freebsd-skq/sys/nfsserver/nfs_srvsubs.c

1533 lines
34 KiB
C
Raw Normal View History

/*-
1994-05-24 10:09:53 +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.
* 4. Neither the name of the University nor the names of its contributors
* 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.
*
* @(#)nfs_subs.c 8.8 (Berkeley) 5/22/95
1994-05-24 10:09:53 +00:00
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
1994-05-24 10:09:53 +00:00
/*
* These functions support the macros and help fiddle mbuf chains for
* the nfs op functions. They do things like create the rpc header and
* copy data between mbuf chains and uio lists.
*/
#include "opt_inet6.h"
1994-05-24 10:09:53 +00:00
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/bio.h>
#include <sys/buf.h>
#include <sys/proc.h>
1994-05-24 10:09:53 +00:00
#include <sys/mount.h>
#include <sys/vnode.h>
#include <sys/namei.h>
#include <sys/mbuf.h>
#include <sys/refcount.h>
1994-05-24 10:09:53 +00:00
#include <sys/socket.h>
#include <sys/stat.h>
#include <sys/malloc.h>
#include <sys/module.h>
1994-09-22 22:10:49 +00:00
#include <sys/sysent.h>
#include <sys/syscall.h>
#include <sys/sysproto.h>
1994-05-24 10:09:53 +00:00
#include <vm/vm.h>
#include <vm/vm_object.h>
#include <vm/vm_extern.h>
#include <vm/uma.h>
1994-05-24 10:09:53 +00:00
#include <nfs/rpcv2.h>
#include <nfs/nfsproto.h>
#include <nfsserver/nfs.h>
1994-05-24 10:09:53 +00:00
#include <nfs/xdr_subs.h>
#include <nfsserver/nfsm_subs.h>
1994-05-24 10:09:53 +00:00
#include <netinet/in.h>
/*
* Data items converted to xdr at startup, since they are constant
* This is kinda hokey, but may save a little time doing byte swaps
*/
u_int32_t nfsrv_nfs_xdrneg1;
u_int32_t nfsrv_rpc_call, nfsrv_rpc_vers, nfsrv_rpc_reply,
nfsrv_rpc_msgdenied, nfsrv_rpc_autherr,
nfsrv_rpc_mismatch, nfsrv_rpc_auth_unix, nfsrv_rpc_msgaccepted;
u_int32_t nfsrv_nfs_prog, nfsrv_nfs_true, nfsrv_nfs_false;
1994-05-24 10:09:53 +00:00
/* And other global data */
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
static const nfstype nfsv2_type[9] = { NFNON, NFREG, NFDIR, NFBLK, NFCHR,
NFLNK, NFNON, NFCHR, NFNON };
#define vtonfsv2_type(a) txdr_unsigned(nfsv2_type[((int32_t)(a))])
#define vtonfsv3_mode(m) txdr_unsigned((m) & ALLPERMS)
1995-12-17 21:14:36 +00:00
int nfsrv_ticks;
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
#ifdef NFS_LEGACYRPC
struct nfssvc_sockhead nfssvc_sockhead;
int nfssvc_sockhead_flag;
struct nfsd_head nfsd_head;
int nfsd_head_flag;
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
#endif
static int nfssvc_offset = SYS_nfssvc;
static struct sysent nfssvc_prev_sysent;
MAKE_SYSENT(nfssvc);
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
struct mtx nfsd_mtx;
/*
* Mapping of old NFS Version 2 RPC numbers to generic numbers.
*/
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
const int nfsrv_nfsv3_procid[NFS_NPROCS] = {
NFSPROC_NULL,
NFSPROC_GETATTR,
NFSPROC_SETATTR,
NFSPROC_NOOP,
NFSPROC_LOOKUP,
NFSPROC_READLINK,
NFSPROC_READ,
NFSPROC_NOOP,
NFSPROC_WRITE,
NFSPROC_CREATE,
NFSPROC_REMOVE,
NFSPROC_RENAME,
NFSPROC_LINK,
NFSPROC_SYMLINK,
NFSPROC_MKDIR,
NFSPROC_RMDIR,
NFSPROC_READDIR,
NFSPROC_FSSTAT,
NFSPROC_NOOP,
NFSPROC_NOOP,
NFSPROC_NOOP,
NFSPROC_NOOP,
NFSPROC_NOOP,
};
/*
* and the reverse mapping from generic to Version 2 procedure numbers
*/
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
const int nfsrvv2_procid[NFS_NPROCS] = {
NFSV2PROC_NULL,
NFSV2PROC_GETATTR,
NFSV2PROC_SETATTR,
NFSV2PROC_LOOKUP,
NFSV2PROC_NOOP,
NFSV2PROC_READLINK,
NFSV2PROC_READ,
NFSV2PROC_WRITE,
NFSV2PROC_CREATE,
NFSV2PROC_MKDIR,
NFSV2PROC_SYMLINK,
NFSV2PROC_CREATE,
NFSV2PROC_REMOVE,
NFSV2PROC_RMDIR,
NFSV2PROC_RENAME,
NFSV2PROC_LINK,
NFSV2PROC_READDIR,
NFSV2PROC_NOOP,
NFSV2PROC_STATFS,
NFSV2PROC_NOOP,
NFSV2PROC_NOOP,
NFSV2PROC_NOOP,
NFSV2PROC_NOOP,
};
/*
* Maps errno values to nfs error numbers.
* Use 0 (which gets converted to NFSERR_IO) as the catch all for ones not
* specifically defined in RFC 1094.
*/
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
static const u_char nfsrv_v2errmap[ELAST] = {
NFSERR_PERM, NFSERR_NOENT, 0, 0, 0,
NFSERR_NXIO, 0, 0, 0, 0,
0, 0, NFSERR_ACCES, 0, 0,
0, NFSERR_EXIST, 0, NFSERR_NODEV, NFSERR_NOTDIR,
NFSERR_ISDIR, 0, 0, 0, 0,
0, NFSERR_FBIG, NFSERR_NOSPC, 0, NFSERR_ROFS,
0, 0, 0, 0, 0,
0, 0, 0, 0, 0,
0, 0, 0, 0, 0,
0, 0, 0, 0, 0,
0, 0, 0, 0, 0,
0, 0, 0, 0, 0,
0, 0, NFSERR_NAMETOL, 0, 0,
NFSERR_NOTEMPTY, 0, 0, NFSERR_DQUOT, NFSERR_STALE,
0
};
/*
* Maps errno values to nfs error numbers.
* Although it is not obvious whether or not NFS clients really care if
* a returned error value is in the specified list for the procedure, the
* safest thing to do is filter them appropriately. For Version 2, the
* X/Open XNFS document is the only specification that defines error values
* for each RPC (The RFC simply lists all possible error values for all RPCs),
* so I have decided to not do this for Version 2.
* The first entry is the default error return and the rest are the valid
* errors for that RPC in increasing numeric order.
*/
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
static const short nfsv3err_null[] = {
0,
0,
};
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
static const short nfsv3err_getattr[] = {
NFSERR_IO,
NFSERR_IO,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_SERVERFAULT,
0,
};
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
static const short nfsv3err_setattr[] = {
NFSERR_IO,
NFSERR_PERM,
NFSERR_IO,
NFSERR_ACCES,
NFSERR_INVAL,
NFSERR_NOSPC,
NFSERR_ROFS,
NFSERR_DQUOT,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_NOT_SYNC,
NFSERR_SERVERFAULT,
0,
};
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
static const short nfsv3err_lookup[] = {
NFSERR_IO,
NFSERR_NOENT,
NFSERR_IO,
NFSERR_ACCES,
NFSERR_NOTDIR,
NFSERR_NAMETOL,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_SERVERFAULT,
0,
};
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
static const short nfsv3err_access[] = {
NFSERR_IO,
NFSERR_IO,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_SERVERFAULT,
0,
};
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
static const short nfsv3err_readlink[] = {
NFSERR_IO,
NFSERR_IO,
NFSERR_ACCES,
NFSERR_INVAL,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_NOTSUPP,
NFSERR_SERVERFAULT,
0,
};
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
static const short nfsv3err_read[] = {
NFSERR_IO,
NFSERR_IO,
NFSERR_NXIO,
NFSERR_ACCES,
NFSERR_INVAL,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_SERVERFAULT,
0,
};
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
static const short nfsv3err_write[] = {
NFSERR_IO,
NFSERR_IO,
NFSERR_ACCES,
NFSERR_INVAL,
NFSERR_FBIG,
NFSERR_NOSPC,
NFSERR_ROFS,
NFSERR_DQUOT,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_SERVERFAULT,
0,
};
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
static const short nfsv3err_create[] = {
NFSERR_IO,
NFSERR_IO,
NFSERR_ACCES,
NFSERR_EXIST,
NFSERR_NOTDIR,
NFSERR_NOSPC,
NFSERR_ROFS,
NFSERR_NAMETOL,
NFSERR_DQUOT,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_NOTSUPP,
NFSERR_SERVERFAULT,
0,
};
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
static const short nfsv3err_mkdir[] = {
NFSERR_IO,
NFSERR_IO,
NFSERR_ACCES,
NFSERR_EXIST,
NFSERR_NOTDIR,
NFSERR_NOSPC,
NFSERR_ROFS,
NFSERR_NAMETOL,
NFSERR_DQUOT,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_NOTSUPP,
NFSERR_SERVERFAULT,
0,
};
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
static const short nfsv3err_symlink[] = {
NFSERR_IO,
NFSERR_IO,
NFSERR_ACCES,
NFSERR_EXIST,
NFSERR_NOTDIR,
NFSERR_NOSPC,
NFSERR_ROFS,
NFSERR_NAMETOL,
NFSERR_DQUOT,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_NOTSUPP,
NFSERR_SERVERFAULT,
0,
};
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
static const short nfsv3err_mknod[] = {
NFSERR_IO,
NFSERR_IO,
NFSERR_ACCES,
NFSERR_EXIST,
NFSERR_NOTDIR,
NFSERR_NOSPC,
NFSERR_ROFS,
NFSERR_NAMETOL,
NFSERR_DQUOT,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_NOTSUPP,
NFSERR_SERVERFAULT,
NFSERR_BADTYPE,
0,
};
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
static const short nfsv3err_remove[] = {
NFSERR_IO,
NFSERR_NOENT,
NFSERR_IO,
NFSERR_ACCES,
NFSERR_NOTDIR,
NFSERR_ROFS,
NFSERR_NAMETOL,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_SERVERFAULT,
0,
};
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
static const short nfsv3err_rmdir[] = {
NFSERR_IO,
NFSERR_NOENT,
NFSERR_IO,
NFSERR_ACCES,
NFSERR_EXIST,
NFSERR_NOTDIR,
NFSERR_INVAL,
NFSERR_ROFS,
NFSERR_NAMETOL,
NFSERR_NOTEMPTY,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_NOTSUPP,
NFSERR_SERVERFAULT,
0,
};
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
static const short nfsv3err_rename[] = {
NFSERR_IO,
NFSERR_NOENT,
NFSERR_IO,
NFSERR_ACCES,
NFSERR_EXIST,
NFSERR_XDEV,
NFSERR_NOTDIR,
NFSERR_ISDIR,
NFSERR_INVAL,
NFSERR_NOSPC,
NFSERR_ROFS,
NFSERR_MLINK,
NFSERR_NAMETOL,
NFSERR_NOTEMPTY,
NFSERR_DQUOT,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_NOTSUPP,
NFSERR_SERVERFAULT,
0,
};
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
static const short nfsv3err_link[] = {
NFSERR_IO,
NFSERR_IO,
NFSERR_ACCES,
NFSERR_EXIST,
NFSERR_XDEV,
NFSERR_NOTDIR,
NFSERR_INVAL,
NFSERR_NOSPC,
NFSERR_ROFS,
NFSERR_MLINK,
NFSERR_NAMETOL,
NFSERR_DQUOT,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_NOTSUPP,
NFSERR_SERVERFAULT,
0,
};
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
static const short nfsv3err_readdir[] = {
NFSERR_IO,
NFSERR_IO,
NFSERR_ACCES,
NFSERR_NOTDIR,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_BAD_COOKIE,
NFSERR_TOOSMALL,
NFSERR_SERVERFAULT,
0,
};
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
static const short nfsv3err_readdirplus[] = {
NFSERR_IO,
NFSERR_IO,
NFSERR_ACCES,
NFSERR_NOTDIR,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_BAD_COOKIE,
NFSERR_NOTSUPP,
NFSERR_TOOSMALL,
NFSERR_SERVERFAULT,
0,
};
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
static const short nfsv3err_fsstat[] = {
NFSERR_IO,
NFSERR_IO,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_SERVERFAULT,
0,
};
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
static const short nfsv3err_fsinfo[] = {
NFSERR_STALE,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_SERVERFAULT,
0,
};
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
static const short nfsv3err_pathconf[] = {
NFSERR_STALE,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_SERVERFAULT,
0,
};
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
static const short nfsv3err_commit[] = {
NFSERR_IO,
NFSERR_IO,
NFSERR_STALE,
NFSERR_BADHANDLE,
NFSERR_SERVERFAULT,
0,
};
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
static const short *nfsrv_v3errmap[] = {
nfsv3err_null,
nfsv3err_getattr,
nfsv3err_setattr,
nfsv3err_lookup,
nfsv3err_access,
nfsv3err_readlink,
nfsv3err_read,
nfsv3err_write,
nfsv3err_create,
nfsv3err_mkdir,
nfsv3err_symlink,
nfsv3err_mknod,
nfsv3err_remove,
nfsv3err_rmdir,
nfsv3err_rename,
nfsv3err_link,
nfsv3err_readdir,
nfsv3err_readdirplus,
nfsv3err_fsstat,
nfsv3err_fsinfo,
nfsv3err_pathconf,
nfsv3err_commit,
};
1994-05-24 10:09:53 +00:00
/*
* Called once to initialize data structures...
*/
static int
nfsrv_modevent(module_t mod, int type, void *data)
1994-05-24 10:09:53 +00:00
{
static int registered;
int error = 0;
switch (type) {
case MOD_LOAD:
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
mtx_init(&nfsd_mtx, "nfsd_mtx", NULL, MTX_DEF);
nfsrv_rpc_vers = txdr_unsigned(RPC_VER2);
nfsrv_rpc_call = txdr_unsigned(RPC_CALL);
nfsrv_rpc_reply = txdr_unsigned(RPC_REPLY);
nfsrv_rpc_msgdenied = txdr_unsigned(RPC_MSGDENIED);
nfsrv_rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED);
nfsrv_rpc_mismatch = txdr_unsigned(RPC_MISMATCH);
nfsrv_rpc_autherr = txdr_unsigned(RPC_AUTHERR);
nfsrv_rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX);
nfsrv_nfs_prog = txdr_unsigned(NFS_PROG);
nfsrv_nfs_true = txdr_unsigned(TRUE);
nfsrv_nfs_false = txdr_unsigned(FALSE);
nfsrv_nfs_xdrneg1 = txdr_unsigned(-1);
nfsrv_ticks = (hz * NFS_TICKINTVL + 500) / 1000;
if (nfsrv_ticks < 1)
nfsrv_ticks = 1;
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
#ifdef NFS_LEGACYRPC
nfsrv_initcache(); /* Init the server request cache */
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
NFSD_LOCK();
nfsrv_init(0); /* Init server data structures */
callout_init(&nfsrv_callout, CALLOUT_MPSAFE);
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
NFSD_UNLOCK();
nfsrv_timer(0);
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
#else
NFSD_LOCK();
nfsrv_init(0); /* Init server data structures */
NFSD_UNLOCK();
#endif
error = syscall_register(&nfssvc_offset, &nfssvc_sysent,
&nfssvc_prev_sysent);
if (error)
break;
registered = 1;
break;
case MOD_UNLOAD:
if (nfsrv_numnfsd != 0) {
error = EBUSY;
break;
}
if (registered)
syscall_deregister(&nfssvc_offset, &nfssvc_prev_sysent);
callout_drain(&nfsrv_callout);
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
#ifdef NFS_LEGACYRPC
nfsrv_destroycache(); /* Free the server request cache */
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
#endif
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
mtx_destroy(&nfsd_mtx);
break;
default:
error = EOPNOTSUPP;
break;
}
return error;
1994-05-24 10:09:53 +00:00
}
static moduledata_t nfsserver_mod = {
"nfsserver",
nfsrv_modevent,
NULL,
};
DECLARE_MODULE(nfsserver, nfsserver_mod, SI_SUB_VFS, SI_ORDER_ANY);
/* So that loader and kldload(2) can find us, wherever we are.. */
MODULE_VERSION(nfsserver, 1);
#ifndef NFS_LEGACYRPC
2008-11-06 10:53:35 +00:00
MODULE_DEPEND(nfsserver, krpc, 1, 1, 1);
#endif
1994-05-24 10:09:53 +00:00
/*
* Set up nameidata for a lookup() call and do it.
*
* If pubflag is set, this call is done for a lookup operation on the
* public filehandle. In that case we allow crossing mountpoints and
* absolute pathnames. However, the caller is expected to check that
* the lookup result is within the public fs, and deny access if
* it is not.
*
* nfs_namei() clears out garbage fields that namei() might leave garbage.
* This is mainly ni_vp and ni_dvp when an error occurs, and ni_dvp when no
* error occurs but the parent was not requested.
*
* dirp may be set whether an error is returned or not, and must be
* released by the caller.
1994-05-24 10:09:53 +00:00
*/
int
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
nfs_namei(struct nameidata *ndp, struct nfsrv_descript *nfsd,
fhandle_t *fhp, int len, struct nfssvc_sock *slp,
struct sockaddr *nam, struct mbuf **mdp,
caddr_t *dposp, struct vnode **retdirp, int v3, struct vattr *retdirattrp,
int *retdirattr_retp, int pubflag)
1994-05-24 10:09:53 +00:00
{
int i, rem;
struct mbuf *md;
char *fromcp, *tocp, *cp;
struct iovec aiov;
struct uio auio;
1994-05-24 10:09:53 +00:00
struct vnode *dp;
int error, rdonly, linklen;
1994-05-24 10:09:53 +00:00
struct componentname *cnp = &ndp->ni_cnd;
int lockleaf = (cnp->cn_flags & LOCKLEAF) != 0;
int dvfslocked;
int vfslocked;
1994-05-24 10:09:53 +00:00
vfslocked = 0;
dvfslocked = 0;
*retdirp = NULL;
cnp->cn_flags |= NOMACCHECK;
cnp->cn_pnbuf = uma_zalloc(namei_zone, M_WAITOK);
1994-05-24 10:09:53 +00:00
/*
* Copy the name from the mbuf list to ndp->ni_pnbuf
* and set the various ndp fields appropriately.
*/
fromcp = *dposp;
tocp = cnp->cn_pnbuf;
md = *mdp;
rem = mtod(md, caddr_t) + md->m_len - fromcp;
for (i = 0; i < len; i++) {
while (rem == 0) {
md = md->m_next;
if (md == NULL) {
error = EBADRPC;
goto out;
1994-05-24 10:09:53 +00:00
}
fromcp = mtod(md, caddr_t);
rem = md->m_len;
}
if (*fromcp == '\0' || (!pubflag && *fromcp == '/')) {
error = EACCES;
goto out;
1994-05-24 10:09:53 +00:00
}
*tocp++ = *fromcp++;
rem--;
}
*tocp = '\0';
*mdp = md;
*dposp = fromcp;
len = nfsm_rndup(len)-len;
if (len > 0) {
if (rem >= len)
*dposp += len;
else if ((error = nfs_adv(mdp, dposp, len, rem)) != 0)
goto out;
1994-05-24 10:09:53 +00:00
}
1994-05-24 10:09:53 +00:00
/*
* Extract and set starting directory.
*/
error = nfsrv_fhtovp(fhp, FALSE, &dp, &dvfslocked,
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
nfsd, slp, nam, &rdonly, pubflag);
if (error)
1994-05-24 10:09:53 +00:00
goto out;
vfslocked = VFS_LOCK_GIANT(dp->v_mount);
1994-05-24 10:09:53 +00:00
if (dp->v_type != VDIR) {
vrele(dp);
1994-05-24 10:09:53 +00:00
error = ENOTDIR;
goto out;
}
if (rdonly)
cnp->cn_flags |= RDONLY;
/*
* Set return directory. Reference to dp is implicitly transfered
* to the returned pointer
*/
*retdirp = dp;
if (v3) {
vn_lock(dp, LK_EXCLUSIVE | LK_RETRY);
*retdirattr_retp = VOP_GETATTR(dp, retdirattrp,
ndp->ni_cnd.cn_cred);
VOP_UNLOCK(dp, 0);
}
if (pubflag) {
/*
* Oh joy. For WebNFS, handle those pesky '%' escapes,
* and the 'native path' indicator.
*/
cp = uma_zalloc(namei_zone, M_WAITOK);
fromcp = cnp->cn_pnbuf;
tocp = cp;
if ((unsigned char)*fromcp >= WEBNFS_SPECCHAR_START) {
switch ((unsigned char)*fromcp) {
case WEBNFS_NATIVE_CHAR:
/*
* 'Native' path for us is the same
* as a path according to the NFS spec,
* just skip the escape char.
*/
fromcp++;
break;
/*
* More may be added in the future, range 0x80-0xff
*/
default:
error = EIO;
uma_zfree(namei_zone, cp);
goto out;
}
}
/*
* Translate the '%' escapes, URL-style.
*/
while (*fromcp != '\0') {
if (*fromcp == WEBNFS_ESC_CHAR) {
if (fromcp[1] != '\0' && fromcp[2] != '\0') {
fromcp++;
*tocp++ = HEXSTRTOI(fromcp);
fromcp += 2;
continue;
} else {
error = ENOENT;
uma_zfree(namei_zone, cp);
goto out;
}
} else
*tocp++ = *fromcp++;
}
*tocp = '\0';
uma_zfree(namei_zone, cnp->cn_pnbuf);
cnp->cn_pnbuf = cp;
}
ndp->ni_pathlen = (tocp - cnp->cn_pnbuf) + 1;
ndp->ni_segflg = UIO_SYSSPACE;
if (pubflag) {
ndp->ni_rootdir = rootvnode;
ndp->ni_loopcnt = 0;
if (cnp->cn_pnbuf[0] == '/') {
int tvfslocked;
tvfslocked = VFS_LOCK_GIANT(rootvnode->v_mount);
VFS_UNLOCK_GIANT(vfslocked);
dp = rootvnode;
vfslocked = tvfslocked;
}
} else {
cnp->cn_flags |= NOCROSSMOUNT;
}
/*
* Initialize for scan, set ni_startdir and bump ref on dp again
* because lookup() will dereference ni_startdir.
*/
cnp->cn_thread = curthread;
VREF(dp);
1994-05-24 10:09:53 +00:00
ndp->ni_startdir = dp;
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.
*/
if (vfslocked)
ndp->ni_cnd.cn_flags |= GIANTHELD;
error = lookup(ndp);
vfslocked = (ndp->ni_cnd.cn_flags & GIANTHELD) != 0;
ndp->ni_cnd.cn_flags &= ~GIANTHELD;
if (error)
break;
/*
* Check for encountering a symbolic link. Trivial
* termination occurs if no symlink encountered.
* Note: zfree is safe because error is 0, so we will
* not zfree it again when we break.
*/
if ((cnp->cn_flags & ISSYMLINK) == 0) {
if (cnp->cn_flags & (SAVENAME | SAVESTART))
cnp->cn_flags |= HASBUF;
else
uma_zfree(namei_zone, cnp->cn_pnbuf);
if (ndp->ni_vp && !lockleaf)
VOP_UNLOCK(ndp->ni_vp, 0);
break;
}
/*
* Validate symlink
*/
1994-05-24 10:09:53 +00:00
if ((cnp->cn_flags & LOCKPARENT) && ndp->ni_pathlen == 1)
VOP_UNLOCK(ndp->ni_dvp, 0);
if (!pubflag) {
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:
vput(ndp->ni_vp);
vrele(ndp->ni_dvp);
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) {
bcopy(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;
1994-05-24 10:09:53 +00:00
}
if (!lockleaf)
cnp->cn_flags &= ~LOCKLEAF;
if (cnp->cn_flags & GIANTHELD) {
mtx_unlock(&Giant);
cnp->cn_flags &= ~GIANTHELD;
}
/*
* nfs_namei() guarentees that fields will not contain garbage
* whether an error occurs or not. This allows the caller to track
* cleanup state trivially.
*/
1994-05-24 10:09:53 +00:00
out:
if (error) {
uma_zfree(namei_zone, cnp->cn_pnbuf);
ndp->ni_vp = NULL;
ndp->ni_dvp = NULL;
ndp->ni_startdir = NULL;
cnp->cn_flags &= ~HASBUF;
VFS_UNLOCK_GIANT(vfslocked);
vfslocked = 0;
} else if ((ndp->ni_cnd.cn_flags & (WANTPARENT|LOCKPARENT)) == 0) {
ndp->ni_dvp = NULL;
}
/*
* This differs from normal namei() in that even on failure we may
* return with Giant held due to the dirp return. Make sure we only
* have not recursed however. The calling code only expects to drop
* one acquire.
*/
if (vfslocked || dvfslocked)
ndp->ni_cnd.cn_flags |= GIANTHELD;
if (vfslocked && dvfslocked)
VFS_UNLOCK_GIANT(vfslocked);
1994-05-24 10:09:53 +00:00
return (error);
}
/*
* A fiddled version of m_adj() that ensures null fill to a long
* boundary and only trims off the back end
*/
void
nfsm_adj(struct mbuf *mp, int len, int nul)
1994-05-24 10:09:53 +00:00
{
struct mbuf *m;
int count, i;
char *cp;
1994-05-24 10:09:53 +00:00
/*
* Trim from tail. Scan the mbuf chain,
* calculating its length and finding the last mbuf.
* If the adjustment only affects this mbuf, then just
* adjust and return. Otherwise, rescan and truncate
* after the remaining size.
*/
count = 0;
m = mp;
for (;;) {
count += m->m_len;
if (m->m_next == NULL)
1994-05-24 10:09:53 +00:00
break;
m = m->m_next;
}
if (m->m_len > len) {
m->m_len -= len;
if (nul > 0) {
cp = mtod(m, caddr_t)+m->m_len-nul;
for (i = 0; i < nul; i++)
*cp++ = '\0';
}
return;
}
count -= len;
if (count < 0)
count = 0;
/*
* Correct length for chain is "count".
* Find the mbuf with last data, adjust its length,
* and toss data from remaining mbufs on chain.
*/
for (m = mp; m; m = m->m_next) {
if (m->m_len >= count) {
m->m_len = count;
if (nul > 0) {
cp = mtod(m, caddr_t)+m->m_len-nul;
for (i = 0; i < nul; i++)
*cp++ = '\0';
}
if (m->m_next != NULL) {
m_freem(m->m_next);
m->m_next = NULL;
}
1994-05-24 10:09:53 +00:00
break;
}
count -= m->m_len;
}
}
/*
* Make these functions instead of macros, so that the kernel text size
* doesn't get too big...
*/
void
nfsm_srvwcc(struct nfsrv_descript *nfsd, int before_ret,
struct vattr *before_vap, int after_ret, struct vattr *after_vap,
struct mbuf **mbp, char **bposp)
{
struct mbuf *mb = *mbp;
char *bpos = *bposp;
u_int32_t *tl;
if (before_ret) {
tl = nfsm_build(u_int32_t *, NFSX_UNSIGNED);
*tl = nfsrv_nfs_false;
} else {
tl = nfsm_build(u_int32_t *, 7 * NFSX_UNSIGNED);
*tl++ = nfsrv_nfs_true;
txdr_hyper(before_vap->va_size, tl);
tl += 2;
txdr_nfsv3time(&(before_vap->va_mtime), tl);
tl += 2;
txdr_nfsv3time(&(before_vap->va_ctime), tl);
}
*bposp = bpos;
*mbp = mb;
nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp);
}
void
nfsm_srvpostopattr(struct nfsrv_descript *nfsd, int after_ret,
struct vattr *after_vap, struct mbuf **mbp, char **bposp)
{
struct mbuf *mb = *mbp;
char *bpos = *bposp;
u_int32_t *tl;
struct nfs_fattr *fp;
if (after_ret) {
tl = nfsm_build(u_int32_t *, NFSX_UNSIGNED);
*tl = nfsrv_nfs_false;
} else {
tl = nfsm_build(u_int32_t *, NFSX_UNSIGNED + NFSX_V3FATTR);
*tl++ = nfsrv_nfs_true;
fp = (struct nfs_fattr *)tl;
nfsm_srvfattr(nfsd, after_vap, fp);
}
*mbp = mb;
*bposp = bpos;
}
void
nfsm_srvfattr(struct nfsrv_descript *nfsd, struct vattr *vap,
struct nfs_fattr *fp)
{
fp->fa_nlink = txdr_unsigned(vap->va_nlink);
fp->fa_uid = txdr_unsigned(vap->va_uid);
fp->fa_gid = txdr_unsigned(vap->va_gid);
if (nfsd->nd_flag & ND_NFSV3) {
fp->fa_type = vtonfsv3_type(vap->va_type);
fp->fa_mode = vtonfsv3_mode(vap->va_mode);
txdr_hyper(vap->va_size, &fp->fa3_size);
txdr_hyper(vap->va_bytes, &fp->fa3_used);
fp->fa3_rdev.specdata1 = txdr_unsigned(major(vap->va_rdev));
fp->fa3_rdev.specdata2 = txdr_unsigned(minor(vap->va_rdev));
fp->fa3_fsid.nfsuquad[0] = 0;
fp->fa3_fsid.nfsuquad[1] = txdr_unsigned(vap->va_fsid);
fp->fa3_fileid.nfsuquad[0] = 0;
fp->fa3_fileid.nfsuquad[1] = txdr_unsigned(vap->va_fileid);
txdr_nfsv3time(&vap->va_atime, &fp->fa3_atime);
txdr_nfsv3time(&vap->va_mtime, &fp->fa3_mtime);
txdr_nfsv3time(&vap->va_ctime, &fp->fa3_ctime);
} else {
fp->fa_type = vtonfsv2_type(vap->va_type);
fp->fa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode);
fp->fa2_size = txdr_unsigned(vap->va_size);
fp->fa2_blocksize = txdr_unsigned(vap->va_blocksize);
if (vap->va_type == VFIFO)
fp->fa2_rdev = 0xffffffff;
else
fp->fa2_rdev = txdr_unsigned(vap->va_rdev);
fp->fa2_blocks = txdr_unsigned(vap->va_bytes / NFS_FABLKSIZE);
fp->fa2_fsid = txdr_unsigned(vap->va_fsid);
fp->fa2_fileid = txdr_unsigned(vap->va_fileid);
txdr_nfsv2time(&vap->va_atime, &fp->fa2_atime);
txdr_nfsv2time(&vap->va_mtime, &fp->fa2_mtime);
txdr_nfsv2time(&vap->va_ctime, &fp->fa2_ctime);
}
}
1994-05-24 10:09:53 +00:00
/*
* nfsrv_fhtovp() - convert a fh to a vnode ptr (optionally locked)
* - look up fsid in mount list (if not found ret error)
* - get vp and export rights by calling VFS_FHTOVP()
* - if cred->cr_uid == 0 or MNT_EXPORTANON set it to credanon
* - if not lockflag unlock it with VOP_UNLOCK()
*/
int
nfsrv_fhtovp(fhandle_t *fhp, int lockflag, struct vnode **vpp, int *vfslockedp,
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
struct nfsrv_descript *nfsd, struct nfssvc_sock *slp,
struct sockaddr *nam, int *rdonlyp, int pubflag)
1994-05-24 10:09:53 +00:00
{
struct mount *mp;
int i;
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
struct ucred *cred, *credanon;
1994-05-24 10:09:53 +00:00
int error, exflags;
#ifdef MNT_EXNORESPORT /* XXX needs mountd and /etc/exports help yet */
struct sockaddr_int *saddr;
#endif
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
int credflavor;
int vfslocked;
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
int numsecflavors, *secflavors;
int authsys;
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
int v3 = nfsd->nd_flag & ND_NFSV3;
int mountreq;
1994-05-24 10:09:53 +00:00
*vfslockedp = 0;
*vpp = NULL;
if (nfs_ispublicfh(fhp)) {
if (!pubflag || !nfs_pub.np_valid)
return (ESTALE);
fhp = &nfs_pub.np_handle;
}
mp = vfs_busyfs(&fhp->fh_fsid);
if (!mp)
1994-05-24 10:09:53 +00:00
return (ESTALE);
vfslocked = VFS_LOCK_GIANT(mp);
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
error = VFS_CHECKEXP(mp, nam, &exflags, &credanon,
&numsecflavors, &secflavors);
if (error) {
vfs_unbusy(mp);
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
goto out;
}
if (numsecflavors == 0) {
/*
* This can happen if the system is running with an
* old mountd that doesn't pass in a secflavor list.
*/
numsecflavors = 1;
authsys = RPCAUTH_UNIX;
secflavors = &authsys;
}
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
credflavor = nfsd->nd_credflavor;
for (i = 0; i < numsecflavors; i++) {
if (secflavors[i] == credflavor)
break;
}
if (i == numsecflavors) {
/*
* RFC 2623 section 2.3.2 - allow certain procedures
* used at NFS client mount time even if they have
* weak authentication.
*/
mountreq = FALSE;
if (v3) {
if (nfsd->nd_procnum == NFSPROC_FSINFO
|| nfsd->nd_procnum == NFSPROC_GETATTR)
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
mountreq = TRUE;
} else {
if (nfsd->nd_procnum == NFSPROC_FSSTAT
|| nfsd->nd_procnum == NFSPROC_GETATTR)
mountreq = TRUE;
}
if (!mountreq) {
error = NFSERR_AUTHERR | AUTH_TOOWEAK;
vfs_unbusy(mp);
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
goto out;
}
}
error = VFS_FHTOVP(mp, &fhp->fh_fid, vpp);
vfs_unbusy(mp);
if (error)
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
goto out;
#ifdef MNT_EXNORESPORT
if (!(exflags & (MNT_EXNORESPORT|MNT_EXPUBLIC))) {
saddr = (struct sockaddr_in *)nam;
if ((saddr->sin_family == AF_INET ||
saddr->sin_family == AF_INET6) &&
/* same code for INET and INET6: sin*_port at same offet */
ntohs(saddr->sin_port) >= IPPORT_RESERVED) {
vput(*vpp);
*vpp = NULL;
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
error = NFSERR_AUTHERR | AUTH_TOOWEAK;
}
}
#endif
1994-05-24 10:09:53 +00:00
/*
* Check/setup credentials.
*/
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
cred = nfsd->nd_cr;
if (cred->cr_uid == 0 || (exflags & MNT_EXPORTANON)) {
1994-05-24 10:09:53 +00:00
cred->cr_uid = credanon->cr_uid;
for (i = 0; i < credanon->cr_ngroups && i < NGROUPS; i++)
cred->cr_groups[i] = credanon->cr_groups[i];
cred->cr_ngroups = i;
1994-05-24 10:09:53 +00:00
}
if (exflags & MNT_EXRDONLY)
*rdonlyp = 1;
else
*rdonlyp = 0;
1994-05-24 10:09:53 +00:00
if (!lockflag)
VOP_UNLOCK(*vpp, 0);
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
out:
if (error) {
VFS_UNLOCK_GIANT(vfslocked);
} else
*vfslockedp = vfslocked;
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
return (error);
1994-05-24 10:09:53 +00:00
}
/*
* WebNFS: check if a filehandle is a public filehandle. For v3, this
* means a length of 0, for v2 it means all zeroes. nfsm_srvmtofh has
* transformed this to all zeroes in both cases, so check for it.
*/
int
nfs_ispublicfh(fhandle_t *fhp)
{
char *cp = (char *)fhp;
int i;
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
NFSD_LOCK_DONTCARE();
for (i = 0; i < NFSX_V3FH; i++)
if (*cp++ != 0)
return (FALSE);
return (TRUE);
}
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
#ifdef NFS_LEGACYRPC
1994-05-24 10:09:53 +00:00
/*
* This function compares two net addresses by family and returns TRUE
* if they are the same host.
* If there is any doubt, return FALSE.
* The AF_INET family is handled as a special case so that address mbufs
* don't need to be saved to store "struct in_addr", which is only 4 bytes.
*/
int
netaddr_match(int family, union nethostaddr *haddr, struct sockaddr *nam)
1994-05-24 10:09:53 +00:00
{
struct sockaddr_in *inetaddr;
1994-05-24 10:09:53 +00:00
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
NFSD_LOCK_DONTCARE();
1994-05-24 10:09:53 +00:00
switch (family) {
case AF_INET:
inetaddr = (struct sockaddr_in *)nam;
1994-05-24 10:09:53 +00:00
if (inetaddr->sin_family == AF_INET &&
inetaddr->sin_addr.s_addr == haddr->had_inetaddr)
return (1);
break;
#ifdef INET6
case AF_INET6:
{
register struct sockaddr_in6 *inet6addr1, *inet6addr2;
inet6addr1 = (struct sockaddr_in6 *)nam;
inet6addr2 = (struct sockaddr_in6 *)haddr->had_nam;
/* XXX - should test sin6_scope_id ? */
if (inet6addr1->sin6_family == AF_INET6 &&
IN6_ARE_ADDR_EQUAL(&inet6addr1->sin6_addr,
&inet6addr2->sin6_addr))
return (1);
break;
}
#endif
1994-05-24 10:09:53 +00:00
default:
break;
};
return (0);
}
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
#endif
/*
* Map errnos to NFS error numbers. For Version 3 also filter out error
* numbers not specified for the associated procedure.
*/
int
nfsrv_errmap(struct nfsrv_descript *nd, int err)
{
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
const short *defaulterrp, *errp;
int e;
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
if (nd->nd_flag & ND_NFSV3) {
if (nd->nd_procnum <= NFSPROC_COMMIT) {
errp = defaulterrp = nfsrv_v3errmap[nd->nd_procnum];
while (*++errp) {
if (*errp == err)
return (err);
else if (*errp > err)
break;
}
return ((int)*defaulterrp);
} else
return (err & 0xffff);
}
e = 0;
if (err <= ELAST)
e = nfsrv_v2errmap[err - 1];
if (e != 0)
return (e);
return (NFSERR_IO);
}
/*
* Sort the group list in increasing numerical order.
* (Insertion sort by Chris Torek, who was grossed out by the bubble sort
* that used to be here.)
*/
void
nfsrvw_sort(gid_t *list, int num)
{
int i, j;
gid_t v;
/* Insertion sort. */
for (i = 1; i < num; i++) {
v = list[i];
/* find correct slot for value v, moving others up */
for (j = i; --j >= 0 && v < list[j];)
list[j + 1] = list[j];
list[j + 1] = v;
}
}
/*
* Helper functions for macros.
*/
void
nfsm_srvfhtom_xx(fhandle_t *f, int v3, struct mbuf **mb, caddr_t *bpos)
{
u_int32_t *tl;
if (v3) {
tl = nfsm_build_xx(NFSX_UNSIGNED + NFSX_V3FH, mb, bpos);
*tl++ = txdr_unsigned(NFSX_V3FH);
bcopy(f, tl, NFSX_V3FH);
} else {
tl = nfsm_build_xx(NFSX_V2FH, mb, bpos);
bcopy(f, tl, NFSX_V2FH);
}
}
void
nfsm_srvpostop_fh_xx(fhandle_t *f, struct mbuf **mb, caddr_t *bpos)
{
u_int32_t *tl;
tl = nfsm_build_xx(2 * NFSX_UNSIGNED + NFSX_V3FH, mb, bpos);
*tl++ = nfsrv_nfs_true;
*tl++ = txdr_unsigned(NFSX_V3FH);
bcopy(f, tl, NFSX_V3FH);
}
int
nfsm_srvstrsiz_xx(int *s, int m, struct mbuf **md, caddr_t *dpos)
{
u_int32_t *tl;
tl = nfsm_dissect_xx_nonblock(NFSX_UNSIGNED, md, dpos);
if (tl == NULL)
return EBADRPC;
*s = fxdr_unsigned(int32_t, *tl);
if (*s > m || *s <= 0)
return EBADRPC;
return 0;
}
int
nfsm_srvnamesiz_xx(int *s, int m, struct mbuf **md, caddr_t *dpos)
{
u_int32_t *tl;
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
NFSD_LOCK_DONTCARE();
tl = nfsm_dissect_xx_nonblock(NFSX_UNSIGNED, md, dpos);
if (tl == NULL)
return EBADRPC;
*s = fxdr_unsigned(int32_t, *tl);
if (*s > m)
return NFSERR_NAMETOL;
if (*s <= 0)
return EBADRPC;
return 0;
}
int
nfsm_srvnamesiz0_xx(int *s, int m, struct mbuf **md, caddr_t *dpos)
{
u_int32_t *tl;
tl = nfsm_dissect_xx_nonblock(NFSX_UNSIGNED, md, dpos);
if (tl == NULL)
return EBADRPC;
*s = fxdr_unsigned(int32_t, *tl);
if (*s > m)
return NFSERR_NAMETOL;
if (*s < 0)
return EBADRPC;
return 0;
}
void
nfsm_clget_xx(u_int32_t **tl, struct mbuf *mb, struct mbuf **mp,
char **bp, char **be, caddr_t bpos)
{
struct mbuf *nmp;
NFSD_UNLOCK_ASSERT();
The socket code upcalls into the NFS server using the so_upcall mechanism so that early processing on mbufs can be performed before a context switch to the NFS server threads. Because of this, if the socket code is running without Giant, the NFS server also needs to be able to run the upcall code without relying on the presence on Giant. This change modifies the NFS server to run using a "giant code lock" covering operation of the whole subsystem. Work is in progress to move to data-based locking as part of the NFSv4 server changes. Introduce an NFS server subsystem lock, 'nfsd_mtx', and a set of macros to operate on the lock: NFSD_LOCK_ASSERT() Assert nfsd_mtx owned by current thread NFSD_UNLOCK_ASSERT() Assert nfsd_mtx not owned by current thread NFSD_LOCK_DONTCARE() Advisory: this function doesn't care NFSD_LOCK() Lock nfsd_mtx NFSD_UNLOCK() Unlock nfsd_mtx Constify a number of global variables/structures in the NFS server code, as they are not modified and contain constants only: nfsrvv2_procid nfsrv_nfsv3_procid nonidempotent nfsv2_repstat nfsv2_type nfsrv_nfsv3_procid nfsrvv2_procid nfsrv_v2errmap nfsv3err_null nfsv3err_getattr nfsv3err_setattr nfsv3err_lookup nfsv3err_access nfsv3err_readlink nfsv3err_read nfsv3err_write nfsv3err_create nfsv3err_mkdir nfsv3err_symlink nfsv3err_mknod nfsv3err_remove nfsv3err_rmdir nfsv3err_rename nfsv3err_link nfsv3err_readdir nfsv3err_readdirplus nfsv3err_fsstat nfsv3err_fsinfo nfsv3err_pathconf nfsv3err_commit nfsrv_v3errmap There are additional structures that should be constified but due to their being passed into general purpose functions without const arguments, I have not yet converted. In general, acquire nfsd_mtx when accessing any of the global NFS structures, including struct nfssvc_sock, struct nfsd, struct nfsrv_descript. Release nfsd_mtx whenever calling into VFS, and acquire Giant for calls into VFS. Giant is not required for any part of the operation of the NFS server with the exception of calls into VFS. Giant will never by acquired in the upcall code path. However, it may operate entirely covered by Giant, or not. If debug.mpsafenet is set to 0, the system calls will acquire Giant across all operations, and the upcall will assert Giant. As such, by default, this enables locking and allows us to test assertions, but should not cause any substantial new amount of code to be run without Giant. Bugs should manifest in the form of lock assertion failures for now. This approach is similar (but not identical) to modifications to the BSD/OS NFS server code snapshot provided by BSDi as part of their SMPng snapshot. The strategy is almost the same (single lock over the NFS server), but differs in the following ways: - Our NFS client and server code bases don't overlap, which means both fewer bugs and easier locking (thanks Peter!). Also means NFSD_*() as opposed to NFS_*(). - We make broad use of assertions, whereas the BSD/OS code does not. - Made slightly different choices about how to handle macros building packets but operating with side effects. - We acquire Giant only when entering VFS from the NFS server daemon threads. - Serious bugs in BSD/OS implementation corrected -- the snapshot we received was clearly a work in progress. Based on ideas from: BSDi SMPng Snapshot Reviewed by: rick@snowhite.cis.uoguelph.ca Extensive testing by: kris
2004-05-24 04:06:14 +00:00
if (*bp >= *be) {
if (*mp == mb)
(*mp)->m_len += *bp - bpos;
MGET(nmp, M_WAIT, MT_DATA);
MCLGET(nmp, M_WAIT);
nmp->m_len = NFSMSIZ(nmp);
(*mp)->m_next = nmp;
*mp = nmp;
*bp = mtod(*mp, caddr_t);
*be = *bp + (*mp)->m_len;
}
*tl = (u_int32_t *)*bp;
}
int
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
nfsm_srvmtofh_xx(fhandle_t *f, int v3, struct mbuf **md, caddr_t *dpos)
{
u_int32_t *tl;
int fhlen;
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
if (v3) {
tl = nfsm_dissect_xx_nonblock(NFSX_UNSIGNED, md, dpos);
if (tl == NULL)
return EBADRPC;
fhlen = fxdr_unsigned(int, *tl);
if (fhlen != 0 && fhlen != NFSX_V3FH)
return EBADRPC;
} else {
fhlen = NFSX_V2FH;
}
if (fhlen != 0) {
tl = nfsm_dissect_xx_nonblock(fhlen, md, dpos);
if (tl == NULL)
return EBADRPC;
bcopy((caddr_t)tl, (caddr_t)(f), fhlen);
} else {
bzero((caddr_t)(f), NFSX_V3FH);
}
return 0;
}
int
nfsm_srvsattr_xx(struct vattr *a, struct mbuf **md, caddr_t *dpos)
{
u_int32_t *tl;
int toclient = 0;
tl = nfsm_dissect_xx_nonblock(NFSX_UNSIGNED, md, dpos);
if (tl == NULL)
return EBADRPC;
if (*tl == nfsrv_nfs_true) {
tl = nfsm_dissect_xx_nonblock(NFSX_UNSIGNED, md, dpos);
if (tl == NULL)
return EBADRPC;
(a)->va_mode = nfstov_mode(*tl);
}
tl = nfsm_dissect_xx_nonblock(NFSX_UNSIGNED, md, dpos);
if (tl == NULL)
return EBADRPC;
if (*tl == nfsrv_nfs_true) {
tl = nfsm_dissect_xx_nonblock(NFSX_UNSIGNED, md, dpos);
if (tl == NULL)
return EBADRPC;
(a)->va_uid = fxdr_unsigned(uid_t, *tl);
}
tl = nfsm_dissect_xx_nonblock(NFSX_UNSIGNED, md, dpos);
if (tl == NULL)
return EBADRPC;
if (*tl == nfsrv_nfs_true) {
tl = nfsm_dissect_xx_nonblock(NFSX_UNSIGNED, md, dpos);
if (tl == NULL)
return EBADRPC;
(a)->va_gid = fxdr_unsigned(gid_t, *tl);
}
tl = nfsm_dissect_xx_nonblock(NFSX_UNSIGNED, md, dpos);
if (tl == NULL)
return EBADRPC;
if (*tl == nfsrv_nfs_true) {
tl = nfsm_dissect_xx_nonblock(2 * NFSX_UNSIGNED, md, dpos);
if (tl == NULL)
return EBADRPC;
(a)->va_size = fxdr_hyper(tl);
}
tl = nfsm_dissect_xx_nonblock(NFSX_UNSIGNED, md, dpos);
if (tl == NULL)
return EBADRPC;
switch (fxdr_unsigned(int, *tl)) {
case NFSV3SATTRTIME_TOCLIENT:
tl = nfsm_dissect_xx_nonblock(2 * NFSX_UNSIGNED, md, dpos);
if (tl == NULL)
return EBADRPC;
fxdr_nfsv3time(tl, &(a)->va_atime);
toclient = 1;
break;
case NFSV3SATTRTIME_TOSERVER:
getnanotime(&(a)->va_atime);
a->va_vaflags |= VA_UTIMES_NULL;
break;
}
tl = nfsm_dissect_xx_nonblock(NFSX_UNSIGNED, md, dpos);
if (tl == NULL)
return EBADRPC;
switch (fxdr_unsigned(int, *tl)) {
case NFSV3SATTRTIME_TOCLIENT:
tl = nfsm_dissect_xx_nonblock(2 * NFSX_UNSIGNED, md, dpos);
if (tl == NULL)
return EBADRPC;
fxdr_nfsv3time(tl, &(a)->va_mtime);
a->va_vaflags &= ~VA_UTIMES_NULL;
break;
case NFSV3SATTRTIME_TOSERVER:
getnanotime(&(a)->va_mtime);
if (toclient == 0)
a->va_vaflags |= VA_UTIMES_NULL;
break;
}
return 0;
}