freebsd-nq/sys/nfsserver/nfs_syscalls.c

738 lines
18 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_syscalls.c 8.5 (Berkeley) 3/30/95
1994-05-24 10:09:53 +00:00
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_inet6.h"
1994-05-24 10:09:53 +00:00
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/sysproto.h>
1994-05-24 10:09:53 +00:00
#include <sys/kernel.h>
#include <sys/sysctl.h>
1994-05-24 10:09:53 +00:00
#include <sys/file.h>
#include <sys/filedesc.h>
1994-05-24 10:09:53 +00:00
#include <sys/vnode.h>
#include <sys/malloc.h>
1994-05-24 10:09:53 +00:00
#include <sys/mount.h>
#include <sys/priv.h>
1994-05-24 10:09:53 +00:00
#include <sys/proc.h>
#include <sys/bio.h>
1994-05-24 10:09:53 +00:00
#include <sys/buf.h>
#include <sys/mbuf.h>
#include <sys/socket.h>
#include <sys/socketvar.h>
#include <sys/domain.h>
#include <sys/protosw.h>
#include <sys/namei.h>
#include <sys/fcntl.h>
#include <sys/lockf.h>
1994-05-24 10:09:53 +00:00
#include <netinet/in.h>
#include <netinet/tcp.h>
#ifdef INET6
#include <net/if.h>
#include <netinet6/in6_var.h>
#endif
#include <nfs/xdr_subs.h>
1994-05-24 10:09:53 +00:00
#include <nfs/rpcv2.h>
#include <nfs/nfsproto.h>
#include <nfsserver/nfs.h>
#include <nfsserver/nfsm_subs.h>
#include <nfsserver/nfsrvcache.h>
1994-05-24 10:09:53 +00:00
#include <security/audit/audit.h>
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
static MALLOC_DEFINE(M_NFSSVC, "nfss_srvsock", "Nfs server structure");
MALLOC_DEFINE(M_NFSRVDESC, "nfss_srvdesc", "NFS server socket descriptor");
MALLOC_DEFINE(M_NFSD, "nfss_daemon", "Nfs server daemon structure");
1994-05-24 10:09:53 +00:00
#define TRUE 1
#define FALSE 0
SYSCTL_DECL(_vfs_nfsrv);
int nfsd_waiting = 0;
int nfsrv_numnfsd = 0;
static int notstarted = 1;
static int nfs_privport = 0;
SYSCTL_INT(_vfs_nfsrv, NFS_NFSPRIVPORT, nfs_privport, CTLFLAG_RW,
&nfs_privport, 0,
"Only allow clients using a privileged port");
SYSCTL_INT(_vfs_nfsrv, OID_AUTO, gatherdelay, CTLFLAG_RW,
&nfsrvw_procrastinate, 0,
"Delay value for write gathering");
SYSCTL_INT(_vfs_nfsrv, OID_AUTO, gatherdelay_v3, CTLFLAG_RW,
&nfsrvw_procrastinate_v3, 0,
"Delay in seconds for NFSv3 write gathering");
static int nfssvc_addsock(struct file *, struct sockaddr *);
static void nfsrv_zapsock(struct nfssvc_sock *slp);
static int nfssvc_nfsd(void);
extern u_long sb_max_adj;
1994-05-24 10:09:53 +00:00
/*
* NFS server system calls
*/
/*
* Nfs server psuedo system call for the nfsd's
* Based on the flag value it either:
* - adds a socket to the selection list
* - remains in the kernel as an nfsd
* - remains in the kernel as an nfsiod
* For INET6 we suppose that nfsd provides only IN6P_IPV6_V6ONLY sockets
* and that mountd provides
* - sockaddr with no IPv4-mapped addresses
* - mask for both INET and INET6 families if there is IPv4-mapped overlap
1994-05-24 10:09:53 +00:00
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct nfssvc_args {
int flag;
caddr_t argp;
};
#endif
int
nfssvc(struct thread *td, struct nfssvc_args *uap)
1994-05-24 10:09:53 +00:00
{
struct file *fp;
struct sockaddr *nam;
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 nfsd_addsock_args nfsdarg;
1994-05-24 10:09:53 +00:00
int error;
KASSERT(!mtx_owned(&Giant), ("nfssvc(): called with Giant"));
AUDIT_ARG(cmd, uap->flag);
error = priv_check(td, PRIV_NFS_DAEMON);
if (error)
return (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
NFSD_LOCK();
while (nfssvc_sockhead_flag & SLP_INIT) {
nfssvc_sockhead_flag |= SLP_WANTINIT;
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
(void) msleep(&nfssvc_sockhead, &nfsd_mtx, PSOCK,
"nfsd init", 0);
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_UNLOCK();
if (uap->flag & NFSSVC_ADDSOCK) {
error = copyin(uap->argp, (caddr_t)&nfsdarg, sizeof(nfsdarg));
if (error)
return (error);
if ((error = fget(td, nfsdarg.sock, &fp)) != 0)
return (error);
if (fp->f_type != DTYPE_SOCKET) {
fdrop(fp, td);
return (error); /* XXXRW: Should be EINVAL? */
}
1994-05-24 10:09:53 +00:00
/*
* Get the client address for connected sockets.
*/
if (nfsdarg.name == NULL || nfsdarg.namelen == 0)
nam = NULL;
else {
error = getsockaddr(&nam, nfsdarg.name,
nfsdarg.namelen);
if (error) {
fdrop(fp, td);
return (error);
}
}
error = nfssvc_addsock(fp, nam);
fdrop(fp, td);
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 if (uap->flag & NFSSVC_OLDNFSD) {
error = nfssvc_nfsd();
} else {
error = ENXIO;
1994-05-24 10:09:53 +00:00
}
if (error == EINTR || error == ERESTART)
error = 0;
return (error);
}
/*
* Adds a socket to the list for servicing by nfsds.
*/
1995-12-17 21:14:36 +00:00
static int
nfssvc_addsock(struct file *fp, struct sockaddr *mynam)
1994-05-24 10:09:53 +00:00
{
int siz;
struct nfssvc_sock *slp;
struct socket *so;
int error;
1994-05-24 10:09:53 +00:00
so = fp->f_data;
#if 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
/*
* XXXRW: If this code is ever enabled, there's a race when running
* MPSAFE.
*/
tslp = NULL;
1994-05-24 10:09:53 +00:00
/*
* Add it to the list, as required.
*/
if (so->so_proto->pr_protocol == IPPROTO_UDP) {
tslp = nfs_udpsock;
if (tslp->ns_flag & SLP_VALID) {
if (mynam != NULL)
free(mynam, M_SONAME);
1994-05-24 10:09:53 +00:00
return (EPERM);
}
}
#endif
siz = sb_max_adj;
1995-05-30 08:16:23 +00:00
error = soreserve(so, siz, siz);
if (error) {
if (mynam != NULL)
free(mynam, M_SONAME);
1994-05-24 10:09:53 +00:00
return (error);
}
/*
* Set protocol specific options { for now TCP only } and
* reserve some space. For datagram sockets, this can get called
* repeatedly for the same socket, but that isn't harmful.
*/
if (so->so_type == SOCK_STREAM) {
struct sockopt sopt;
int val;
bzero(&sopt, sizeof sopt);
sopt.sopt_dir = SOPT_SET;
sopt.sopt_level = SOL_SOCKET;
sopt.sopt_name = SO_KEEPALIVE;
sopt.sopt_val = &val;
sopt.sopt_valsize = sizeof val;
val = 1;
sosetopt(so, &sopt);
1994-05-24 10:09:53 +00:00
}
if (so->so_proto->pr_protocol == IPPROTO_TCP) {
struct sockopt sopt;
int val;
bzero(&sopt, sizeof sopt);
sopt.sopt_dir = SOPT_SET;
sopt.sopt_level = IPPROTO_TCP;
sopt.sopt_name = TCP_NODELAY;
sopt.sopt_val = &val;
sopt.sopt_valsize = sizeof val;
val = 1;
sosetopt(so, &sopt);
1994-05-24 10:09:53 +00:00
}
SOCKBUF_LOCK(&so->so_rcv);
1994-05-24 10:09:53 +00:00
so->so_rcv.sb_flags &= ~SB_NOINTR;
so->so_rcv.sb_timeo = 0;
SOCKBUF_UNLOCK(&so->so_rcv);
SOCKBUF_LOCK(&so->so_snd);
1994-05-24 10:09:53 +00:00
so->so_snd.sb_flags &= ~SB_NOINTR;
so->so_snd.sb_timeo = 0;
SOCKBUF_UNLOCK(&so->so_snd);
slp = (struct nfssvc_sock *)
malloc(sizeof (struct nfssvc_sock), M_NFSSVC,
M_WAITOK | M_ZERO);
STAILQ_INIT(&slp->ns_rec);
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();
TAILQ_INSERT_TAIL(&nfssvc_sockhead, slp, ns_chain);
1994-05-24 10:09:53 +00:00
slp->ns_so = so;
slp->ns_nam = mynam;
fhold(fp);
1994-05-24 10:09:53 +00:00
slp->ns_fp = fp;
SOCKBUF_LOCK(&so->so_rcv);
1994-05-24 10:09:53 +00:00
so->so_upcallarg = (caddr_t)slp;
so->so_upcall = nfsrv_rcv;
so->so_rcv.sb_flags |= SB_UPCALL;
SOCKBUF_UNLOCK(&so->so_rcv);
1994-05-24 10:09:53 +00:00
slp->ns_flag = (SLP_VALID | SLP_NEEDQ);
nfsrv_wakenfsd(slp);
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();
1994-05-24 10:09:53 +00:00
return (0);
}
/*
* Called by nfssvc() for nfsds. Just loops around servicing rpc requests
* until it is killed by a signal.
*/
1995-12-17 21:14:36 +00:00
static int
nfssvc_nfsd()
1994-05-24 10:09:53 +00:00
{
int siz;
struct nfssvc_sock *slp;
struct nfsd *nfsd;
struct nfsrv_descript *nd = NULL;
struct mbuf *m, *mreq;
int error = 0, cacherep, s, sotype, writes_todo;
int procrastinate;
u_quad_t cur_usec;
1994-05-24 10:09:53 +00:00
#ifndef nolint
cacherep = RC_DOIT;
writes_todo = 0;
#endif
nfsd = (struct nfsd *)
malloc(sizeof (struct nfsd), M_NFSD, M_WAITOK | M_ZERO);
s = splnet();
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();
TAILQ_INSERT_TAIL(&nfsd_head, nfsd, nfsd_chain);
nfsrv_numnfsd++;
1994-05-24 10:09:53 +00:00
/*
* Loop getting rpc requests until SIGKILL.
*/
for (;;) {
if ((nfsd->nfsd_flag & NFSD_REQINPROG) == 0) {
while (nfsd->nfsd_slp == NULL &&
(nfsd_head_flag & NFSD_CHECKSLP) == 0) {
nfsd->nfsd_flag |= NFSD_WAITING;
1994-05-24 10:09:53 +00:00
nfsd_waiting++;
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 = msleep(nfsd, &nfsd_mtx,
PSOCK | PCATCH, "-", 0);
1994-05-24 10:09:53 +00:00
nfsd_waiting--;
if (error)
goto done;
}
if (nfsd->nfsd_slp == NULL &&
(nfsd_head_flag & NFSD_CHECKSLP) != 0) {
TAILQ_FOREACH(slp, &nfssvc_sockhead, ns_chain) {
1994-05-24 10:09:53 +00:00
if ((slp->ns_flag & (SLP_VALID | SLP_DOREC))
== (SLP_VALID | SLP_DOREC)) {
slp->ns_flag &= ~SLP_DOREC;
slp->ns_sref++;
nfsd->nfsd_slp = slp;
1994-05-24 10:09:53 +00:00
break;
}
}
if (slp == NULL)
nfsd_head_flag &= ~NFSD_CHECKSLP;
1994-05-24 10:09:53 +00:00
}
if ((slp = nfsd->nfsd_slp) == NULL)
1994-05-24 10:09:53 +00:00
continue;
if (slp->ns_flag & SLP_VALID) {
if (slp->ns_flag & SLP_DISCONN)
nfsrv_zapsock(slp);
else if (slp->ns_flag & SLP_NEEDQ) {
slp->ns_flag &= ~SLP_NEEDQ;
(void) nfs_slplock(slp, 1);
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();
1994-05-24 10:09:53 +00:00
nfsrv_rcv(slp->ns_so, (caddr_t)slp,
M_WAIT);
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();
nfs_slpunlock(slp);
1994-05-24 10:09:53 +00:00
}
error = nfsrv_dorec(slp, nfsd, &nd);
cur_usec = nfs_curusec();
if (error && LIST_FIRST(&slp->ns_tq) &&
LIST_FIRST(&slp->ns_tq)->nd_time <= cur_usec) {
error = 0;
cacherep = RC_DOIT;
writes_todo = 1;
} else
writes_todo = 0;
nfsd->nfsd_flag |= NFSD_REQINPROG;
1994-05-24 10:09:53 +00:00
}
} else {
error = 0;
slp = nfsd->nfsd_slp;
1994-05-24 10:09:53 +00:00
}
if (error || (slp->ns_flag & SLP_VALID) == 0) {
if (nd) {
if (nd->nd_cr != NULL)
crfree(nd->nd_cr);
free((caddr_t)nd, M_NFSRVDESC);
nd = NULL;
}
nfsd->nfsd_slp = NULL;
nfsd->nfsd_flag &= ~NFSD_REQINPROG;
1994-05-24 10:09:53 +00:00
nfsrv_slpderef(slp);
continue;
}
splx(s);
sotype = slp->ns_so->so_type;
if (nd) {
getmicrotime(&nd->nd_starttime);
if (nd->nd_nam2)
nd->nd_nam = nd->nd_nam2;
else
nd->nd_nam = slp->ns_nam;
1994-05-24 10:09:53 +00:00
/*
* Check to see if authorization is needed.
*/
cacherep = nfsrv_getcache(nd, &mreq);
1994-05-24 10:09:53 +00:00
if (nfs_privport) {
/* Check if source port is privileged */
u_short port;
struct sockaddr *nam = nd->nd_nam;
struct sockaddr_in *sin;
sin = (struct sockaddr_in *)nam;
/*
* INET/INET6 - same code:
* sin_port and sin6_port are at same offset
*/
port = ntohs(sin->sin_port);
if (port >= IPPORT_RESERVED &&
nd->nd_procnum != NFSPROC_NULL) {
#ifdef INET6
char b6[INET6_ADDRSTRLEN];
#if defined(KLD_MODULE)
/* Do not use ip6_sprintf: the nfs module should work without INET6. */
#define ip6_sprintf(buf, a) \
(sprintf((buf), "%x:%x:%x:%x:%x:%x:%x:%x", \
(a)->s6_addr16[0], (a)->s6_addr16[1], \
(a)->s6_addr16[2], (a)->s6_addr16[3], \
(a)->s6_addr16[4], (a)->s6_addr16[5], \
(a)->s6_addr16[6], (a)->s6_addr16[7]), \
(buf))
#endif
#endif
nd->nd_procnum = NFSPROC_NOOP;
nd->nd_repstat = (NFSERR_AUTHERR | AUTH_TOOWEAK);
cacherep = RC_DOIT;
printf("NFS request from unprivileged port (%s:%d)\n",
#ifdef INET6
sin->sin_family == AF_INET6 ?
ip6_sprintf(b6, &satosin6(sin)->sin6_addr) :
#if defined(KLD_MODULE)
#undef ip6_sprintf
#endif
#endif
inet_ntoa(sin->sin_addr), port);
}
}
1994-05-24 10:09:53 +00:00
}
/*
* Loop to get all the write rpc relies that have been
* gathered together.
*/
do {
switch (cacherep) {
case RC_DOIT:
if (nd && (nd->nd_flag & ND_NFSV3))
procrastinate = nfsrvw_procrastinate_v3;
else
procrastinate = nfsrvw_procrastinate;
NFSD_UNLOCK();
if (writes_todo || (!(nd->nd_flag & ND_NFSV3) &&
nd->nd_procnum == NFSPROC_WRITE &&
procrastinate > 0 && !notstarted))
error = nfsrv_writegather(&nd, slp, &mreq);
else
error = (*(nfsrv3_procs[nd->nd_procnum]))(nd,
slp, &mreq);
NFSD_LOCK();
if (mreq == NULL)
break;
if (error != 0 && error != NFSERR_RETVOID) {
nfsrvstats.srv_errs++;
nfsrv_updatecache(nd, FALSE, mreq);
if (nd->nd_nam2)
free(nd->nd_nam2, M_SONAME);
1994-05-24 10:09:53 +00:00
break;
}
nfsrvstats.srvrpccnt[nd->nd_procnum]++;
nfsrv_updatecache(nd, TRUE, mreq);
nd->nd_mrep = NULL;
/* FALLTHROUGH */
case RC_REPLY:
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();
siz = m_length(mreq, NULL);
1994-05-24 10:09:53 +00:00
if (siz <= 0 || siz > NFS_MAXPACKET) {
printf("mbuf siz=%d\n",siz);
panic("Bad nfs svc reply");
}
m = mreq;
m->m_pkthdr.len = siz;
m->m_pkthdr.rcvif = NULL;
1994-05-24 10:09:53 +00:00
/*
* For stream protocols, prepend a Sun RPC
* Record Mark.
*/
if (sotype == SOCK_STREAM) {
M_PREPEND(m, NFSX_UNSIGNED, M_WAIT);
*mtod(m, u_int32_t *) = htonl(0x80000000 | siz);
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();
if (slp->ns_so->so_proto->pr_flags & PR_CONNREQUIRED)
(void) nfs_slplock(slp, 1);
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 (slp->ns_flag & SLP_VALID) {
NFSD_UNLOCK();
error = nfsrv_send(slp->ns_so, nd->nd_nam2, m);
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();
} else {
1994-05-24 10:09:53 +00:00
error = EPIPE;
m_freem(m);
}
if (nd->nd_nam2)
free(nd->nd_nam2, M_SONAME);
1994-05-24 10:09:53 +00:00
if (nd->nd_mrep)
m_freem(nd->nd_mrep);
if (error == EPIPE)
nfsrv_zapsock(slp);
if (slp->ns_so->so_proto->pr_flags & PR_CONNREQUIRED)
nfs_slpunlock(slp);
1994-05-24 10:09:53 +00:00
if (error == EINTR || error == ERESTART) {
if (nd->nd_cr != NULL)
crfree(nd->nd_cr);
free((caddr_t)nd, M_NFSRVDESC);
1994-05-24 10:09:53 +00:00
nfsrv_slpderef(slp);
s = splnet();
goto done;
}
break;
case RC_DROPIT:
1994-05-24 10:09:53 +00:00
m_freem(nd->nd_mrep);
if (nd->nd_nam2)
free(nd->nd_nam2, M_SONAME);
1994-05-24 10:09:53 +00:00
break;
};
if (nd) {
if (nd->nd_cr != NULL)
crfree(nd->nd_cr);
free((caddr_t)nd, M_NFSRVDESC);
nd = NULL;
}
/*
* Check to see if there are outstanding writes that
* need to be serviced.
*/
cur_usec = nfs_curusec();
s = splsoftclock();
if (LIST_FIRST(&slp->ns_tq) &&
LIST_FIRST(&slp->ns_tq)->nd_time <= cur_usec) {
cacherep = RC_DOIT;
writes_todo = 1;
} else
writes_todo = 0;
splx(s);
} while (writes_todo);
1994-05-24 10:09:53 +00:00
s = splnet();
if (nfsrv_dorec(slp, nfsd, &nd)) {
nfsd->nfsd_flag &= ~NFSD_REQINPROG;
nfsd->nfsd_slp = NULL;
1994-05-24 10:09:53 +00:00
nfsrv_slpderef(slp);
}
mtx_assert(&Giant, MA_NOTOWNED);
1994-05-24 10:09:53 +00:00
}
done:
mtx_assert(&Giant, MA_NOTOWNED);
TAILQ_REMOVE(&nfsd_head, nfsd, nfsd_chain);
1994-05-24 10:09:53 +00:00
splx(s);
free((caddr_t)nfsd, M_NFSD);
if (--nfsrv_numnfsd == 0)
1994-05-24 10:09:53 +00:00
nfsrv_init(TRUE); /* Reinitialize everything */
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();
1994-05-24 10:09:53 +00:00
return (error);
}
/*
* Shut down a socket associated with an nfssvc_sock structure.
* Should be called with the send lock set, if required.
* The trick here is to increment the sref at the start, so that the nfsds
* will stop using it and clear ns_flag at the end so that it will not be
* reassigned during cleanup.
*/
static void
nfsrv_zapsock(struct nfssvc_sock *slp)
{
struct nfsrv_descript *nwp, *nnwp;
struct socket *so;
struct file *fp;
struct nfsrv_rec *rec;
int s;
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_ASSERT();
/*
* XXXRW: By clearing all flags, other threads/etc should ignore
* this slp and we can safely release nfsd_mtx so we can clean
* up the slp safely.
*/
slp->ns_flag &= ~SLP_ALLFLAGS;
fp = slp->ns_fp;
if (fp) {
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();
slp->ns_fp = NULL;
so = slp->ns_so;
SOCKBUF_LOCK(&so->so_rcv);
so->so_rcv.sb_flags &= ~SB_UPCALL;
so->so_upcall = NULL;
so->so_upcallarg = NULL;
SOCKBUF_UNLOCK(&so->so_rcv);
soshutdown(so, SHUT_RDWR);
closef(fp, 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
NFSD_LOCK();
if (slp->ns_nam)
free(slp->ns_nam, M_SONAME);
m_freem(slp->ns_raw);
while ((rec = STAILQ_FIRST(&slp->ns_rec)) != NULL) {
STAILQ_REMOVE_HEAD(&slp->ns_rec, nr_link);
if (rec->nr_address)
free(rec->nr_address, M_SONAME);
m_freem(rec->nr_packet);
free(rec, M_NFSRVDESC);
}
s = splsoftclock();
for (nwp = LIST_FIRST(&slp->ns_tq); nwp; nwp = nnwp) {
nnwp = LIST_NEXT(nwp, nd_tq);
LIST_REMOVE(nwp, nd_tq);
if (nwp->nd_cr != NULL)
crfree(nwp->nd_cr);
free((caddr_t)nwp, M_NFSRVDESC);
}
LIST_INIT(&slp->ns_tq);
splx(s);
}
}
/*
* Derefence a server socket structure. If it has no more references and
* is no longer valid, you can throw it away.
*/
void
nfsrv_slpderef(struct nfssvc_sock *slp)
{
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_ASSERT();
if (--(slp->ns_sref) == 0 && (slp->ns_flag & SLP_VALID) == 0) {
TAILQ_REMOVE(&nfssvc_sockhead, slp, ns_chain);
free((caddr_t)slp, M_NFSSVC);
}
}
/*
* Lock a socket against others.
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
*
* XXXRW: Wait argument is always 1 in the caller. Replace with a real
* sleep lock?
*/
int
nfs_slplock(struct nfssvc_sock *slp, int wait)
{
int *statep = &slp->ns_solock;
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_ASSERT();
if (!wait && (*statep & NFSRV_SNDLOCK))
return(0); /* already locked, fail */
while (*statep & NFSRV_SNDLOCK) {
*statep |= NFSRV_WANTSND;
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
(void) msleep(statep, &nfsd_mtx, PZERO - 1, "nfsslplck", 0);
}
*statep |= NFSRV_SNDLOCK;
return (1);
}
/*
* Unlock the stream socket for others.
*/
void
nfs_slpunlock(struct nfssvc_sock *slp)
{
int *statep = &slp->ns_solock;
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_ASSERT();
if ((*statep & NFSRV_SNDLOCK) == 0)
panic("nfs slpunlock");
*statep &= ~NFSRV_SNDLOCK;
if (*statep & NFSRV_WANTSND) {
*statep &= ~NFSRV_WANTSND;
wakeup(statep);
}
}
/*
* Initialize the data structures for the server.
* Handshake with any new nfsds starting up to avoid any chance of
* corruption.
*/
void
nfsrv_init(int terminating)
{
struct nfssvc_sock *slp, *nslp;
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_ASSERT();
if (nfssvc_sockhead_flag & SLP_INIT)
panic("nfsd init");
nfssvc_sockhead_flag |= SLP_INIT;
if (terminating) {
TAILQ_FOREACH_SAFE(slp, &nfssvc_sockhead, ns_chain, nslp) {
if (slp->ns_flag & SLP_VALID)
nfsrv_zapsock(slp);
TAILQ_REMOVE(&nfssvc_sockhead, slp, ns_chain);
free((caddr_t)slp, M_NFSSVC);
}
nfsrv_cleancache(); /* And clear out server cache */
} else
nfs_pub.np_valid = 0;
TAILQ_INIT(&nfssvc_sockhead);
nfssvc_sockhead_flag &= ~SLP_INIT;
if (nfssvc_sockhead_flag & SLP_WANTINIT) {
nfssvc_sockhead_flag &= ~SLP_WANTINIT;
wakeup(&nfssvc_sockhead);
}
TAILQ_INIT(&nfsd_head);
nfsd_head_flag &= ~NFSD_CHECKSLP;
#if 0
nfs_udpsock = (struct nfssvc_sock *)
malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK | M_ZERO);
STAILQ_INIT(&nfs_udpsock->ns_rec);
TAILQ_INSERT_HEAD(&nfssvc_sockhead, nfs_udpsock, ns_chain);
nfs_cltpsock = (struct nfssvc_sock *)
malloc(sizeof (struct nfssvc_sock), M_NFSSVC, M_WAITOK | M_ZERO);
STAILQ_INIT(&nfs_cltpsock->ns_rec);
TAILQ_INSERT_TAIL(&nfssvc_sockhead, nfs_cltpsock, ns_chain);
#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
#endif /* NFS_LEGACYRPC */