ret;
}
nmpp++;
fhp++;
}
m = m_copym(*mpp, offs, NFSM_RNDUP(len), M_WAITOK);
ret = nfsrv_writedsdorpc(*nmpp, fhp, off, len, &na, m, cred, p);
if (nfsds_failerr(ret) && *failposp == -1 && mirrorcnt > 1)
*failposp = mirrorcnt - 1;
else if (error == 0 && ret != 0)
error = ret;
if (error == 0)
error = nfsrv_setextattr(vp, &na, p);
NFSD_DEBUG(4, "nfsrv_writedsrpc: aft setextat=%d\n", error);
tdrpc = drpc;
timo = hz / 50; /* Wait for 20msec. */
if (timo < 1)
timo = 1;
for (i = 0; i < mirrorcnt - 1; i++, tdrpc++) {
/* Wait for RPCs on separate threads to complete. */
while (tdrpc->inprog != 0 && tdrpc->done == 0)
tsleep(&tdrpc->tsk, PVFS, "srvwrds", timo);
if (nfsds_failerr(tdrpc->err) && *failposp == -1)
*failposp = i;
else if (error == 0 && tdrpc->err != 0)
error = tdrpc->err;
}
free(drpc, M_TEMP);
return (error);
}
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
/*
* Do a allocate RPC on a DS data file, using this structure for the arguments,
* so that this function can be executed by a separate kernel process.
*/
struct nfsrvallocatedsdorpc {
int done;
int inprog;
struct task tsk;
fhandle_t fh;
off_t off;
off_t len;
struct nfsmount *nmp;
struct ucred *cred;
NFSPROC_T *p;
int err;
};
static int
nfsrv_allocatedsdorpc(struct nfsmount *nmp, fhandle_t *fhp, off_t off,
off_t len, struct nfsvattr *nap, struct ucred *cred, NFSPROC_T *p)
{
uint32_t *tl;
struct nfsrv_descript *nd;
nfsattrbit_t attrbits;
nfsv4stateid_t st;
int error;
nd = malloc(sizeof(*nd), M_TEMP, M_WAITOK | M_ZERO);
nfscl_reqstart(nd, NFSPROC_ALLOCATE, nmp, (u_int8_t *)fhp,
Delete the unused "use_ext" argument to nfscl_reqstart(). This is a partial revert of r363210, since the "use_ext" argument added by that commit is not actually useful. This patch should not result in any semantics change.
2020-08-18 01:41:12 +00:00
sizeof(fhandle_t), NULL, NULL, 0, 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
/*
* Use a stateid where other is an alternating 01010 pattern and
* seqid is 0xffffffff. This value is not defined as special by
* the RFC and is used by the FreeBSD NFS server to indicate an
* MDS->DS proxy operation.
*/
st.other[0] = 0x55555555;
st.other[1] = 0x55555555;
st.other[2] = 0x55555555;
st.seqid = 0xffffffff;
nfsm_stateidtom(nd, &st, NFSSTATEID_PUTSTATEID);
NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_UNSIGNED);
txdr_hyper(off, tl); tl += 2;
txdr_hyper(len, tl); tl += 2;
NFSD_DEBUG(4, "nfsrv_allocatedsdorpc: len=%jd\n", (intmax_t)len);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
NFSGETATTR_ATTRBIT(&attrbits);
nfsrv_putattrbit(nd, &attrbits);
error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
if (error != 0) {
free(nd, M_TEMP);
return (error);
}
NFSD_DEBUG(4, "nfsrv_allocatedsdorpc: aft allocaterpc=%d\n",
nd->nd_repstat);
if (nd->nd_repstat == 0) {
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0, NULL, NULL,
NULL, NULL, NULL, 0, NULL, NULL, NULL, NULL, NULL);
} else
error = nd->nd_repstat;
NFSD_DEBUG(4, "nfsrv_allocatedsdorpc: aft loadattr=%d\n", error);
nfsmout:
m_freem(nd->nd_mrep);
free(nd, M_TEMP);
NFSD_DEBUG(4, "nfsrv_allocatedsdorpc error=%d\n", error);
return (error);
}
/*
* Start up the thread that will execute nfsrv_allocatedsdorpc().
*/
static void
start_allocatedsdorpc(void *arg, int pending)
{
struct nfsrvallocatedsdorpc *drpc;
drpc = (struct nfsrvallocatedsdorpc *)arg;
drpc->err = nfsrv_allocatedsdorpc(drpc->nmp, &drpc->fh, drpc->off,
drpc->len, NULL, drpc->cred, drpc->p);
drpc->done = 1;
NFSD_DEBUG(4, "start_allocatedsdorpc: err=%d\n", drpc->err);
}
static int
nfsrv_allocatedsrpc(fhandle_t *fhp, off_t off, off_t len, struct ucred *cred,
NFSPROC_T *p, struct vnode *vp, struct nfsmount **nmpp, int mirrorcnt,
int *failposp)
{
Add some more initializations to quiet riscv build. The one case in nfs_copy_file_range() was a legitimate case, although it would probably never occur in practice.
2019-12-13 01:34:25 +00:00
struct nfsrvallocatedsdorpc *drpc, *tdrpc = NULL;
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 nfsvattr na;
int error, i, ret, timo;
NFSD_DEBUG(4, "in nfsrv_allocatedsrpc\n");
drpc = NULL;
if (mirrorcnt > 1)
tdrpc = drpc = malloc(sizeof(*drpc) * (mirrorcnt - 1), M_TEMP,
M_WAITOK);
/*
* Do the allocate RPC for every DS, using a separate kernel process
* for every DS except the last one.
*/
error = 0;
for (i = 0; i < mirrorcnt - 1; i++, tdrpc++) {
tdrpc->done = 0;
NFSBCOPY(fhp, &tdrpc->fh, sizeof(*fhp));
tdrpc->off = off;
tdrpc->len = len;
tdrpc->nmp = *nmpp;
tdrpc->cred = cred;
tdrpc->p = p;
tdrpc->inprog = 0;
tdrpc->err = 0;
ret = EIO;
if (nfs_pnfsiothreads != 0) {
ret = nfs_pnfsio(start_allocatedsdorpc, tdrpc);
NFSD_DEBUG(4, "nfsrv_allocatedsrpc: nfs_pnfsio=%d\n",
ret);
}
if (ret != 0) {
ret = nfsrv_allocatedsdorpc(*nmpp, fhp, off, len, NULL,
cred, p);
if (nfsds_failerr(ret) && *failposp == -1)
*failposp = i;
else if (error == 0 && ret != 0)
error = ret;
}
nmpp++;
fhp++;
}
ret = nfsrv_allocatedsdorpc(*nmpp, fhp, off, len, &na, cred, p);
if (nfsds_failerr(ret) && *failposp == -1 && mirrorcnt > 1)
*failposp = mirrorcnt - 1;
else if (error == 0 && ret != 0)
error = ret;
if (error == 0)
error = nfsrv_setextattr(vp, &na, p);
NFSD_DEBUG(4, "nfsrv_allocatedsrpc: aft setextat=%d\n", error);
tdrpc = drpc;
timo = hz / 50; /* Wait for 20msec. */
if (timo < 1)
timo = 1;
for (i = 0; i < mirrorcnt - 1; i++, tdrpc++) {
/* Wait for RPCs on separate threads to complete. */
while (tdrpc->inprog != 0 && tdrpc->done == 0)
tsleep(&tdrpc->tsk, PVFS, "srvalds", timo);
if (nfsds_failerr(tdrpc->err) && *failposp == -1)
*failposp = i;
else if (error == 0 && tdrpc->err != 0)
error = tdrpc->err;
}
free(drpc, M_TEMP);
return (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
static int
nfsrv_setattrdsdorpc(fhandle_t *fhp, struct ucred *cred, NFSPROC_T *p,
struct vnode *vp, struct nfsmount *nmp, struct nfsvattr *nap,
struct nfsvattr *dsnap)
{
uint32_t *tl;
struct nfsrv_descript *nd;
nfsv4stateid_t st;
nfsattrbit_t attrbits;
int error;
NFSD_DEBUG(4, "in nfsrv_setattrdsdorpc\n");
nd = malloc(sizeof(*nd), M_TEMP, M_WAITOK | M_ZERO);
/*
* Use a stateid where other is an alternating 01010 pattern and
* seqid is 0xffffffff. This value is not defined as special by
* the RFC and is used by the FreeBSD NFS server to indicate an
* MDS->DS proxy operation.
*/
st.other[0] = 0x55555555;
st.other[1] = 0x55555555;
st.other[2] = 0x55555555;
st.seqid = 0xffffffff;
nfscl_reqstart(nd, NFSPROC_SETATTR, nmp, (u_int8_t *)fhp, sizeof(*fhp),
Delete the unused "use_ext" argument to nfscl_reqstart(). This is a partial revert of r363210, since the "use_ext" argument added by that commit is not actually useful. This patch should not result in any semantics change.
2020-08-18 01:41:12 +00:00
NULL, NULL, 0, 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_stateidtom(nd, &st, NFSSTATEID_PUTSTATEID);
nfscl_fillsattr(nd, &nap->na_vattr, vp, NFSSATTR_FULL, 0);
Fix the pNFS server's reporting of SpaceUsed (va_bytes). The pNFS server currently reports SpaceUsed (va_bytes) for the metadata file. This in not correct, since the metadata file is always empty and, as such, va_bytes is just the allocation for the empty file. This patch adds va_bytes to the list of attributes acquired from the DS for a file, so that it includes the allocated data size and is updated when the file is written. For files created on a pNFS server before this patch is applied, the va_bytes value is estimated by rounding va_size up to a multiple of BLKDEV_IOSIZE. Once the file is written after this patch has been applied to the metadata server, the va_bytes returned for the file will be correct. This patch only affects a pNFS metadata server. Found during testing of the NFSv4.2 pNFS server for the Allocate operation. (Not yet in head/current.) MFC after: 2 weeks
2019-11-22 00:22:55 +00:00
/* Do a Getattr for the attributes that change due to writing. */
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
NFSZERO_ATTRBIT(&attrbits);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_SIZE);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEACCESS);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
Fix the pNFS server's reporting of SpaceUsed (va_bytes). The pNFS server currently reports SpaceUsed (va_bytes) for the metadata file. This in not correct, since the metadata file is always empty and, as such, va_bytes is just the allocation for the empty file. This patch adds va_bytes to the list of attributes acquired from the DS for a file, so that it includes the allocated data size and is updated when the file is written. For files created on a pNFS server before this patch is applied, the va_bytes value is estimated by rounding va_size up to a multiple of BLKDEV_IOSIZE. Once the file is written after this patch has been applied to the metadata server, the va_bytes returned for the file will be correct. This patch only affects a pNFS metadata server. Found during testing of the NFSv4.2 pNFS server for the Allocate operation. (Not yet in head/current.) MFC after: 2 weeks
2019-11-22 00:22:55 +00:00
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_SPACEUSED);
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 *, NFSX_UNSIGNED);
*tl = txdr_unsigned(NFSV4OP_GETATTR);
(void) nfsrv_putattrbit(nd, &attrbits);
error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
if (error != 0) {
free(nd, M_TEMP);
return (error);
}
NFSD_DEBUG(4, "nfsrv_setattrdsdorpc: aft setattrrpc=%d\n",
nd->nd_repstat);
/* Get rid of weak cache consistency data for now. */
if ((nd->nd_flag & (ND_NOMOREDATA | ND_NFSV4 | ND_V4WCCATTR)) ==
(ND_NFSV4 | ND_V4WCCATTR)) {
error = nfsv4_loadattr(nd, NULL, dsnap, NULL, NULL, 0, NULL,
NULL, NULL, NULL, NULL, 0, NULL, NULL, NULL, NULL, NULL);
NFSD_DEBUG(4, "nfsrv_setattrdsdorpc: wcc attr=%d\n", error);
if (error != 0)
goto nfsmout;
/*
* Get rid of Op# and status for next op.
*/
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
if (*++tl != 0)
nd->nd_flag |= ND_NOMOREDATA;
}
error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
if (error != 0)
goto nfsmout;
if (nd->nd_repstat != 0)
error = nd->nd_repstat;
/*
* Get the Change, Size, Access Time and Modify Time attributes and set
* on the Metadata file, so its attributes will be what the file's
* would be if it had been written.
*/
if (error == 0) {
NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
error = nfsv4_loadattr(nd, NULL, dsnap, NULL, NULL, 0, NULL,
NULL, NULL, NULL, NULL, 0, NULL, NULL, NULL, NULL, NULL);
}
NFSD_DEBUG(4, "nfsrv_setattrdsdorpc: aft setattr loadattr=%d\n", error);
nfsmout:
m_freem(nd->nd_mrep);
free(nd, M_TEMP);
NFSD_DEBUG(4, "nfsrv_setattrdsdorpc error=%d\n", error);
return (error);
}
struct nfsrvsetattrdsdorpc {
int done;
int inprog;
struct task tsk;
fhandle_t fh;
struct nfsmount *nmp;
struct vnode *vp;
struct ucred *cred;
NFSPROC_T *p;
struct nfsvattr na;
struct nfsvattr dsna;
int err;
};
/*
* Start up the thread that will execute nfsrv_setattrdsdorpc().
*/
static void
start_setattrdsdorpc(void *arg, int pending)
{
struct nfsrvsetattrdsdorpc *drpc;
drpc = (struct nfsrvsetattrdsdorpc *)arg;
drpc->err = nfsrv_setattrdsdorpc(&drpc->fh, drpc->cred, drpc->p,
drpc->vp, drpc->nmp, &drpc->na, &drpc->dsna);
drpc->done = 1;
}
static int
nfsrv_setattrdsrpc(fhandle_t *fhp, struct ucred *cred, NFSPROC_T *p,
struct vnode *vp, struct nfsmount **nmpp, int mirrorcnt,
struct nfsvattr *nap, int *failposp)
{
Add some more initializations to quiet riscv build. The one case in nfs_copy_file_range() was a legitimate case, although it would probably never occur in practice.
2019-12-13 01:34:25 +00:00
struct nfsrvsetattrdsdorpc *drpc, *tdrpc = 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
struct nfsvattr na;
int error, i, ret, timo;
NFSD_DEBUG(4, "in nfsrv_setattrdsrpc\n");
drpc = NULL;
if (mirrorcnt > 1)
tdrpc = drpc = malloc(sizeof(*drpc) * (mirrorcnt - 1), M_TEMP,
M_WAITOK);
/*
* Do the setattr RPC for every DS, using a separate kernel process
* for every DS except the last one.
*/
error = 0;
for (i = 0; i < mirrorcnt - 1; i++, tdrpc++) {
tdrpc->done = 0;
tdrpc->inprog = 0;
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(fhp, &tdrpc->fh, sizeof(*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
tdrpc->nmp = *nmpp;
tdrpc->vp = vp;
tdrpc->cred = cred;
tdrpc->p = p;
tdrpc->na = *nap;
tdrpc->err = 0;
ret = EIO;
if (nfs_pnfsiothreads != 0) {
ret = nfs_pnfsio(start_setattrdsdorpc, tdrpc);
NFSD_DEBUG(4, "nfsrv_setattrdsrpc: nfs_pnfsio=%d\n",
ret);
}
if (ret != 0) {
ret = nfsrv_setattrdsdorpc(fhp, cred, p, vp, *nmpp, nap,
&na);
if (nfsds_failerr(ret) && *failposp == -1)
*failposp = i;
else if (error == 0 && ret != 0)
error = ret;
}
nmpp++;
fhp++;
}
ret = nfsrv_setattrdsdorpc(fhp, cred, p, vp, *nmpp, nap, &na);
if (nfsds_failerr(ret) && *failposp == -1 && mirrorcnt > 1)
*failposp = mirrorcnt - 1;
else if (error == 0 && ret != 0)
error = ret;
if (error == 0)
error = nfsrv_setextattr(vp, &na, p);
NFSD_DEBUG(4, "nfsrv_setattrdsrpc: aft setextat=%d\n", error);
tdrpc = drpc;
timo = hz / 50; /* Wait for 20msec. */
if (timo < 1)
timo = 1;
for (i = 0; i < mirrorcnt - 1; i++, tdrpc++) {
/* Wait for RPCs on separate threads to complete. */
while (tdrpc->inprog != 0 && tdrpc->done == 0)
tsleep(&tdrpc->tsk, PVFS, "srvsads", timo);
if (nfsds_failerr(tdrpc->err) && *failposp == -1)
*failposp = i;
else if (error == 0 && tdrpc->err != 0)
error = tdrpc->err;
}
free(drpc, M_TEMP);
return (error);
}
/*
* Do a Setattr of an NFSv4 ACL on the DS file.
*/
static int
nfsrv_setacldsdorpc(fhandle_t *fhp, struct ucred *cred, NFSPROC_T *p,
struct vnode *vp, struct nfsmount *nmp, struct acl *aclp)
{
struct nfsrv_descript *nd;
nfsv4stateid_t st;
nfsattrbit_t attrbits;
int error;
NFSD_DEBUG(4, "in nfsrv_setacldsdorpc\n");
nd = malloc(sizeof(*nd), M_TEMP, M_WAITOK | M_ZERO);
/*
* Use a stateid where other is an alternating 01010 pattern and
* seqid is 0xffffffff. This value is not defined as special by
* the RFC and is used by the FreeBSD NFS server to indicate an
* MDS->DS proxy operation.
*/
st.other[0] = 0x55555555;
st.other[1] = 0x55555555;
st.other[2] = 0x55555555;
st.seqid = 0xffffffff;
nfscl_reqstart(nd, NFSPROC_SETACL, nmp, (u_int8_t *)fhp, sizeof(*fhp),
Delete the unused "use_ext" argument to nfscl_reqstart(). This is a partial revert of r363210, since the "use_ext" argument added by that commit is not actually useful. This patch should not result in any semantics change.
2020-08-18 01:41:12 +00:00
NULL, NULL, 0, 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_stateidtom(nd, &st, NFSSTATEID_PUTSTATEID);
NFSZERO_ATTRBIT(&attrbits);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
/*
* The "vp" argument to nfsv4_fillattr() is only used for vnode_type(),
* so passing in the metadata "vp" will be ok, since it is of
* the same type (VREG).
*/
nfsv4_fillattr(nd, NULL, vp, aclp, NULL, NULL, 0, &attrbits, NULL,
NULL, 0, 0, 0, 0, 0, NULL);
error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
if (error != 0) {
free(nd, M_TEMP);
return (error);
}
NFSD_DEBUG(4, "nfsrv_setacldsdorpc: aft setaclrpc=%d\n",
nd->nd_repstat);
error = nd->nd_repstat;
m_freem(nd->nd_mrep);
free(nd, M_TEMP);
return (error);
}
struct nfsrvsetacldsdorpc {
int done;
int inprog;
struct task tsk;
fhandle_t fh;
struct nfsmount *nmp;
struct vnode *vp;
struct ucred *cred;
NFSPROC_T *p;
struct acl *aclp;
int err;
};
/*
* Start up the thread that will execute nfsrv_setacldsdorpc().
*/
static void
start_setacldsdorpc(void *arg, int pending)
{
struct nfsrvsetacldsdorpc *drpc;
drpc = (struct nfsrvsetacldsdorpc *)arg;
drpc->err = nfsrv_setacldsdorpc(&drpc->fh, drpc->cred, drpc->p,
drpc->vp, drpc->nmp, drpc->aclp);
drpc->done = 1;
}
static int
nfsrv_setacldsrpc(fhandle_t *fhp, struct ucred *cred, NFSPROC_T *p,
struct vnode *vp, struct nfsmount **nmpp, int mirrorcnt, struct acl *aclp,
int *failposp)
{
Add some more initializations to quiet riscv build. The one case in nfs_copy_file_range() was a legitimate case, although it would probably never occur in practice.
2019-12-13 01:34:25 +00:00
struct nfsrvsetacldsdorpc *drpc, *tdrpc = 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
int error, i, ret, timo;
NFSD_DEBUG(4, "in nfsrv_setacldsrpc\n");
drpc = NULL;
if (mirrorcnt > 1)
tdrpc = drpc = malloc(sizeof(*drpc) * (mirrorcnt - 1), M_TEMP,
M_WAITOK);
/*
* Do the setattr RPC for every DS, using a separate kernel process
* for every DS except the last one.
*/
error = 0;
for (i = 0; i < mirrorcnt - 1; i++, tdrpc++) {
tdrpc->done = 0;
tdrpc->inprog = 0;
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(fhp, &tdrpc->fh, sizeof(*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
tdrpc->nmp = *nmpp;
tdrpc->vp = vp;
tdrpc->cred = cred;
tdrpc->p = p;
tdrpc->aclp = aclp;
tdrpc->err = 0;
ret = EIO;
if (nfs_pnfsiothreads != 0) {
ret = nfs_pnfsio(start_setacldsdorpc, tdrpc);
NFSD_DEBUG(4, "nfsrv_setacldsrpc: nfs_pnfsio=%d\n",
ret);
}
if (ret != 0) {
ret = nfsrv_setacldsdorpc(fhp, cred, p, vp, *nmpp,
aclp);
if (nfsds_failerr(ret) && *failposp == -1)
*failposp = i;
else if (error == 0 && ret != 0)
error = ret;
}
nmpp++;
fhp++;
}
ret = nfsrv_setacldsdorpc(fhp, cred, p, vp, *nmpp, aclp);
if (nfsds_failerr(ret) && *failposp == -1 && mirrorcnt > 1)
*failposp = mirrorcnt - 1;
else if (error == 0 && ret != 0)
error = ret;
NFSD_DEBUG(4, "nfsrv_setacldsrpc: aft setextat=%d\n", error);
tdrpc = drpc;
timo = hz / 50; /* Wait for 20msec. */
if (timo < 1)
timo = 1;
for (i = 0; i < mirrorcnt - 1; i++, tdrpc++) {
/* Wait for RPCs on separate threads to complete. */
while (tdrpc->inprog != 0 && tdrpc->done == 0)
tsleep(&tdrpc->tsk, PVFS, "srvacds", timo);
if (nfsds_failerr(tdrpc->err) && *failposp == -1)
*failposp = i;
else if (error == 0 && tdrpc->err != 0)
error = tdrpc->err;
}
free(drpc, M_TEMP);
return (error);
}
/*
Fix the pNFS server's reporting of SpaceUsed (va_bytes). The pNFS server currently reports SpaceUsed (va_bytes) for the metadata file. This in not correct, since the metadata file is always empty and, as such, va_bytes is just the allocation for the empty file. This patch adds va_bytes to the list of attributes acquired from the DS for a file, so that it includes the allocated data size and is updated when the file is written. For files created on a pNFS server before this patch is applied, the va_bytes value is estimated by rounding va_size up to a multiple of BLKDEV_IOSIZE. Once the file is written after this patch has been applied to the metadata server, the va_bytes returned for the file will be correct. This patch only affects a pNFS metadata server. Found during testing of the NFSv4.2 pNFS server for the Allocate operation. (Not yet in head/current.) MFC after: 2 weeks
2019-11-22 00:22:55 +00:00
* Getattr call to the DS for the attributes that change due to writing.
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_getattrdsrpc(fhandle_t *fhp, struct ucred *cred, NFSPROC_T *p,
struct vnode *vp, struct nfsmount *nmp, struct nfsvattr *nap)
{
struct nfsrv_descript *nd;
int error;
nfsattrbit_t attrbits;
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
NFSD_DEBUG(4, "in nfsrv_getattrdsrpc\n");
nd = malloc(sizeof(*nd), M_TEMP, M_WAITOK | M_ZERO);
nfscl_reqstart(nd, NFSPROC_GETATTR, nmp, (u_int8_t *)fhp,
Delete the unused "use_ext" argument to nfscl_reqstart(). This is a partial revert of r363210, since the "use_ext" argument added by that commit is not actually useful. This patch should not result in any semantics change.
2020-08-18 01:41:12 +00:00
sizeof(fhandle_t), NULL, NULL, 0, 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
NFSZERO_ATTRBIT(&attrbits);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_SIZE);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEACCESS);
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
Fix the pNFS server's reporting of SpaceUsed (va_bytes). The pNFS server currently reports SpaceUsed (va_bytes) for the metadata file. This in not correct, since the metadata file is always empty and, as such, va_bytes is just the allocation for the empty file. This patch adds va_bytes to the list of attributes acquired from the DS for a file, so that it includes the allocated data size and is updated when the file is written. For files created on a pNFS server before this patch is applied, the va_bytes value is estimated by rounding va_size up to a multiple of BLKDEV_IOSIZE. Once the file is written after this patch has been applied to the metadata server, the va_bytes returned for the file will be correct. This patch only affects a pNFS metadata server. Found during testing of the NFSv4.2 pNFS server for the Allocate operation. (Not yet in head/current.) MFC after: 2 weeks
2019-11-22 00:22:55 +00:00
NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_SPACEUSED);
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
(void) nfsrv_putattrbit(nd, &attrbits);
error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
if (error != 0) {
free(nd, M_TEMP);
return (error);
}
NFSD_DEBUG(4, "nfsrv_getattrdsrpc: aft getattrrpc=%d\n",
nd->nd_repstat);
if (nd->nd_repstat == 0) {
error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
NULL, NULL, NULL, NULL, NULL, 0, NULL, NULL, NULL,
NULL, NULL);
/*
* We can only save the updated values in the extended
* attribute if the vp is exclusively locked.
* This should happen when any of the following operations
* occur on the vnode:
* Close, Delegreturn, LayoutCommit, LayoutReturn
* As such, the updated extended attribute should get saved
* before nfsrv_checkdsattr() returns 0 and allows the cached
* attributes to be returned without calling this function.
*/
if (error == 0 && VOP_ISLOCKED(vp) == LK_EXCLUSIVE) {
error = nfsrv_setextattr(vp, nap, p);
NFSD_DEBUG(4, "nfsrv_getattrdsrpc: aft setextat=%d\n",
error);
}
} else
error = nd->nd_repstat;
m_freem(nd->nd_mrep);
free(nd, M_TEMP);
NFSD_DEBUG(4, "nfsrv_getattrdsrpc error=%d\n", error);
return (error);
}
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
/*
* Seek call to a DS.
*/
static int
nfsrv_seekdsrpc(fhandle_t *fhp, off_t *offp, int content, bool *eofp,
struct ucred *cred, NFSPROC_T *p, struct nfsmount *nmp)
{
uint32_t *tl;
struct nfsrv_descript *nd;
nfsv4stateid_t st;
int error;
fs: clean up empty lines in .c and .h files
2020-09-01 21:18:40 +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
NFSD_DEBUG(4, "in nfsrv_seekdsrpc\n");
/*
* Use a stateid where other is an alternating 01010 pattern and
* seqid is 0xffffffff. This value is not defined as special by
* the RFC and is used by the FreeBSD NFS server to indicate an
* MDS->DS proxy operation.
*/
st.other[0] = 0x55555555;
st.other[1] = 0x55555555;
st.other[2] = 0x55555555;
st.seqid = 0xffffffff;
nd = malloc(sizeof(*nd), M_TEMP, M_WAITOK | M_ZERO);
nfscl_reqstart(nd, NFSPROC_SEEKDS, nmp, (u_int8_t *)fhp,
Delete the unused "use_ext" argument to nfscl_reqstart(). This is a partial revert of r363210, since the "use_ext" argument added by that commit is not actually useful. This patch should not result in any semantics change.
2020-08-18 01:41:12 +00:00
sizeof(fhandle_t), NULL, NULL, 0, 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
nfsm_stateidtom(nd, &st, NFSSTATEID_PUTSTATEID);
NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
txdr_hyper(*offp, tl); tl += 2;
*tl = txdr_unsigned(content);
error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
if (error != 0) {
free(nd, M_TEMP);
return (error);
}
NFSD_DEBUG(4, "nfsrv_seekdsrpc: aft seekrpc=%d\n", nd->nd_repstat);
if (nd->nd_repstat == 0) {
NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED + NFSX_HYPER);
if (*tl++ == newnfs_true)
*eofp = true;
else
*eofp = false;
*offp = fxdr_hyper(tl);
} else
error = nd->nd_repstat;
nfsmout:
m_freem(nd->nd_mrep);
free(nd, M_TEMP);
NFSD_DEBUG(4, "nfsrv_seekdsrpc error=%d\n", error);
return (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
/*
* Get the device id and file handle for a DS file.
*/
int
nfsrv_dsgetdevandfh(struct vnode *vp, NFSPROC_T *p, int *mirrorcntp,
fhandle_t *fhp, char *devid)
{
int buflen, error;
char *buf;
buflen = 1024;
buf = malloc(buflen, M_TEMP, M_WAITOK);
error = nfsrv_dsgetsockmnt(vp, 0, buf, &buflen, mirrorcntp, p, NULL,
fhp, devid, NULL, NULL, NULL, NULL, NULL, NULL);
free(buf, M_TEMP);
return (error);
}
/*
* Do a Lookup against the DS for the filename.
*/
static int
nfsrv_pnfslookupds(struct vnode *vp, struct vnode *dvp, struct pnfsdsfile *pf,
struct vnode **nvpp, NFSPROC_T *p)
{
struct nameidata named;
struct ucred *tcred;
char *bufp;
u_long *hashp;
struct vnode *nvp;
int error;
tcred = newnfs_getcred();
named.ni_cnd.cn_nameiop = LOOKUP;
named.ni_cnd.cn_lkflags = LK_SHARED | LK_RETRY;
named.ni_cnd.cn_cred = tcred;
named.ni_cnd.cn_thread = p;
named.ni_cnd.cn_flags = ISLASTCN | LOCKPARENT | LOCKLEAF | SAVENAME;
nfsvno_setpathbuf(&named, &bufp, &hashp);
named.ni_cnd.cn_nameptr = bufp;
named.ni_cnd.cn_namelen = strlen(pf->dsf_filename);
strlcpy(bufp, pf->dsf_filename, NAME_MAX);
NFSD_DEBUG(4, "nfsrv_pnfslookupds: filename=%s\n", bufp);
error = VOP_LOOKUP(dvp, &nvp, &named.ni_cnd);
NFSD_DEBUG(4, "nfsrv_pnfslookupds: aft LOOKUP=%d\n", error);
NFSFREECRED(tcred);
nfsvno_relpathbuf(&named);
if (error == 0)
*nvpp = nvp;
NFSD_DEBUG(4, "eo nfsrv_pnfslookupds=%d\n", error);
return (error);
}
/*
* Set the file handle to the correct one.
*/
static void
Fix LORs between vn_start_write() and vn_lock() in the pNFS server. When coding the pNFS server, I added several vn_start_write() calls done while the vnode was locked, not realizing I had introduced LORs and possible deadlock when an exported file system on the MDS is suspended. This patch fixes this by removing the added vn_start_write() calls and modifying the code so that the extant vn_start_write() call before the NFS RPC/operation is done when needed by the pNFS server. Flags are changed so that LayoutCommit and LayoutReturn now get a vn_start_write() done for them. When the pNFS server is enabled, the code now also changes the flags for Getattr, so that the vn_start_write() is done for Getattr, since it may need to do a vn_set_extattr(). The nfs_writerpc flag array was made global to the NFS server and renamed nfsrv_writerpc, which is consistent naming for globals in the NFS server. Thanks go to kib@ for reporting that doing vn_start_write() while the vnode is locked results in a LOR. This patch only affects the behaviour of the pNFS server.
2018-08-17 21:12:16 +00:00
nfsrv_pnfssetfh(struct vnode *vp, struct pnfsdsfile *pf, char *devid,
char *fnamep, struct vnode *nvp, 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 nfsnode *np;
Fix the build for MAC not defined and a couple of might not be initialized. r355677 broke the build for the not MAC defined case and a couple of might not be initialized warnings were generated for riscv. Others seem to be erroneous. Hopefully there won't be too many more build errors. Pointy hat goes on me.
2019-12-13 00:45:14 +00:00
int ret = 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
np = VTONFS(nvp);
NFSBCOPY(np->n_fhp->nfh_fh, &pf->dsf_fh, NFSX_MYFH);
/*
Fix LORs between vn_start_write() and vn_lock() in the pNFS server. When coding the pNFS server, I added several vn_start_write() calls done while the vnode was locked, not realizing I had introduced LORs and possible deadlock when an exported file system on the MDS is suspended. This patch fixes this by removing the added vn_start_write() calls and modifying the code so that the extant vn_start_write() call before the NFS RPC/operation is done when needed by the pNFS server. Flags are changed so that LayoutCommit and LayoutReturn now get a vn_start_write() done for them. When the pNFS server is enabled, the code now also changes the flags for Getattr, so that the vn_start_write() is done for Getattr, since it may need to do a vn_set_extattr(). The nfs_writerpc flag array was made global to the NFS server and renamed nfsrv_writerpc, which is consistent naming for globals in the NFS server. Thanks go to kib@ for reporting that doing vn_start_write() while the vnode is locked results in a LOR. This patch only affects the behaviour of the pNFS server.
2018-08-17 21:12:16 +00:00
* We can only do a vn_set_extattr() if the vnode is exclusively
* locked and vn_start_write() has been done. If devid != NULL or
* fnamep != NULL or the vnode is shared locked, vn_start_write()
* may not have been done.
* If not done now, it will be done on a future call.
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
*/
Fix LORs between vn_start_write() and vn_lock() in the pNFS server. When coding the pNFS server, I added several vn_start_write() calls done while the vnode was locked, not realizing I had introduced LORs and possible deadlock when an exported file system on the MDS is suspended. This patch fixes this by removing the added vn_start_write() calls and modifying the code so that the extant vn_start_write() call before the NFS RPC/operation is done when needed by the pNFS server. Flags are changed so that LayoutCommit and LayoutReturn now get a vn_start_write() done for them. When the pNFS server is enabled, the code now also changes the flags for Getattr, so that the vn_start_write() is done for Getattr, since it may need to do a vn_set_extattr(). The nfs_writerpc flag array was made global to the NFS server and renamed nfsrv_writerpc, which is consistent naming for globals in the NFS server. Thanks go to kib@ for reporting that doing vn_start_write() while the vnode is locked results in a LOR. This patch only affects the behaviour of the pNFS server.
2018-08-17 21:12:16 +00:00
if (devid == NULL && fnamep == NULL && NFSVOPISLOCKED(vp) ==
LK_EXCLUSIVE)
ret = vn_extattr_set(vp, IO_NODELOCKED,
EXTATTR_NAMESPACE_SYSTEM, "pnfsd.dsfile", sizeof(*pf),
(char *)pf, 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, "eo nfsrv_pnfssetfh=%d\n", ret);
}
/*
* Cause RPCs waiting on "nmp" to fail. This is called for a DS mount point
* when the DS has failed.
*/
void
nfsrv_killrpcs(struct nfsmount *nmp)
{
/*
* Call newnfs_nmcancelreqs() to cause
* any RPCs in progress on the mount point to
* fail.
* This will cause any process waiting for an
* RPC to complete while holding a vnode lock
* on the mounted-on vnode (such as "df" or
* a non-forced "umount") to fail.
* This will unlock the mounted-on vnode so
* a forced dismount can succeed.
* The NFSMNTP_CANCELRPCS flag should be set when this function is
* called.
*/
newnfs_nmcancelreqs(nmp);
}
/*
* Sum up the statfs info for each of the DSs, so that the client will
* receive the total for all DSs.
*/
static int
Fix the pNFS server's reporting of disk space usage for the "#<path>" case. The pNFS server would report the total disk space used and free for all of the DSs, even when certain DSs are assigned to the file system via the "#<path>" suffix used in the "nfsd -p" option argument. This patch fixes this case. It only reports usage for the file system that the argument vnode resides on. This is consistent with the non-pNFS NFSv4 server. In NFSv4 it is possible to have subtrees on other file systems, but these are not included in the usage information for NFSv4. Approved by: re (gjb)
2018-10-09 01:10:50 +00:00
nfsrv_pnfsstatfs(struct statfs *sf, 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
{
struct statfs *tsf;
struct nfsdevice *ds;
struct vnode **dvpp, **tdvpp, *dvp;
uint64_t tot;
int cnt, error = 0, i;
if (nfsrv_devidcnt <= 0)
return (ENXIO);
dvpp = mallocarray(nfsrv_devidcnt, sizeof(*dvpp), M_TEMP, M_WAITOK);
tsf = malloc(sizeof(*tsf), M_TEMP, M_WAITOK);
/* Get an array of the dvps for the DSs. */
tdvpp = dvpp;
i = 0;
NFSDDSLOCK();
Fix the pNFS server's reporting of disk space usage for the "#<path>" case. The pNFS server would report the total disk space used and free for all of the DSs, even when certain DSs are assigned to the file system via the "#<path>" suffix used in the "nfsd -p" option argument. This patch fixes this case. It only reports usage for the file system that the argument vnode resides on. This is consistent with the non-pNFS NFSv4 server. In NFSv4 it is possible to have subtrees on other file systems, but these are not included in the usage information for NFSv4. Approved by: re (gjb)
2018-10-09 01:10:50 +00:00
/* First, search for matches for same file system. */
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_FOREACH(ds, &nfsrv_devidhead, nfsdev_list) {
Fix the pNFS server's reporting of disk space usage for the "#<path>" case. The pNFS server would report the total disk space used and free for all of the DSs, even when certain DSs are assigned to the file system via the "#<path>" suffix used in the "nfsd -p" option argument. This patch fixes this case. It only reports usage for the file system that the argument vnode resides on. This is consistent with the non-pNFS NFSv4 server. In NFSv4 it is possible to have subtrees on other file systems, but these are not included in the usage information for NFSv4. Approved by: re (gjb)
2018-10-09 01:10:50 +00:00
if (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, &mp->mnt_stat.f_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
if (++i > nfsrv_devidcnt)
break;
*tdvpp++ = ds->nfsdev_dvp;
}
}
Fix the pNFS server's reporting of disk space usage for the "#<path>" case. The pNFS server would report the total disk space used and free for all of the DSs, even when certain DSs are assigned to the file system via the "#<path>" suffix used in the "nfsd -p" option argument. This patch fixes this case. It only reports usage for the file system that the argument vnode resides on. This is consistent with the non-pNFS NFSv4 server. In NFSv4 it is possible to have subtrees on other file systems, but these are not included in the usage information for NFSv4. Approved by: re (gjb)
2018-10-09 01:10:50 +00:00
/*
* If no matches for same file system, total all servers not assigned
* to a file system.
*/
if (i == 0) {
TAILQ_FOREACH(ds, &nfsrv_devidhead, nfsdev_list) {
if (ds->nfsdev_nmp != NULL &&
ds->nfsdev_mdsisset == 0) {
if (++i > nfsrv_devidcnt)
break;
*tdvpp++ = ds->nfsdev_dvp;
}
}
}
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();
cnt = i;
/* Do a VFS_STATFS() for each of the DSs and sum them up. */
tdvpp = dvpp;
for (i = 0; i < cnt && error == 0; i++) {
dvp = *tdvpp++;
error = VFS_STATFS(dvp->v_mount, tsf);
if (error == 0) {
if (sf->f_bsize == 0) {
if (tsf->f_bsize > 0)
sf->f_bsize = tsf->f_bsize;
else
sf->f_bsize = 8192;
}
if (tsf->f_blocks > 0) {
if (sf->f_bsize != tsf->f_bsize) {
tot = tsf->f_blocks * tsf->f_bsize;
sf->f_blocks += (tot / sf->f_bsize);
} else
sf->f_blocks += tsf->f_blocks;
}
if (tsf->f_bfree > 0) {
if (sf->f_bsize != tsf->f_bsize) {
tot = tsf->f_bfree * tsf->f_bsize;
sf->f_bfree += (tot / sf->f_bsize);
} else
sf->f_bfree += tsf->f_bfree;
}
if (tsf->f_bavail > 0) {
if (sf->f_bsize != tsf->f_bsize) {
tot = tsf->f_bavail * tsf->f_bsize;
sf->f_bavail += (tot / sf->f_bsize);
} else
sf->f_bavail += tsf->f_bavail;
}
}
}
free(tsf, M_TEMP);
free(dvpp, M_TEMP);
return (error);
}
/*
* Set an NFSv4 acl.
*/
int
nfsrv_setacl(struct vnode *vp, NFSACL_T *aclp, struct ucred *cred, NFSPROC_T *p)
{
int error;
if (nfsrv_useacl == 0 || nfs_supportsnfsv4acls(vp) == 0) {
error = NFSERR_ATTRNOTSUPP;
goto out;
}
/*
* With NFSv4 ACLs, chmod(2) may need to add additional entries.
* Make sure it has enough room for that - splitting every entry
* into two and appending "canonical six" entries at the end.
* Cribbed out of kern/vfs_acl.c - Rick M.
*/
if (aclp->acl_cnt > (ACL_MAX_ENTRIES - 6) / 2) {
error = NFSERR_ATTRNOTSUPP;
goto out;
}
error = VOP_SETACL(vp, ACL_TYPE_NFS4, aclp, cred, p);
if (error == 0) {
error = nfsrv_dssetacl(vp, aclp, cred, p);
if (error == ENOENT)
error = 0;
}
out:
NFSEXITCODE(error);
return (error);
}
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
/*
* Seek vnode op call (actually it is a VOP_IOCTL()).
* This function is called with the vnode locked, but unlocks and vrele()s
* the vp before returning.
*/
int
nfsvno_seek(struct nfsrv_descript *nd, struct vnode *vp, u_long cmd,
off_t *offp, int content, bool *eofp, struct ucred *cred, NFSPROC_T *p)
{
struct nfsvattr at;
int error, ret;
ASSERT_VOP_LOCKED(vp, "nfsvno_seek vp");
/*
* Attempt to seek on a DS file. A return of ENOENT implies
* there is no DS file to seek on.
*/
error = nfsrv_proxyds(vp, 0, 0, cred, p, NFSPROC_SEEKDS, NULL,
NULL, NULL, NULL, NULL, offp, content, eofp);
if (error != ENOENT) {
vput(vp);
return (error);
}
/*
* Do the VOP_IOCTL() call. For the case where *offp == file_size,
* VOP_IOCTL() will return ENXIO. However, the correct reply for
* NFSv4.2 is *eofp == true and error == 0 for this case.
*/
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 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 = VOP_IOCTL(vp, cmd, offp, 0, cred, p);
*eofp = false;
if (error == ENXIO || (error == 0 && cmd == FIOSEEKHOLE)) {
/* Handle the cases where we might be at EOF. */
ret = nfsvno_getattr(vp, &at, nd, p, 0, NULL);
if (ret == 0 && *offp == at.na_size) {
*eofp = true;
error = 0;
}
if (ret != 0 && error == 0)
error = ret;
}
vrele(vp);
NFSEXITCODE(error);
return (error);
}
/*
* Allocate vnode op call.
*/
int
nfsvno_allocate(struct vnode *vp, off_t off, off_t len, struct ucred *cred,
NFSPROC_T *p)
{
int error, trycnt;
ASSERT_VOP_ELOCKED(vp, "nfsvno_allocate vp");
/*
* Attempt to allocate on a DS file. A return of ENOENT implies
* there is no DS file to allocate on.
*/
error = nfsrv_proxyds(vp, off, 0, cred, p, NFSPROC_ALLOCATE, NULL,
NULL, NULL, NULL, NULL, &len, 0, NULL);
if (error != ENOENT)
return (error);
error = 0;
/*
* Do the actual VOP_ALLOCATE(), looping a reasonable number of
* times to achieve completion.
*/
trycnt = 0;
while (error == 0 && len > 0 && trycnt++ < 20)
error = VOP_ALLOCATE(vp, &off, &len);
if (error == 0 && len > 0)
error = NFSERR_IO;
NFSEXITCODE(error);
return (error);
}
/*
* Get Extended Atribute vnode op into an mbuf list.
*/
int
nfsvno_getxattr(struct vnode *vp, char *name, uint32_t maxresp,
Add optional support for ext_pgs mbufs to the NFS server's read, readlink and getxattr operations. This patch optionally enables generation of read, readlink and getxattr replies in ext_pgs mbufs. Since neither of ND_EXTPG or ND_TLS are currently ever set, there is no change in semantics at this time. It also corrects the message in a couple of panic()s that should never occur. This is another in the series of commits that add support to the NFS client and server for building RPC messages in ext_pgs mbufs with anonymous pages. This is useful so that the entire mbuf list does not need to be copied before calling sosend() when NFS over TLS is enabled. Use of ext_pgs mbufs will not be enabled until the kernel RPC is updated to handle TLS.
2020-07-31 23:35:49 +00:00
struct ucred *cred, uint64_t flag, int maxextsiz, struct thread *p,
struct mbuf **mpp, struct mbuf **mpendp, int *lenp)
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 iovec *iv;
struct uio io, *uiop = &io;
struct mbuf *m, *m2;
int alen, error, len, tlen;
size_t siz;
/* First, find out the size of the extended attribute. */
error = VOP_GETEXTATTR(vp, EXTATTR_NAMESPACE_USER, name, NULL,
&siz, cred, p);
if (error != 0)
return (NFSERR_NOXATTR);
if (siz > maxresp - NFS_MAXXDR)
return (NFSERR_XATTR2BIG);
len = siz;
tlen = NFSM_RNDUP(len);
Fix the NFSv2 extended attribute support to handle 0 length attributes. I did not realize that zero length attributes are allowed, but they are. This patch fixes the NFSv4.2 client and server to handle zero length extended attributes correctly. Submitted by: Frank van der Linden <fllinden@amazon.com> (earlier version) Reported by: Frank van der Linden <fllinder@amazon.com>
2020-04-14 22:57:21 +00:00
if (tlen > 0) {
Add optional support for ext_pgs mbufs to the NFS server's read, readlink and getxattr operations. This patch optionally enables generation of read, readlink and getxattr replies in ext_pgs mbufs. Since neither of ND_EXTPG or ND_TLS are currently ever set, there is no change in semantics at this time. It also corrects the message in a couple of panic()s that should never occur. This is another in the series of commits that add support to the NFS client and server for building RPC messages in ext_pgs mbufs with anonymous pages. This is useful so that the entire mbuf list does not need to be copied before calling sosend() when NFS over TLS is enabled. Use of ext_pgs mbufs will not be enabled until the kernel RPC is updated to handle TLS.
2020-07-31 23:35:49 +00:00
/*
* If cnt > MCLBYTES and the reply will not be saved, use
* ext_pgs mbufs for TLS.
* For NFSv4.0, we do not know for sure if the reply will
* be saved, so do not use ext_pgs mbufs for NFSv4.0.
* Always use ext_pgs mbufs if ND_EXTPG is set.
*/
if ((flag & ND_EXTPG) != 0 || (tlen > MCLBYTES &&
(flag & (ND_TLS | ND_SAVEREPLY)) == ND_TLS &&
(flag & (ND_NFSV4 | ND_NFSV41)) != ND_NFSV4))
uiop->uio_iovcnt = nfsrv_createiovec_extpgs(tlen,
maxextsiz, &m, &m2, &iv);
else
uiop->uio_iovcnt = nfsrv_createiovec(tlen, &m, &m2,
&iv);
Fix the NFSv2 extended attribute support to handle 0 length attributes. I did not realize that zero length attributes are allowed, but they are. This patch fixes the NFSv4.2 client and server to handle zero length extended attributes correctly. Submitted by: Frank van der Linden <fllinden@amazon.com> (earlier version) Reported by: Frank van der Linden <fllinder@amazon.com>
2020-04-14 22:57:21 +00:00
uiop->uio_iov = iv;
} else {
uiop->uio_iovcnt = 0;
uiop->uio_iov = iv = NULL;
m = m2 = NULL;
}
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
uiop->uio_offset = 0;
uiop->uio_resid = tlen;
uiop->uio_rw = UIO_READ;
uiop->uio_segflg = UIO_SYSSPACE;
uiop->uio_td = p;
#ifdef MAC
error = mac_vnode_check_getextattr(cred, vp, EXTATTR_NAMESPACE_USER,
name);
if (error != 0)
goto out;
#endif
Fix the NFSv2 extended attribute support to handle 0 length attributes. I did not realize that zero length attributes are allowed, but they are. This patch fixes the NFSv4.2 client and server to handle zero length extended attributes correctly. Submitted by: Frank van der Linden <fllinden@amazon.com> (earlier version) Reported by: Frank van der Linden <fllinder@amazon.com>
2020-04-14 22:57:21 +00:00
if (tlen > 0)
error = VOP_GETEXTATTR(vp, EXTATTR_NAMESPACE_USER, name, uiop,
NULL, cred, p);
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)
goto out;
if (uiop->uio_resid > 0) {
alen = tlen;
len = tlen - uiop->uio_resid;
tlen = NFSM_RNDUP(len);
if (alen != tlen)
printf("nfsvno_getxattr: weird size read\n");
Add support for ext_pgs mbufs to nfsrv_adj(). This patch uses a slightly different algorithm for nfsrv_adj() since ext_pgs mbuf lists are not permitted to have m_len == 0 mbufs. As such, the code now frees mbufs after the adjustment in the list instead of setting their m_len field to 0. Since mbuf(s) may be trimmed off the tail of the list, the function now returns a pointer to the last mbuf in the list. This saves the caller from needing to use m_last() to find the last mbuf. It also implies that it might return a nul list, which required a check for that in nfsrvd_readlink(). This is another in the series of commits that add support to the NFS client and server for building RPC messages in ext_pgs mbufs with anonymous pages. This is useful so that the entire mbuf list does not need to be copied before calling sosend() when NFS over TLS is enabled. Use of ext_pgs mbufs will not be enabled until the kernel RPC is updated to handle TLS.
2020-07-26 02:42:09 +00:00
if (tlen == 0) {
m_freem(m);
m = m2 = NULL;
} else if (alen != tlen || tlen != len)
m2 = nfsrv_adj(m, alen - tlen, tlen - len);
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
}
*lenp = len;
*mpp = m;
*mpendp = m2;
out:
if (error != 0) {
Fix the NFSv2 extended attribute support to handle 0 length attributes. I did not realize that zero length attributes are allowed, but they are. This patch fixes the NFSv4.2 client and server to handle zero length extended attributes correctly. Submitted by: Frank van der Linden <fllinden@amazon.com> (earlier version) Reported by: Frank van der Linden <fllinder@amazon.com>
2020-04-14 22:57:21 +00:00
if (m != NULL)
m_freem(m);
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
*lenp = 0;
}
free(iv, M_TEMP);
NFSEXITCODE(error);
return (error);
}
/*
* Set Extended attribute vnode op from an mbuf list.
*/
int
nfsvno_setxattr(struct vnode *vp, char *name, int len, struct mbuf *m,
char *cp, struct ucred *cred, struct thread *p)
{
struct iovec *iv;
struct uio uio, *uiop = &uio;
int cnt, error;
Fix the build for MAC not defined and a couple of might not be initialized. r355677 broke the build for the not MAC defined case and a couple of might not be initialized warnings were generated for riscv. Others seem to be erroneous. Hopefully there won't be too many more build errors. Pointy hat goes on me.
2019-12-13 00:45:14 +00:00
error = 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
#ifdef MAC
error = mac_vnode_check_setextattr(cred, vp, EXTATTR_NAMESPACE_USER,
name);
Fix the build for MAC not defined and a couple of might not be initialized. r355677 broke the build for the not MAC defined case and a couple of might not be initialized warnings were generated for riscv. Others seem to be erroneous. Hopefully there won't be too many more build errors. Pointy hat goes on me.
2019-12-13 00:45:14 +00:00
#endif
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)
goto out;
uiop->uio_rw = UIO_WRITE;
uiop->uio_segflg = UIO_SYSSPACE;
uiop->uio_td = p;
uiop->uio_offset = 0;
uiop->uio_resid = len;
Fix the NFSv2 extended attribute support to handle 0 length attributes. I did not realize that zero length attributes are allowed, but they are. This patch fixes the NFSv4.2 client and server to handle zero length extended attributes correctly. Submitted by: Frank van der Linden <fllinden@amazon.com> (earlier version) Reported by: Frank van der Linden <fllinder@amazon.com>
2020-04-14 22:57:21 +00:00
if (len > 0) {
error = nfsrv_createiovecw(len, m, cp, &iv, &cnt);
uiop->uio_iov = iv;
uiop->uio_iovcnt = cnt;
} else {
uiop->uio_iov = iv = NULL;
uiop->uio_iovcnt = 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 (error == 0) {
error = VOP_SETEXTATTR(vp, EXTATTR_NAMESPACE_USER, name, uiop,
cred, p);
free(iv, M_TEMP);
}
out:
NFSEXITCODE(error);
return (error);
}
/*
* Remove Extended attribute vnode op.
*/
int
nfsvno_rmxattr(struct nfsrv_descript *nd, struct vnode *vp, char *name,
struct ucred *cred, struct thread *p)
{
int error;
/*
* Get rid of any delegations. I am not sure why this is required,
* but RFC-8276 says so.
*/
error = nfsrv_checkremove(vp, 0, nd, nd->nd_clientid, p);
if (error != 0)
goto out;
#ifdef MAC
error = mac_vnode_check_deleteextattr(cred, vp, EXTATTR_NAMESPACE_USER,
name);
if (error != 0)
goto out;
#endif
error = VOP_DELETEEXTATTR(vp, EXTATTR_NAMESPACE_USER, name, cred, p);
if (error == EOPNOTSUPP)
error = VOP_SETEXTATTR(vp, EXTATTR_NAMESPACE_USER, name, NULL,
cred, p);
out:
NFSEXITCODE(error);
return (error);
}
/*
* List Extended Atribute vnode op into an mbuf list.
*/
int
nfsvno_listxattr(struct vnode *vp, uint64_t cookie, struct ucred *cred,
struct thread *p, u_char **bufp, uint32_t *lenp, bool *eofp)
{
struct iovec iv;
struct uio io;
int error;
size_t siz;
*bufp = NULL;
/* First, find out the size of the extended attribute. */
error = VOP_LISTEXTATTR(vp, EXTATTR_NAMESPACE_USER, NULL, &siz, cred,
p);
if (error != 0)
return (NFSERR_NOXATTR);
if (siz <= cookie) {
*lenp = 0;
*eofp = true;
goto out;
}
if (siz > cookie + *lenp) {
siz = cookie + *lenp;
*eofp = false;
} else
*eofp = true;
/* Just choose a sanity limit of 10Mbytes for malloc(M_TEMP). */
if (siz > 10 * 1024 * 1024) {
error = NFSERR_XATTR2BIG;
goto out;
}
*bufp = malloc(siz, M_TEMP, M_WAITOK);
iv.iov_base = *bufp;
iv.iov_len = siz;
io.uio_iovcnt = 1;
io.uio_iov = &iv;
io.uio_offset = 0;
io.uio_resid = siz;
io.uio_rw = UIO_READ;
io.uio_segflg = UIO_SYSSPACE;
io.uio_td = p;
#ifdef MAC
error = mac_vnode_check_listextattr(cred, vp, EXTATTR_NAMESPACE_USER);
if (error != 0)
goto out;
#endif
error = VOP_LISTEXTATTR(vp, EXTATTR_NAMESPACE_USER, &io, NULL, cred,
p);
if (error != 0)
goto out;
if (io.uio_resid > 0)
siz -= io.uio_resid;
*lenp = siz;
out:
if (error != 0) {
free(*bufp, M_TEMP);
*bufp = NULL;
}
NFSEXITCODE(error);
return (error);
}
Add support for ext_pgs mbufs to nfsrvd_readdir() and nfsrvd_readdirplus(). This patch code that optionally (based on ND_TLS, never set yet) generates readdir replies in ext_pgs mbufs. To trim the list back, a new function that is ext_pgs aware called nfsm_trimtrailing() replaces newnfs_trimtrailing(). newnfs_trimtrailing() is no longer used, but will be removed in a future commit, since its removal does modify the internal kpi between the NFS modules. This is another in the series of commits that add support to the NFS client and server for building RPC messages in ext_pgs mbufs with anonymous pages. This is useful so that the entire mbuf list does not need to be copied before calling sosend() when NFS over TLS is enabled. Use of ext_pgs mbufs will not be enabled until the kernel RPC is updated to handle TLS.
2020-07-29 22:58:08 +00:00
/*
* Trim trailing data off the mbuf list being built.
*/
nfsd: fix NFS server for ERELOOKUP r367672 modified UFS such that certain VOPs, such as VOP_CREATE() will intermittently return ERELOOKUP. When this happens, the entire system call, or NFS operation in the case of the NFS server, must be redone. This patch adds that support to the NFS server by rolling back the state of the NFS request arguments and NFS reply arguments mbuf lists to the condition they were in before the operation and then redoing the operation. Tested by: pho Reviewed by: kib Differential Revision: https://reviews.freebsd.org/D27875
2021-01-01 21:55:51 +00:00
void
Add support for ext_pgs mbufs to nfsrvd_readdir() and nfsrvd_readdirplus(). This patch code that optionally (based on ND_TLS, never set yet) generates readdir replies in ext_pgs mbufs. To trim the list back, a new function that is ext_pgs aware called nfsm_trimtrailing() replaces newnfs_trimtrailing(). newnfs_trimtrailing() is no longer used, but will be removed in a future commit, since its removal does modify the internal kpi between the NFS modules. This is another in the series of commits that add support to the NFS client and server for building RPC messages in ext_pgs mbufs with anonymous pages. This is useful so that the entire mbuf list does not need to be copied before calling sosend() when NFS over TLS is enabled. Use of ext_pgs mbufs will not be enabled until the kernel RPC is updated to handle TLS.
2020-07-29 22:58:08 +00:00
nfsm_trimtrailing(struct nfsrv_descript *nd, struct mbuf *mb, char *bpos,
int bextpg, int bextpgsiz)
{
vm_page_t pg;
int fullpgsiz, i;
if (mb->m_next != NULL) {
m_freem(mb->m_next);
mb->m_next = NULL;
}
if ((mb->m_flags & M_EXTPG) != 0) {
nfsd: add KASSERTs to nfsm_trimtrailing() for M_EXTPG mbufs Add KASSERTS to nfsm_trimtrailing() to confirm the sanity of the arguments for the M_EXTPG case. Suggested by: kib Reviewed by: kib Differential Revision: https://reviews.freebsd.org/D28053
2021-01-10 21:50:15 +00:00
KASSERT(bextpg >= 0 && bextpg < mb->m_epg_npgs,
("nfsm_trimtrailing: bextpg out of range"));
KASSERT(bpos == (char *)(void *)
PHYS_TO_DMAP(mb->m_epg_pa[bextpg]) + PAGE_SIZE - bextpgsiz,
("nfsm_trimtrailing: bextpgsiz bad!"));
Add support for ext_pgs mbufs to nfsrvd_readdir() and nfsrvd_readdirplus(). This patch code that optionally (based on ND_TLS, never set yet) generates readdir replies in ext_pgs mbufs. To trim the list back, a new function that is ext_pgs aware called nfsm_trimtrailing() replaces newnfs_trimtrailing(). newnfs_trimtrailing() is no longer used, but will be removed in a future commit, since its removal does modify the internal kpi between the NFS modules. This is another in the series of commits that add support to the NFS client and server for building RPC messages in ext_pgs mbufs with anonymous pages. This is useful so that the entire mbuf list does not need to be copied before calling sosend() when NFS over TLS is enabled. Use of ext_pgs mbufs will not be enabled until the kernel RPC is updated to handle TLS.
2020-07-29 22:58:08 +00:00
/* First, get rid of any pages after this position. */
for (i = mb->m_epg_npgs - 1; i > bextpg; i--) {
pg = PHYS_TO_VM_PAGE(mb->m_epg_pa[i]);
vm_page_unwire_noq(pg);
vm_page_free(pg);
}
mb->m_epg_npgs = bextpg + 1;
if (bextpg == 0)
fullpgsiz = PAGE_SIZE - mb->m_epg_1st_off;
else
fullpgsiz = PAGE_SIZE;
mb->m_epg_last_len = fullpgsiz - bextpgsiz;
mb->m_len = m_epg_pagelen(mb, 0, mb->m_epg_1st_off);
for (i = 1; i < mb->m_epg_npgs; i++)
mb->m_len += m_epg_pagelen(mb, i, 0);
nd->nd_bextpgsiz = bextpgsiz;
nd->nd_bextpg = bextpg;
} else
mb->m_len = bpos - mtod(mb, char *);
nd->nd_mb = mb;
nd->nd_bpos = bpos;
}
nfsd: Fix when NFSERR_WRONGSEC may be replied to NFSv4 clients Commit d224f05fcfc1 pre-parsed the next operation number for the put file handle operations. This patch uses this next operation number, plus the type of the file handle being set by the put file handle operation, to implement the rules in RFC5661 Sec. 2.6 with respect to replying NFSERR_WRONGSEC. This patch also adds a check to see if NFSERR_WRONGSEC should be replied when about to perform Lookup, Lookupp or Open with a file name component, so that the NFSERR_WRONGSEC reply is done for these operations, as required by RFC5661 Sec. 2.6. This patch does not have any practical effect for the FreeBSD NFSv4 client and I believe that the same is true for the Linux client, since NFSERR_WRONGSEC is considered a fatal error at this time. MFC after: 2 weeks
2021-06-05 23:53:07 +00:00
/*
* Check to see if a put file handle operation should test for
* NFSERR_WRONGSEC, although NFSv3 actually returns NFSERR_AUTHERR.
* When Open is the next operation, NFSERR_WRONGSEC cannot be
* replied for the Open cases that use a component. Thia can
* be identified by the fact that the file handle's type is VDIR.
*/
bool
nfsrv_checkwrongsec(struct nfsrv_descript *nd, int nextop, enum vtype vtyp)
{
if ((nd->nd_flag & ND_NFSV4) == 0) {
if (nd->nd_procnum == NFSPROC_FSINFO)
return (false);
return (true);
}
if ((nd->nd_flag & ND_LASTOP) != 0)
return (false);
if (nextop == NFSV4OP_PUTROOTFH || nextop == NFSV4OP_PUTFH ||
nextop == NFSV4OP_PUTPUBFH || nextop == NFSV4OP_RESTOREFH ||
nextop == NFSV4OP_LOOKUP || nextop == NFSV4OP_LOOKUPP ||
nextop == NFSV4OP_SECINFO || nextop == NFSV4OP_SECINFONONAME)
return (false);
if (nextop == NFSV4OP_OPEN && vtyp == VDIR)
return (false);
return (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
extern int (*nfsd_call_nfsd)(struct thread *, struct nfssvc_args *);
/*
* Called once to initialize data structures...
*/
static int
nfsd_modevent(module_t mod, int type, void *data)
{
int error = 0, i;
static int loaded = 0;
switch (type) {
case MOD_LOAD:
if (loaded)
goto out;
newnfs_portinit();
for (i = 0; i < NFSRVCACHE_HASHSIZE; i++) {
mtx_init(&nfsrchash_table[i].mtx, "nfsrtc", NULL,
MTX_DEF);
mtx_init(&nfsrcahash_table[i].mtx, "nfsrtca", NULL,
MTX_DEF);
}
mtx_init(&nfsrc_udpmtx, "nfsuc", NULL, MTX_DEF);
mtx_init(&nfs_v4root_mutex, "nfs4rt", NULL, MTX_DEF);
mtx_init(&nfsv4root_mnt.mnt_mtx, "nfs4mnt", NULL, MTX_DEF);
mtx_init(&nfsrv_dontlistlock_mtx, "nfs4dnl", NULL, MTX_DEF);
mtx_init(&nfsrv_recalllock_mtx, "nfs4rec", NULL, MTX_DEF);
lockinit(&nfsv4root_mnt.mnt_explock, PVFS, "explock", 0, 0);
nfsrvd_initcache();
nfsd_init();
NFSD_LOCK();
nfsrvd_init(0);
NFSD_UNLOCK();
nfsd_mntinit();
#ifdef VV_DISABLEDELEG
vn_deleg_ops.vndeleg_recall = nfsd_recalldelegation;
vn_deleg_ops.vndeleg_disable = nfsd_disabledelegation;
#endif
nfsd_call_servertimer = nfsrv_servertimer;
nfsd_call_nfsd = nfssvc_nfsd;
loaded = 1;
break;
case MOD_UNLOAD:
if (newnfs_numnfsd != 0) {
error = EBUSY;
break;
}
#ifdef VV_DISABLEDELEG
vn_deleg_ops.vndeleg_recall = NULL;
vn_deleg_ops.vndeleg_disable = NULL;
#endif
nfsd_call_servertimer = NULL;
nfsd_call_nfsd = NULL;
/* Clean out all NFSv4 state. */
nfsrv_throwawayallstate(curthread);
/* Clean the NFS server reply cache */
nfsrvd_cleancache();
/* Free up the krpc server pool. */
if (nfsrvd_pool != NULL)
svcpool_destroy(nfsrvd_pool);
/* and get rid of the locks */
for (i = 0; i < NFSRVCACHE_HASHSIZE; i++) {
mtx_destroy(&nfsrchash_table[i].mtx);
mtx_destroy(&nfsrcahash_table[i].mtx);
}
mtx_destroy(&nfsrc_udpmtx);
mtx_destroy(&nfs_v4root_mutex);
mtx_destroy(&nfsv4root_mnt.mnt_mtx);
mtx_destroy(&nfsrv_dontlistlock_mtx);
mtx_destroy(&nfsrv_recalllock_mtx);
for (i = 0; i < nfsrv_sessionhashsize; i++)
mtx_destroy(&nfssessionhash[i].mtx);
if (nfslayouthash != NULL) {
for (i = 0; i < nfsrv_layouthashsize; i++)
mtx_destroy(&nfslayouthash[i].mtx);
free(nfslayouthash, M_NFSDSESSION);
Rework NFS Duplicate Request Cache cleanup logic. - Introduce additional hash to group requests by hash of sockref. This allows to process TCP acknowledgements without looping though all the cache, and as result allows to do it every time. - Indroduce additional callbacks to notify application layer about sockets disconnection. Without this last few requests processed just before socket disconnection never processed their ACKs and stuck in cache for many hours. - Implement transport-specific method for tracking reply acknowledgements. New implementation does not cross multiple stack layers to get the data and does not have race conditions that previously made some requests stuck in cache. This could be done more efficiently at sockbuf layer, but that would broke some KBIs, while I don't know other consumers for it aside NFS. - Instead of traversing all DRC twice per request, run cleaning only once per request, and except in some conditions traverse only single hash slot at a time. Together this limits NFS DRC growth only to situations of real connectivity problems. If network is working well, and so all replies are acknowledged, cache remains almost empty even after hours of heavy load. Without this change on the same test cache was growing to many thousand requests even with perfectly working local network. As another result this reduces CPU time spent on the DRC handling during SPEC NFS benchmark from about 10% to 0.5%. Sponsored by: iXsystems, Inc.
2014-01-03 15:09:59 +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
lockdestroy(&nfsv4root_mnt.mnt_explock);
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(nfsclienthash, M_NFSDCLIENT);
free(nfslockhash, M_NFSDLOCKFILE);
free(nfssessionhash, M_NFSDSESSION);
Add the experimental nfs subtree to the kernel, that includes support for NFSv4 as well as NFSv2 and 3. It lives in 3 subdirs under sys/fs: nfs - functions that are common to the client and server nfsclient - a mutation of sys/nfsclient that call generic functions to do RPCs and handle state. As such, it retains the buffer cache handling characteristics and vnode semantics that are found in sys/nfsclient, for the most part. nfsserver - the server. It includes a DRC designed specifically for NFSv4, that is used instead of the generic DRC in sys/rpc. The build glue will be checked in later, so at this point, it consists of 3 new subdirs that should not affect kernel building. Approved by: kib (mentor)
2009-05-04 15:23:58 +00:00
loaded = 0;
break;
default:
error = EOPNOTSUPP;
break;
}
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
}
static moduledata_t nfsd_mod = {
"nfsd",
nfsd_modevent,
NULL,
};
DECLARE_MODULE(nfsd, nfsd_mod, SI_SUB_VFS, SI_ORDER_ANY);
/* So that loader and kldload(2) can find us, wherever we are.. */
MODULE_VERSION(nfsd, 1);
MODULE_DEPEND(nfsd, nfscommon, 1, 1, 1);
Fix the experimental nfs server so that it depends on the nlm, since it now calls nlm_acquire_next_sysid(). Approved by: kib (mentor)
2009-05-22 01:15:07 +00:00
MODULE_DEPEND(nfsd, nfslockd, 1, 1, 1);
Add MODULE_DEPEND() macros to the experimental NFS client and server so that the modules will load when kernels are built with none of the NFS* configuration options specified. I believe this resolves the problems reported by PR kern/144458 and the email on freebsd-stable@ posted by Dmitry Pryanishnikov on June 13. Tested by: kib PR: kern/144458 Reviewed by: kib MFC after: 1 week
2010-06-15 00:25:04 +00:00
MODULE_DEPEND(nfsd, krpc, 1, 1, 1);
MODULE_DEPEND(nfsd, nfssvc, 1, 1, 1);
Reference in New Issue Copy Permalink