freebsd-nq/sys/nfsclient/nfs_subs.c

1207 lines
30 KiB
C
Raw Normal View History

/*-
1994-05-24 10:09:53 +00:00
* Copyright (c) 1989, 1993
* The Regents of the University of California. All rights reserved.
*
* This code is derived from software contributed to Berkeley by
* Rick Macklem at The University of Guelph.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 4. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* @(#)nfs_subs.c 8.8 (Berkeley) 5/22/95
1994-05-24 10:09:53 +00:00
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
1994-05-24 10:09:53 +00:00
/*
* These functions support the macros and help fiddle mbuf chains for
* the nfs op functions. They do things like create the rpc header and
* copy data between mbuf chains and uio lists.
*/
1994-05-24 10:09:53 +00:00
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/bio.h>
#include <sys/buf.h>
#include <sys/proc.h>
1994-05-24 10:09:53 +00:00
#include <sys/mount.h>
#include <sys/vnode.h>
#include <sys/namei.h>
#include <sys/mbuf.h>
#include <sys/socket.h>
#include <sys/stat.h>
#include <sys/malloc.h>
1994-09-22 22:10:49 +00:00
#include <sys/sysent.h>
#include <sys/syscall.h>
#include <sys/sysproto.h>
1994-05-24 10:09:53 +00:00
#include <vm/vm.h>
#include <vm/vm_object.h>
#include <vm/vm_extern.h>
#include <vm/uma.h>
#include <rpc/rpcclnt.h>
1994-05-24 10:09:53 +00:00
#include <nfs/rpcv2.h>
#include <nfs/nfsproto.h>
#include <nfsclient/nfs.h>
#include <nfsclient/nfsnode.h>
1994-05-24 10:09:53 +00:00
#include <nfs/xdr_subs.h>
#include <nfsclient/nfsm_subs.h>
#include <nfsclient/nfsmount.h>
1994-05-24 10:09:53 +00:00
#include <netinet/in.h>
/*
* Note that stdarg.h and the ANSI style va_start macro is used for both
* ANSI and traditional C compilers.
*/
#include <machine/stdarg.h>
1994-05-24 10:09:53 +00:00
/*
* Data items converted to xdr at startup, since they are constant
* This is kinda hokey, but may save a little time doing byte swaps
*/
u_int32_t nfs_xdrneg1;
u_int32_t rpc_call, rpc_vers, rpc_reply, rpc_msgdenied, rpc_autherr,
rpc_mismatch, rpc_auth_unix, rpc_msgaccepted;
u_int32_t nfs_true, nfs_false;
1994-05-24 10:09:53 +00:00
/* And other global data */
static u_int32_t nfs_xid = 0;
1995-12-17 21:14:36 +00:00
static enum vtype nv2tov_type[8]= {
VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON, VNON
1995-12-17 21:14:36 +00:00
};
int nfs_ticks;
int nfs_pbuf_freecnt = -1; /* start out unlimited */
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
#ifdef NFS_LEGACYRPC
struct nfs_reqq nfs_reqq;
struct mtx nfs_reqq_mtx;
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
struct nfs_bufq nfs_bufq;
static struct mtx nfs_xid_mtx;
/*
* and the reverse mapping from generic to Version 2 procedure numbers
*/
int nfsv2_procid[NFS_NPROCS] = {
NFSV2PROC_NULL,
NFSV2PROC_GETATTR,
NFSV2PROC_SETATTR,
NFSV2PROC_LOOKUP,
NFSV2PROC_NOOP,
NFSV2PROC_READLINK,
NFSV2PROC_READ,
NFSV2PROC_WRITE,
NFSV2PROC_CREATE,
NFSV2PROC_MKDIR,
NFSV2PROC_SYMLINK,
NFSV2PROC_CREATE,
NFSV2PROC_REMOVE,
NFSV2PROC_RMDIR,
NFSV2PROC_RENAME,
NFSV2PROC_LINK,
NFSV2PROC_READDIR,
NFSV2PROC_NOOP,
NFSV2PROC_STATFS,
NFSV2PROC_NOOP,
NFSV2PROC_NOOP,
NFSV2PROC_NOOP,
NFSV2PROC_NOOP,
};
LIST_HEAD(nfsnodehashhead, nfsnode);
u_int32_t
nfs_xid_gen(void)
{
uint32_t xid;
mtx_lock(&nfs_xid_mtx);
/* Get a pretty random xid to start with */
if (!nfs_xid)
nfs_xid = random();
/*
* Skip zero xid if it should ever happen.
*/
if (++nfs_xid == 0)
nfs_xid++;
xid = nfs_xid;
mtx_unlock(&nfs_xid_mtx);
return xid;
}
1994-05-24 10:09:53 +00:00
/*
* Create the header for an rpc request packet
* The hsiz is the size of the rest of the nfs request header.
* (just used to decide if a cluster is a good idea)
*/
struct mbuf *
nfsm_reqhead(struct vnode *vp, u_long procid, int hsiz)
1994-05-24 10:09:53 +00:00
{
struct mbuf *mb;
1994-05-24 10:09:53 +00:00
MGET(mb, M_WAIT, MT_DATA);
1994-05-24 10:09:53 +00:00
if (hsiz >= MINCLSIZE)
MCLGET(mb, M_WAIT);
1994-05-24 10:09:53 +00:00
mb->m_len = 0;
return (mb);
}
/*
* Build the RPC header and fill in the authorization info.
* The authorization string argument is only used when the credentials
* come from outside of the kernel.
* Returns the head of the mbuf list.
*/
struct mbuf *
nfsm_rpchead(struct ucred *cr, int nmflag, int procid, int auth_type,
int auth_len, struct mbuf *mrest, int mrest_len, struct mbuf **mbp,
u_int32_t **xidpp)
1994-05-24 10:09:53 +00:00
{
struct mbuf *mb;
u_int32_t *tl;
caddr_t bpos;
int i;
struct mbuf *mreq;
int grpsiz, authsiz;
1994-05-24 10:09:53 +00:00
authsiz = nfsm_rndup(auth_len);
MGETHDR(mb, M_WAIT, MT_DATA);
if ((authsiz + 10 * NFSX_UNSIGNED) >= MINCLSIZE) {
MCLGET(mb, M_WAIT);
} else if ((authsiz + 10 * NFSX_UNSIGNED) < MHLEN) {
MH_ALIGN(mb, authsiz + 10 * NFSX_UNSIGNED);
1994-05-24 10:09:53 +00:00
} else {
MH_ALIGN(mb, 8 * NFSX_UNSIGNED);
1994-05-24 10:09:53 +00:00
}
mb->m_len = 0;
mreq = mb;
bpos = mtod(mb, caddr_t);
/*
* First the RPC header.
*/
tl = nfsm_build(u_int32_t *, 8 * NFSX_UNSIGNED);
*xidpp = tl;
*tl++ = txdr_unsigned(nfs_xid_gen());
1994-05-24 10:09:53 +00:00
*tl++ = rpc_call;
*tl++ = rpc_vers;
*tl++ = txdr_unsigned(NFS_PROG);
if (nmflag & NFSMNT_NFSV3) {
*tl++ = txdr_unsigned(NFS_VER3);
*tl++ = txdr_unsigned(procid);
} else {
*tl++ = txdr_unsigned(NFS_VER2);
*tl++ = txdr_unsigned(nfsv2_procid[procid]);
}
1994-05-24 10:09:53 +00:00
/*
* And then the authorization cred.
*/
*tl++ = txdr_unsigned(auth_type);
*tl = txdr_unsigned(authsiz);
switch (auth_type) {
case RPCAUTH_UNIX:
tl = nfsm_build(u_int32_t *, auth_len);
1994-05-24 10:09:53 +00:00
*tl++ = 0; /* stamp ?? */
*tl++ = 0; /* NULL hostname */
*tl++ = txdr_unsigned(cr->cr_uid);
*tl++ = txdr_unsigned(cr->cr_groups[0]);
grpsiz = (auth_len >> 2) - 5;
*tl++ = txdr_unsigned(grpsiz);
for (i = 1; i <= grpsiz; i++)
*tl++ = txdr_unsigned(cr->cr_groups[i]);
break;
}
/*
* And the verifier...
*/
tl = nfsm_build(u_int32_t *, 2 * NFSX_UNSIGNED);
*tl++ = txdr_unsigned(RPCAUTH_NULL);
*tl = 0;
1994-05-24 10:09:53 +00:00
mb->m_next = mrest;
mreq->m_pkthdr.len = authsiz + 10 * NFSX_UNSIGNED + mrest_len;
mreq->m_pkthdr.rcvif = NULL;
1994-05-24 10:09:53 +00:00
*mbp = mb;
return (mreq);
}
/*
* copies a uio scatter/gather list to an mbuf chain.
* NOTE: can ony handle iovcnt == 1
1994-05-24 10:09:53 +00:00
*/
int
nfsm_uiotombuf(struct uio *uiop, struct mbuf **mq, int siz, caddr_t *bpos)
1994-05-24 10:09:53 +00:00
{
char *uiocp;
struct mbuf *mp, *mp2;
int xfer, left, mlen;
1994-05-24 10:09:53 +00:00
int uiosiz, clflg, rem;
char *cp;
#ifdef DIAGNOSTIC
if (uiop->uio_iovcnt != 1)
panic("nfsm_uiotombuf: iovcnt != 1");
#endif
1994-05-24 10:09:53 +00:00
if (siz > MLEN) /* or should it >= MCLBYTES ?? */
clflg = 1;
else
clflg = 0;
rem = nfsm_rndup(siz)-siz;
mp = mp2 = *mq;
while (siz > 0) {
left = uiop->uio_iov->iov_len;
uiocp = uiop->uio_iov->iov_base;
if (left > siz)
left = siz;
uiosiz = left;
while (left > 0) {
mlen = M_TRAILINGSPACE(mp);
if (mlen == 0) {
MGET(mp, M_WAIT, MT_DATA);
1994-05-24 10:09:53 +00:00
if (clflg)
MCLGET(mp, M_WAIT);
1994-05-24 10:09:53 +00:00
mp->m_len = 0;
mp2->m_next = mp;
mp2 = mp;
mlen = M_TRAILINGSPACE(mp);
}
xfer = (left > mlen) ? mlen : left;
#ifdef notdef
/* Not Yet.. */
if (uiop->uio_iov->iov_op != NULL)
(*(uiop->uio_iov->iov_op))
(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
else
#endif
if (uiop->uio_segflg == UIO_SYSSPACE)
bcopy(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
else
copyin(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
mp->m_len += xfer;
left -= xfer;
uiocp += xfer;
uiop->uio_offset += xfer;
uiop->uio_resid -= xfer;
}
uiop->uio_iov->iov_base =
(char *)uiop->uio_iov->iov_base + uiosiz;
uiop->uio_iov->iov_len -= uiosiz;
1994-05-24 10:09:53 +00:00
siz -= uiosiz;
}
if (rem > 0) {
if (rem > M_TRAILINGSPACE(mp)) {
MGET(mp, M_WAIT, MT_DATA);
1994-05-24 10:09:53 +00:00
mp->m_len = 0;
mp2->m_next = mp;
}
cp = mtod(mp, caddr_t)+mp->m_len;
for (left = 0; left < rem; left++)
*cp++ = '\0';
mp->m_len += rem;
*bpos = cp;
} else
*bpos = mtod(mp, caddr_t)+mp->m_len;
*mq = mp;
return (0);
}
/*
* Copy a string into mbufs for the hard cases...
*/
int
nfsm_strtmbuf(struct mbuf **mb, char **bpos, const char *cp, long siz)
1994-05-24 10:09:53 +00:00
{
struct mbuf *m1 = NULL, *m2;
1994-05-24 10:09:53 +00:00
long left, xfer, len, tlen;
u_int32_t *tl;
1994-05-24 10:09:53 +00:00
int putsize;
putsize = 1;
m2 = *mb;
left = M_TRAILINGSPACE(m2);
if (left > 0) {
tl = ((u_int32_t *)(*bpos));
1994-05-24 10:09:53 +00:00
*tl++ = txdr_unsigned(siz);
putsize = 0;
left -= NFSX_UNSIGNED;
m2->m_len += NFSX_UNSIGNED;
if (left > 0) {
bcopy(cp, (caddr_t) tl, left);
siz -= left;
cp += left;
m2->m_len += left;
left = 0;
}
}
/* Loop around adding mbufs */
while (siz > 0) {
MGET(m1, M_WAIT, MT_DATA);
1994-05-24 10:09:53 +00:00
if (siz > MLEN)
MCLGET(m1, M_WAIT);
1994-05-24 10:09:53 +00:00
m1->m_len = NFSMSIZ(m1);
m2->m_next = m1;
m2 = m1;
tl = mtod(m1, u_int32_t *);
1994-05-24 10:09:53 +00:00
tlen = 0;
if (putsize) {
*tl++ = txdr_unsigned(siz);
m1->m_len -= NFSX_UNSIGNED;
tlen = NFSX_UNSIGNED;
putsize = 0;
}
if (siz < m1->m_len) {
len = nfsm_rndup(siz);
xfer = siz;
if (xfer < len)
*(tl+(xfer>>2)) = 0;
} else {
xfer = len = m1->m_len;
}
bcopy(cp, (caddr_t) tl, xfer);
m1->m_len = len+tlen;
siz -= xfer;
cp += xfer;
}
*mb = m1;
*bpos = mtod(m1, caddr_t)+m1->m_len;
return (0);
}
/*
* Called once to initialize data structures...
*/
int
nfs_init(struct vfsconf *vfsp)
1994-05-24 10:09:53 +00:00
{
int i;
1994-05-24 10:09:53 +00:00
nfsmount_zone = uma_zcreate("NFSMOUNT", sizeof(struct nfsmount),
NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
1994-05-24 10:09:53 +00:00
rpc_vers = txdr_unsigned(RPC_VER2);
rpc_call = txdr_unsigned(RPC_CALL);
rpc_reply = txdr_unsigned(RPC_REPLY);
rpc_msgdenied = txdr_unsigned(RPC_MSGDENIED);
rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED);
rpc_mismatch = txdr_unsigned(RPC_MISMATCH);
rpc_autherr = txdr_unsigned(RPC_AUTHERR);
rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX);
nfs_true = txdr_unsigned(TRUE);
nfs_false = txdr_unsigned(FALSE);
nfs_xdrneg1 = txdr_unsigned(-1);
nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000;
if (nfs_ticks < 1)
nfs_ticks = 1;
1994-05-24 10:09:53 +00:00
/* Ensure async daemons disabled */
for (i = 0; i < NFS_MAXASYNCDAEMON; i++) {
nfs_iodwant[i] = NULL;
nfs_iodmount[i] = NULL;
}
1994-05-24 10:09:53 +00:00
nfs_nhinit(); /* Init the nfsnode table */
/*
* Initialize reply list and start timer
*/
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
TAILQ_INIT(&nfs_reqq);
mtx_init(&nfs_reqq_mtx, "NFS reqq lock", NULL, MTX_DEF);
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
callout_init(&nfs_callout, CALLOUT_MPSAFE);
#endif
mtx_init(&nfs_iod_mtx, "NFS iod lock", NULL, MTX_DEF);
mtx_init(&nfs_xid_mtx, "NFS xid lock", NULL, MTX_DEF);
nfs_pbuf_freecnt = nswbuf / 2 + 1;
return (0);
1994-05-24 10:09:53 +00:00
}
int
nfs_uninit(struct vfsconf *vfsp)
{
int i;
Implement support for RPCSEC_GSS authentication to both the NFS client and server. This replaces the RPC implementation of the NFS client and server with the newer RPC implementation originally developed (actually ported from the userland sunrpc code) to support the NFS Lock Manager. I have tested this code extensively and I believe it is stable and that performance is at least equal to the legacy RPC implementation. The NFS code currently contains support for both the new RPC implementation and the older legacy implementation inherited from the original NFS codebase. The default is to use the new implementation - add the NFS_LEGACYRPC option to fall back to the old code. When I merge this support back to RELENG_7, I will probably change this so that users have to 'opt in' to get the new code. To use RPCSEC_GSS on either client or server, you must build a kernel which includes the KGSSAPI option and the crypto device. On the userland side, you must build at least a new libc, mountd, mount_nfs and gssd. You must install new versions of /etc/rc.d/gssd and /etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf. As long as gssd is running, you should be able to mount an NFS filesystem from a server that requires RPCSEC_GSS authentication. The mount itself can happen without any kerberos credentials but all access to the filesystem will be denied unless the accessing user has a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There is currently no support for situations where the ticket file is in a different place, such as when the user logged in via SSH and has delegated credentials from that login. This restriction is also present in Solaris and Linux. In theory, we could improve this in future, possibly using Brooks Davis' implementation of variant symlinks. Supporting RPCSEC_GSS on a server is nearly as simple. You must create service creds for the server in the form 'nfs/<fqdn>@<REALM>' and install them in /etc/krb5.keytab. The standard heimdal utility ktutil makes this fairly easy. After the service creds have been created, you can add a '-sec=krb5' option to /etc/exports and restart both mountd and nfsd. The only other difference an administrator should notice is that nfsd doesn't fork to create service threads any more. In normal operation, there will be two nfsd processes, one in userland waiting for TCP connections and one in the kernel handling requests. The latter process will create as many kthreads as required - these should be visible via 'top -H'. The code has some support for varying the number of service threads according to load but initially at least, nfsd uses a fixed number of threads according to the value supplied to its '-n' option. Sponsored by: Isilon Systems MFC after: 1 month
2008-11-03 10:38:00 +00:00
#ifdef NFS_LEGACYRPC
callout_stop(&nfs_callout);
KASSERT(TAILQ_EMPTY(&nfs_reqq),
("nfs_uninit: request queue not empty"));
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
/*
* Tell all nfsiod processes to exit. Clear nfs_iodmax, and wakeup
* any sleeping nfsiods so they check nfs_iodmax and exit.
*/
mtx_lock(&nfs_iod_mtx);
nfs_iodmax = 0;
for (i = 0; i < nfs_numasync; i++)
if (nfs_iodwant[i])
wakeup(&nfs_iodwant[i]);
/* The last nfsiod to exit will wake us up when nfs_numasync hits 0 */
while (nfs_numasync)
msleep(&nfs_numasync, &nfs_iod_mtx, PWAIT, "ioddie", 0);
mtx_unlock(&nfs_iod_mtx);
nfs_nhuninit();
uma_zdestroy(nfsmount_zone);
return (0);
}
void
nfs_dircookie_lock(struct nfsnode *np)
{
mtx_lock(&np->n_mtx);
while (np->n_flag & NDIRCOOKIELK)
(void) msleep(&np->n_flag, &np->n_mtx, PZERO, "nfsdirlk", 0);
np->n_flag |= NDIRCOOKIELK;
mtx_unlock(&np->n_mtx);
}
void
nfs_dircookie_unlock(struct nfsnode *np)
{
mtx_lock(&np->n_mtx);
np->n_flag &= ~NDIRCOOKIELK;
wakeup(&np->n_flag);
mtx_unlock(&np->n_mtx);
}
int
nfs_upgrade_vnlock(struct vnode *vp)
{
int old_lock;
if ((old_lock = VOP_ISLOCKED(vp)) != LK_EXCLUSIVE) {
if (old_lock == LK_SHARED) {
/* Upgrade to exclusive lock, this might block */
vn_lock(vp, LK_UPGRADE | LK_RETRY);
} else {
vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
}
}
return old_lock;
}
void
nfs_downgrade_vnlock(struct vnode *vp, int old_lock)
{
if (old_lock != LK_EXCLUSIVE) {
if (old_lock == LK_SHARED) {
/* Downgrade from exclusive lock, this might block */
vn_lock(vp, LK_DOWNGRADE);
} else {
VOP_UNLOCK(vp, 0);
}
}
}
void
nfs_printf(const char *fmt, ...)
{
va_list ap;
mtx_lock(&Giant);
va_start(ap, fmt);
printf(fmt, ap);
va_end(ap);
mtx_unlock(&Giant);
}
1994-05-24 10:09:53 +00:00
/*
* Attribute cache routines.
* nfs_loadattrcache() - loads or updates the cache contents from attributes
* that are on the mbuf list
* nfs_getattrcache() - returns valid attributes if found in cache, returns
* error otherwise
*/
/*
* Load the attribute cache (that lives in the nfsnode entry) with
* the values on the mbuf list and
* Iff vap not NULL
* copy the attributes to *vaper
*/
int
nfs_loadattrcache(struct vnode **vpp, struct mbuf **mdp, caddr_t *dposp,
struct vattr *vaper, int dontshrink)
1994-05-24 10:09:53 +00:00
{
struct vnode *vp = *vpp;
struct vattr *vap;
struct nfs_fattr *fp;
struct nfsnode *np;
int32_t t1;
caddr_t cp2;
int rdev;
1994-05-24 10:09:53 +00:00
struct mbuf *md;
enum vtype vtyp;
u_short vmode;
struct timespec mtime, mtime_save;
int v3 = NFS_ISV3(vp);
struct thread *td = curthread;
1994-05-24 10:09:53 +00:00
md = *mdp;
t1 = (mtod(md, caddr_t) + md->m_len) - *dposp;
cp2 = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, M_WAIT);
if (cp2 == NULL)
return EBADRPC;
fp = (struct nfs_fattr *)cp2;
if (v3) {
vtyp = nfsv3tov_type(fp->fa_type);
vmode = fxdr_unsigned(u_short, fp->fa_mode);
rdev = makedev(fxdr_unsigned(int, fp->fa3_rdev.specdata1),
fxdr_unsigned(int, fp->fa3_rdev.specdata2));
fxdr_nfsv3time(&fp->fa3_mtime, &mtime);
1994-05-24 10:09:53 +00:00
} else {
vtyp = nfsv2tov_type(fp->fa_type);
vmode = fxdr_unsigned(u_short, fp->fa_mode);
/*
* XXX
*
* The duplicate information returned in fa_type and fa_mode
* is an ambiguity in the NFS version 2 protocol.
*
* VREG should be taken literally as a regular file. If a
* server intents to return some type information differently
* in the upper bits of the mode field (e.g. for sockets, or
* FIFOs), NFSv2 mandates fa_type to be VNON. Anyway, we
* leave the examination of the mode bits even in the VREG
* case to avoid breakage for bogus servers, but we make sure
* that there are actually type bits set in the upper part of
* fa_mode (and failing that, trust the va_type field).
*
* NFSv3 cleared the issue, and requires fa_mode to not
* contain any type information (while also introduing sockets
* and FIFOs for fa_type).
*/
if (vtyp == VNON || (vtyp == VREG && (vmode & S_IFMT) != 0))
vtyp = IFTOVT(vmode);
rdev = fxdr_unsigned(int32_t, fp->fa2_rdev);
fxdr_nfsv2time(&fp->fa2_mtime, &mtime);
/*
* Really ugly NFSv2 kludge.
*/
if (vtyp == VCHR && rdev == 0xffffffff)
vtyp = VFIFO;
1994-05-24 10:09:53 +00:00
}
1994-05-24 10:09:53 +00:00
/*
* If v_type == VNON it is a new node, so fill in the v_type,
1995-05-30 08:16:23 +00:00
* n_mtime fields. Check to see if it represents a special
1994-05-24 10:09:53 +00:00
* device, and if so, check for a possible alias. Once the
* correct vnode has been obtained, fill in the rest of the
* information.
*/
np = VTONFS(vp);
mtx_lock(&np->n_mtx);
if (vp->v_type != vtyp) {
vp->v_type = vtyp;
if (vp->v_type == VFIFO)
vp->v_op = &nfs_fifoops;
np->n_mtime = mtime;
1994-05-24 10:09:53 +00:00
}
vap = &np->n_vattr;
vap->va_type = vtyp;
vap->va_mode = (vmode & 07777);
Divorce "dev_t" from the "major|minor" bitmap, which is now called udev_t in the kernel but still called dev_t in userland. Provide functions to manipulate both types: major() umajor() minor() uminor() makedev() umakedev() dev2udev() udev2dev() For now they're functions, they will become in-line functions after one of the next two steps in this process. Return major/minor/makedev to macro-hood for userland. Register a name in cdevsw[] for the "filedescriptor" driver. In the kernel the udev_t appears in places where we have the major/minor number combination, (ie: a potential device: we may not have the driver nor the device), like in inodes, vattr, cdevsw registration and so on, whereas the dev_t appears where we carry around a reference to a actual device. In the future the cdevsw and the aliased-from vnode will be hung directly from the dev_t, along with up to two softc pointers for the device driver and a few houskeeping bits. This will essentially replace the current "alias" check code (same buck, bigger bang). A little stunt has been provided to try to catch places where the wrong type is being used (dev_t vs udev_t), if you see something not working, #undef DEVT_FASCIST in kern/kern_conf.c and see if it makes a difference. If it does, please try to track it down (many hands make light work) or at least try to reproduce it as simply as possible, and describe how to do that. Without DEVT_FASCIST I belive this patch is a no-op. Stylistic/posixoid comments about the userland view of the <sys/*.h> files welcome now, from userland they now contain the end result. Next planned step: make all dev_t's refer to the same devsw[] which means convert BLK's to CHR's at the perimeter of the vnodes and other places where they enter the game (bootdev, mknod, sysctl).
1999-05-11 19:55:07 +00:00
vap->va_rdev = rdev;
mtime_save = vap->va_mtime;
1994-05-24 10:09:53 +00:00
vap->va_mtime = mtime;
vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0];
if (v3) {
vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
vap->va_size = fxdr_hyper(&fp->fa3_size);
vap->va_blocksize = NFS_FABLKSIZE;
vap->va_bytes = fxdr_hyper(&fp->fa3_used);
vap->va_fileid = fxdr_unsigned(int32_t,
fp->fa3_fileid.nfsuquad[1]);
fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime);
fxdr_nfsv3time(&fp->fa3_ctime, &vap->va_ctime);
vap->va_flags = 0;
vap->va_filerev = 0;
1994-05-24 10:09:53 +00:00
} else {
vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
vap->va_size = fxdr_unsigned(u_int32_t, fp->fa2_size);
vap->va_blocksize = fxdr_unsigned(int32_t, fp->fa2_blocksize);
vap->va_bytes = (u_quad_t)fxdr_unsigned(int32_t, fp->fa2_blocks)
* NFS_FABLKSIZE;
vap->va_fileid = fxdr_unsigned(int32_t, fp->fa2_fileid);
fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime);
1994-05-24 10:09:53 +00:00
vap->va_flags = 0;
vap->va_ctime.tv_sec = fxdr_unsigned(u_int32_t,
fp->fa2_ctime.nfsv2_sec);
vap->va_ctime.tv_nsec = 0;
vap->va_gen = fxdr_unsigned(u_int32_t, fp->fa2_ctime.nfsv2_usec);
1994-05-24 10:09:53 +00:00
vap->va_filerev = 0;
}
np->n_attrstamp = time_second;
/* Timestamp the NFS otw getattr fetch */
if (td->td_proc) {
np->n_ac_ts_tid = td->td_tid;
np->n_ac_ts_pid = td->td_proc->p_pid;
np->n_ac_ts_syscalls = td->td_syscalls;
} else
bzero(&np->n_ac_ts, sizeof(struct nfs_attrcache_timestamp));
1994-05-24 10:09:53 +00:00
if (vap->va_size != np->n_size) {
if (vap->va_type == VREG) {
if (dontshrink && vap->va_size < np->n_size) {
/*
* We've been told not to shrink the file;
* zero np->n_attrstamp to indicate that
* the attributes are stale.
*/
vap->va_size = np->n_size;
np->n_attrstamp = 0;
} else if (np->n_flag & NMODIFIED) {
/*
* We've modified the file: Use the larger
* of our size, and the server's size.
*/
if (vap->va_size < np->n_size) {
1994-05-24 10:09:53 +00:00
vap->va_size = np->n_size;
} else {
1994-05-24 10:09:53 +00:00
np->n_size = vap->va_size;
np->n_flag |= NSIZECHANGED;
}
Synopsis of problem being fixed: Dan Nelson originally reported that blocks of zeros could wind up in a file written to over NFS by a client. The problem only occurs a few times per several gigabytes of data. This problem turned out to be bug #3 below. bug #1: B_CLUSTEROK must be cleared when an NFS buffer is reverted from stage 2 (ready for commit rpc) to stage 1 (ready for write). Reversions can occur when a dirty NFS buffer is redirtied with new data. Otherwise the VFS/BIO system may end up thinking that a stage 1 NFS buffer is clusterable. Stage 1 NFS buffers are not clusterable. bug #2: B_CLUSTEROK was inappropriately set for a 'short' NFS buffer (short buffers only occur near the EOF of the file). Change to only set when the buffer is a full biosize (usually 8K). This bug has no effect but should be fixed in -current anyway. It need not be backported. bug #3: B_NEEDCOMMIT was inappropriately set in nfs_flush() (which is typically only called by the update daemon). nfs_flush() does a multi-pass loop but due to the lack of vnode locking it is possible for new buffers to be added to the dirtyblkhd list while a flush operation is going on. This may result in nfs_flush() setting B_NEEDCOMMIT on a buffer which has *NOT* yet gone through its stage 1 write, causing only the commit rpc to be made and thus causing the contents of the buffer to be thrown away (never sent to the server). The patch also contains some cleanup, which only applies to the commit into -current. Reviewed by: dg, julian Originally Reported by: Dan Nelson <dnelson@emsphone.com>
1999-12-12 06:09:57 +00:00
} else {
1994-05-24 10:09:53 +00:00
np->n_size = vap->va_size;
np->n_flag |= NSIZECHANGED;
Synopsis of problem being fixed: Dan Nelson originally reported that blocks of zeros could wind up in a file written to over NFS by a client. The problem only occurs a few times per several gigabytes of data. This problem turned out to be bug #3 below. bug #1: B_CLUSTEROK must be cleared when an NFS buffer is reverted from stage 2 (ready for commit rpc) to stage 1 (ready for write). Reversions can occur when a dirty NFS buffer is redirtied with new data. Otherwise the VFS/BIO system may end up thinking that a stage 1 NFS buffer is clusterable. Stage 1 NFS buffers are not clusterable. bug #2: B_CLUSTEROK was inappropriately set for a 'short' NFS buffer (short buffers only occur near the EOF of the file). Change to only set when the buffer is a full biosize (usually 8K). This bug has no effect but should be fixed in -current anyway. It need not be backported. bug #3: B_NEEDCOMMIT was inappropriately set in nfs_flush() (which is typically only called by the update daemon). nfs_flush() does a multi-pass loop but due to the lack of vnode locking it is possible for new buffers to be added to the dirtyblkhd list while a flush operation is going on. This may result in nfs_flush() setting B_NEEDCOMMIT on a buffer which has *NOT* yet gone through its stage 1 write, causing only the commit rpc to be made and thus causing the contents of the buffer to be thrown away (never sent to the server). The patch also contains some cleanup, which only applies to the commit into -current. Reviewed by: dg, julian Originally Reported by: Dan Nelson <dnelson@emsphone.com>
1999-12-12 06:09:57 +00:00
}
vnode_pager_setsize(vp, np->n_size);
Synopsis of problem being fixed: Dan Nelson originally reported that blocks of zeros could wind up in a file written to over NFS by a client. The problem only occurs a few times per several gigabytes of data. This problem turned out to be bug #3 below. bug #1: B_CLUSTEROK must be cleared when an NFS buffer is reverted from stage 2 (ready for commit rpc) to stage 1 (ready for write). Reversions can occur when a dirty NFS buffer is redirtied with new data. Otherwise the VFS/BIO system may end up thinking that a stage 1 NFS buffer is clusterable. Stage 1 NFS buffers are not clusterable. bug #2: B_CLUSTEROK was inappropriately set for a 'short' NFS buffer (short buffers only occur near the EOF of the file). Change to only set when the buffer is a full biosize (usually 8K). This bug has no effect but should be fixed in -current anyway. It need not be backported. bug #3: B_NEEDCOMMIT was inappropriately set in nfs_flush() (which is typically only called by the update daemon). nfs_flush() does a multi-pass loop but due to the lack of vnode locking it is possible for new buffers to be added to the dirtyblkhd list while a flush operation is going on. This may result in nfs_flush() setting B_NEEDCOMMIT on a buffer which has *NOT* yet gone through its stage 1 write, causing only the commit rpc to be made and thus causing the contents of the buffer to be thrown away (never sent to the server). The patch also contains some cleanup, which only applies to the commit into -current. Reviewed by: dg, julian Originally Reported by: Dan Nelson <dnelson@emsphone.com>
1999-12-12 06:09:57 +00:00
} else {
1994-05-24 10:09:53 +00:00
np->n_size = vap->va_size;
Synopsis of problem being fixed: Dan Nelson originally reported that blocks of zeros could wind up in a file written to over NFS by a client. The problem only occurs a few times per several gigabytes of data. This problem turned out to be bug #3 below. bug #1: B_CLUSTEROK must be cleared when an NFS buffer is reverted from stage 2 (ready for commit rpc) to stage 1 (ready for write). Reversions can occur when a dirty NFS buffer is redirtied with new data. Otherwise the VFS/BIO system may end up thinking that a stage 1 NFS buffer is clusterable. Stage 1 NFS buffers are not clusterable. bug #2: B_CLUSTEROK was inappropriately set for a 'short' NFS buffer (short buffers only occur near the EOF of the file). Change to only set when the buffer is a full biosize (usually 8K). This bug has no effect but should be fixed in -current anyway. It need not be backported. bug #3: B_NEEDCOMMIT was inappropriately set in nfs_flush() (which is typically only called by the update daemon). nfs_flush() does a multi-pass loop but due to the lack of vnode locking it is possible for new buffers to be added to the dirtyblkhd list while a flush operation is going on. This may result in nfs_flush() setting B_NEEDCOMMIT on a buffer which has *NOT* yet gone through its stage 1 write, causing only the commit rpc to be made and thus causing the contents of the buffer to be thrown away (never sent to the server). The patch also contains some cleanup, which only applies to the commit into -current. Reviewed by: dg, julian Originally Reported by: Dan Nelson <dnelson@emsphone.com>
1999-12-12 06:09:57 +00:00
}
1994-05-24 10:09:53 +00:00
}
/*
* The following checks are added to prevent a race between (say)
* a READDIR+ and a WRITE.
* READDIR+, WRITE requests sent out.
* READDIR+ resp, WRITE resp received on client.
* However, the WRITE resp was handled before the READDIR+ resp
* causing the post op attrs from the write to be loaded first
* and the attrs from the READDIR+ to be loaded later. If this
* happens, we have stale attrs loaded into the attrcache.
* We detect this by for the mtime moving back. We invalidate the
* attrcache when this happens.
*/
if (timespeccmp(&mtime_save, &vap->va_mtime, >))
/* Size changed or mtime went backwards */
np->n_attrstamp = 0;
1994-05-24 10:09:53 +00:00
if (vaper != NULL) {
bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(*vap));
if (np->n_flag & NCHG) {
if (np->n_flag & NACC)
vaper->va_atime = np->n_atim;
if (np->n_flag & NUPD)
vaper->va_mtime = np->n_mtim;
1994-05-24 10:09:53 +00:00
}
}
mtx_unlock(&np->n_mtx);
1994-05-24 10:09:53 +00:00
return (0);
}
#ifdef NFS_ACDEBUG
#include <sys/sysctl.h>
1999-02-17 13:59:29 +00:00
SYSCTL_DECL(_vfs_nfs);
static int nfs_acdebug;
SYSCTL_INT(_vfs_nfs, OID_AUTO, acdebug, CTLFLAG_RW, &nfs_acdebug, 0,
"Toggle acdebug (access cache debug) flag");
#endif
1994-05-24 10:09:53 +00:00
/*
* Check the time stamp
* If the cache is valid, copy contents to *vap and return 0
* otherwise return an error
*/
int
nfs_getattrcache(struct vnode *vp, struct vattr *vaper)
1994-05-24 10:09:53 +00:00
{
struct nfsnode *np;
struct vattr *vap;
struct nfsmount *nmp;
int timeo;
np = VTONFS(vp);
vap = &np->n_vattr;
nmp = VFSTONFS(vp->v_mount);
#ifdef NFS_ACDEBUG
mtx_lock(&Giant); /* nfs_printf() */
#endif
mtx_lock(&np->n_mtx);
/* XXX n_mtime doesn't seem to be updated on a miss-and-reload */
timeo = (time_second - np->n_mtime.tv_sec) / 10;
#ifdef NFS_ACDEBUG
if (nfs_acdebug>1)
nfs_printf("nfs_getattrcache: initial timeo = %d\n", timeo);
#endif
if (vap->va_type == VDIR) {
if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acdirmin)
timeo = nmp->nm_acdirmin;
else if (timeo > nmp->nm_acdirmax)
timeo = nmp->nm_acdirmax;
} else {
if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acregmin)
timeo = nmp->nm_acregmin;
else if (timeo > nmp->nm_acregmax)
timeo = nmp->nm_acregmax;
}
1994-05-24 10:09:53 +00:00
#ifdef NFS_ACDEBUG
if (nfs_acdebug > 2)
nfs_printf("acregmin %d; acregmax %d; acdirmin %d; acdirmax %d\n",
nmp->nm_acregmin, nmp->nm_acregmax,
nmp->nm_acdirmin, nmp->nm_acdirmax);
if (nfs_acdebug)
nfs_printf("nfs_getattrcache: age = %d; final timeo = %d\n",
(time_second - np->n_attrstamp), timeo);
#endif
if ((time_second - np->n_attrstamp) >= timeo) {
1994-05-24 10:09:53 +00:00
nfsstats.attrcache_misses++;
mtx_unlock(&np->n_mtx);
return( ENOENT);
1994-05-24 10:09:53 +00:00
}
nfsstats.attrcache_hits++;
if (vap->va_size != np->n_size) {
if (vap->va_type == VREG) {
if (np->n_flag & NMODIFIED) {
if (vap->va_size < np->n_size)
vap->va_size = np->n_size;
else
np->n_size = vap->va_size;
Synopsis of problem being fixed: Dan Nelson originally reported that blocks of zeros could wind up in a file written to over NFS by a client. The problem only occurs a few times per several gigabytes of data. This problem turned out to be bug #3 below. bug #1: B_CLUSTEROK must be cleared when an NFS buffer is reverted from stage 2 (ready for commit rpc) to stage 1 (ready for write). Reversions can occur when a dirty NFS buffer is redirtied with new data. Otherwise the VFS/BIO system may end up thinking that a stage 1 NFS buffer is clusterable. Stage 1 NFS buffers are not clusterable. bug #2: B_CLUSTEROK was inappropriately set for a 'short' NFS buffer (short buffers only occur near the EOF of the file). Change to only set when the buffer is a full biosize (usually 8K). This bug has no effect but should be fixed in -current anyway. It need not be backported. bug #3: B_NEEDCOMMIT was inappropriately set in nfs_flush() (which is typically only called by the update daemon). nfs_flush() does a multi-pass loop but due to the lack of vnode locking it is possible for new buffers to be added to the dirtyblkhd list while a flush operation is going on. This may result in nfs_flush() setting B_NEEDCOMMIT on a buffer which has *NOT* yet gone through its stage 1 write, causing only the commit rpc to be made and thus causing the contents of the buffer to be thrown away (never sent to the server). The patch also contains some cleanup, which only applies to the commit into -current. Reviewed by: dg, julian Originally Reported by: Dan Nelson <dnelson@emsphone.com>
1999-12-12 06:09:57 +00:00
} else {
1994-05-24 10:09:53 +00:00
np->n_size = vap->va_size;
Synopsis of problem being fixed: Dan Nelson originally reported that blocks of zeros could wind up in a file written to over NFS by a client. The problem only occurs a few times per several gigabytes of data. This problem turned out to be bug #3 below. bug #1: B_CLUSTEROK must be cleared when an NFS buffer is reverted from stage 2 (ready for commit rpc) to stage 1 (ready for write). Reversions can occur when a dirty NFS buffer is redirtied with new data. Otherwise the VFS/BIO system may end up thinking that a stage 1 NFS buffer is clusterable. Stage 1 NFS buffers are not clusterable. bug #2: B_CLUSTEROK was inappropriately set for a 'short' NFS buffer (short buffers only occur near the EOF of the file). Change to only set when the buffer is a full biosize (usually 8K). This bug has no effect but should be fixed in -current anyway. It need not be backported. bug #3: B_NEEDCOMMIT was inappropriately set in nfs_flush() (which is typically only called by the update daemon). nfs_flush() does a multi-pass loop but due to the lack of vnode locking it is possible for new buffers to be added to the dirtyblkhd list while a flush operation is going on. This may result in nfs_flush() setting B_NEEDCOMMIT on a buffer which has *NOT* yet gone through its stage 1 write, causing only the commit rpc to be made and thus causing the contents of the buffer to be thrown away (never sent to the server). The patch also contains some cleanup, which only applies to the commit into -current. Reviewed by: dg, julian Originally Reported by: Dan Nelson <dnelson@emsphone.com>
1999-12-12 06:09:57 +00:00
}
vnode_pager_setsize(vp, np->n_size);
Synopsis of problem being fixed: Dan Nelson originally reported that blocks of zeros could wind up in a file written to over NFS by a client. The problem only occurs a few times per several gigabytes of data. This problem turned out to be bug #3 below. bug #1: B_CLUSTEROK must be cleared when an NFS buffer is reverted from stage 2 (ready for commit rpc) to stage 1 (ready for write). Reversions can occur when a dirty NFS buffer is redirtied with new data. Otherwise the VFS/BIO system may end up thinking that a stage 1 NFS buffer is clusterable. Stage 1 NFS buffers are not clusterable. bug #2: B_CLUSTEROK was inappropriately set for a 'short' NFS buffer (short buffers only occur near the EOF of the file). Change to only set when the buffer is a full biosize (usually 8K). This bug has no effect but should be fixed in -current anyway. It need not be backported. bug #3: B_NEEDCOMMIT was inappropriately set in nfs_flush() (which is typically only called by the update daemon). nfs_flush() does a multi-pass loop but due to the lack of vnode locking it is possible for new buffers to be added to the dirtyblkhd list while a flush operation is going on. This may result in nfs_flush() setting B_NEEDCOMMIT on a buffer which has *NOT* yet gone through its stage 1 write, causing only the commit rpc to be made and thus causing the contents of the buffer to be thrown away (never sent to the server). The patch also contains some cleanup, which only applies to the commit into -current. Reviewed by: dg, julian Originally Reported by: Dan Nelson <dnelson@emsphone.com>
1999-12-12 06:09:57 +00:00
} else {
1994-05-24 10:09:53 +00:00
np->n_size = vap->va_size;
Synopsis of problem being fixed: Dan Nelson originally reported that blocks of zeros could wind up in a file written to over NFS by a client. The problem only occurs a few times per several gigabytes of data. This problem turned out to be bug #3 below. bug #1: B_CLUSTEROK must be cleared when an NFS buffer is reverted from stage 2 (ready for commit rpc) to stage 1 (ready for write). Reversions can occur when a dirty NFS buffer is redirtied with new data. Otherwise the VFS/BIO system may end up thinking that a stage 1 NFS buffer is clusterable. Stage 1 NFS buffers are not clusterable. bug #2: B_CLUSTEROK was inappropriately set for a 'short' NFS buffer (short buffers only occur near the EOF of the file). Change to only set when the buffer is a full biosize (usually 8K). This bug has no effect but should be fixed in -current anyway. It need not be backported. bug #3: B_NEEDCOMMIT was inappropriately set in nfs_flush() (which is typically only called by the update daemon). nfs_flush() does a multi-pass loop but due to the lack of vnode locking it is possible for new buffers to be added to the dirtyblkhd list while a flush operation is going on. This may result in nfs_flush() setting B_NEEDCOMMIT on a buffer which has *NOT* yet gone through its stage 1 write, causing only the commit rpc to be made and thus causing the contents of the buffer to be thrown away (never sent to the server). The patch also contains some cleanup, which only applies to the commit into -current. Reviewed by: dg, julian Originally Reported by: Dan Nelson <dnelson@emsphone.com>
1999-12-12 06:09:57 +00:00
}
1994-05-24 10:09:53 +00:00
}
bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(struct vattr));
if (np->n_flag & NCHG) {
if (np->n_flag & NACC)
vaper->va_atime = np->n_atim;
if (np->n_flag & NUPD)
vaper->va_mtime = np->n_mtim;
1994-05-24 10:09:53 +00:00
}
mtx_unlock(&np->n_mtx);
#ifdef NFS_ACDEBUG
mtx_unlock(&Giant); /* nfs_printf() */
#endif
1994-05-24 10:09:53 +00:00
return (0);
}
static nfsuint64 nfs_nullcookie = { { 0, 0 } };
/*
* This function finds the directory cookie that corresponds to the
* logical byte offset given.
*/
nfsuint64 *
nfs_getcookie(struct nfsnode *np, off_t off, int add)
{
struct nfsdmap *dp, *dp2;
int pos;
nfsuint64 *retval = NULL;
pos = (uoff_t)off / NFS_DIRBLKSIZ;
if (pos == 0 || off < 0) {
#ifdef DIAGNOSTIC
if (add)
panic("nfs getcookie add at <= 0");
#endif
return (&nfs_nullcookie);
}
pos--;
dp = LIST_FIRST(&np->n_cookies);
if (!dp) {
if (add) {
dp = malloc(sizeof (struct nfsdmap),
M_NFSDIROFF, M_WAITOK);
dp->ndm_eocookie = 0;
LIST_INSERT_HEAD(&np->n_cookies, dp, ndm_list);
} else
goto out;
}
while (pos >= NFSNUMCOOKIES) {
pos -= NFSNUMCOOKIES;
if (LIST_NEXT(dp, ndm_list)) {
if (!add && dp->ndm_eocookie < NFSNUMCOOKIES &&
pos >= dp->ndm_eocookie)
goto out;
dp = LIST_NEXT(dp, ndm_list);
} else if (add) {
dp2 = malloc(sizeof (struct nfsdmap),
M_NFSDIROFF, M_WAITOK);
dp2->ndm_eocookie = 0;
LIST_INSERT_AFTER(dp, dp2, ndm_list);
dp = dp2;
} else
goto out;
}
if (pos >= dp->ndm_eocookie) {
if (add)
dp->ndm_eocookie = pos + 1;
else
goto out;
}
retval = &dp->ndm_cookies[pos];
out:
return (retval);
}
/*
* Invalidate cached directory information, except for the actual directory
* blocks (which are invalidated separately).
* Done mainly to avoid the use of stale offset cookies.
*/
void
nfs_invaldir(struct vnode *vp)
{
struct nfsnode *np = VTONFS(vp);
#ifdef DIAGNOSTIC
if (vp->v_type != VDIR)
panic("nfs: invaldir not dir");
#endif
nfs_dircookie_lock(np);
np->n_direofoffset = 0;
np->n_cookieverf.nfsuquad[0] = 0;
np->n_cookieverf.nfsuquad[1] = 0;
if (LIST_FIRST(&np->n_cookies))
LIST_FIRST(&np->n_cookies)->ndm_eocookie = 0;
nfs_dircookie_unlock(np);
}
/*
* The write verifier has changed (probably due to a server reboot), so all
* B_NEEDCOMMIT blocks will have to be written again. Since they are on the
* dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT
Synopsis of problem being fixed: Dan Nelson originally reported that blocks of zeros could wind up in a file written to over NFS by a client. The problem only occurs a few times per several gigabytes of data. This problem turned out to be bug #3 below. bug #1: B_CLUSTEROK must be cleared when an NFS buffer is reverted from stage 2 (ready for commit rpc) to stage 1 (ready for write). Reversions can occur when a dirty NFS buffer is redirtied with new data. Otherwise the VFS/BIO system may end up thinking that a stage 1 NFS buffer is clusterable. Stage 1 NFS buffers are not clusterable. bug #2: B_CLUSTEROK was inappropriately set for a 'short' NFS buffer (short buffers only occur near the EOF of the file). Change to only set when the buffer is a full biosize (usually 8K). This bug has no effect but should be fixed in -current anyway. It need not be backported. bug #3: B_NEEDCOMMIT was inappropriately set in nfs_flush() (which is typically only called by the update daemon). nfs_flush() does a multi-pass loop but due to the lack of vnode locking it is possible for new buffers to be added to the dirtyblkhd list while a flush operation is going on. This may result in nfs_flush() setting B_NEEDCOMMIT on a buffer which has *NOT* yet gone through its stage 1 write, causing only the commit rpc to be made and thus causing the contents of the buffer to be thrown away (never sent to the server). The patch also contains some cleanup, which only applies to the commit into -current. Reviewed by: dg, julian Originally Reported by: Dan Nelson <dnelson@emsphone.com>
1999-12-12 06:09:57 +00:00
* and B_CLUSTEROK flags. Once done the new write verifier can be set for the
* mount point.
*
* B_CLUSTEROK must be cleared along with B_NEEDCOMMIT because stage 1 data
Synopsis of problem being fixed: Dan Nelson originally reported that blocks of zeros could wind up in a file written to over NFS by a client. The problem only occurs a few times per several gigabytes of data. This problem turned out to be bug #3 below. bug #1: B_CLUSTEROK must be cleared when an NFS buffer is reverted from stage 2 (ready for commit rpc) to stage 1 (ready for write). Reversions can occur when a dirty NFS buffer is redirtied with new data. Otherwise the VFS/BIO system may end up thinking that a stage 1 NFS buffer is clusterable. Stage 1 NFS buffers are not clusterable. bug #2: B_CLUSTEROK was inappropriately set for a 'short' NFS buffer (short buffers only occur near the EOF of the file). Change to only set when the buffer is a full biosize (usually 8K). This bug has no effect but should be fixed in -current anyway. It need not be backported. bug #3: B_NEEDCOMMIT was inappropriately set in nfs_flush() (which is typically only called by the update daemon). nfs_flush() does a multi-pass loop but due to the lack of vnode locking it is possible for new buffers to be added to the dirtyblkhd list while a flush operation is going on. This may result in nfs_flush() setting B_NEEDCOMMIT on a buffer which has *NOT* yet gone through its stage 1 write, causing only the commit rpc to be made and thus causing the contents of the buffer to be thrown away (never sent to the server). The patch also contains some cleanup, which only applies to the commit into -current. Reviewed by: dg, julian Originally Reported by: Dan Nelson <dnelson@emsphone.com>
1999-12-12 06:09:57 +00:00
* writes are not clusterable.
*/
void
nfs_clearcommit(struct mount *mp)
{
struct vnode *vp, *nvp;
struct buf *bp, *nbp;
struct bufobj *bo;
MNT_ILOCK(mp);
When we traverse the vnodes on a mountpoint we need to look out for our cached 'next vnode' being removed from this mountpoint. If we find that it was recycled, we restart our traversal from the start of the list. Code to do that is in all local disk filesystems (and a few other places) and looks roughly like this: MNT_ILOCK(mp); loop: for (vp = TAILQ_FIRST(&mp...); (vp = nvp) != NULL; nvp = TAILQ_NEXT(vp,...)) { if (vp->v_mount != mp) goto loop; MNT_IUNLOCK(mp); ... MNT_ILOCK(mp); } MNT_IUNLOCK(mp); The code which takes vnodes off a mountpoint looks like this: MNT_ILOCK(vp->v_mount); ... TAILQ_REMOVE(&vp->v_mount->mnt_nvnodelist, vp, v_nmntvnodes); ... MNT_IUNLOCK(vp->v_mount); ... vp->v_mount = something; (Take a moment and try to spot the locking error before you read on.) On a SMP system, one CPU could have removed nvp from our mountlist but not yet gotten to assign a new value to vp->v_mount while another CPU simultaneously get to the top of the traversal loop where it finds that (vp->v_mount != mp) is not true despite the fact that the vnode has indeed been removed from our mountpoint. Fix: Introduce the macro MNT_VNODE_FOREACH() to traverse the list of vnodes on a mountpoint while taking into account that vnodes may be removed from the list as we go. This saves approx 65 lines of duplicated code. Split the insmntque() which potentially moves a vnode from one mount point to another into delmntque() and insmntque() which does just what the names say. Fix delmntque() to set vp->v_mount to NULL while holding the mountpoint lock.
2004-07-04 08:52:35 +00:00
MNT_VNODE_FOREACH(vp, mp, nvp) {
bo = &vp->v_bufobj;
VI_LOCK(vp);
if (vp->v_iflag & VI_DOOMED) {
VI_UNLOCK(vp);
continue;
}
vholdl(vp);
VI_UNLOCK(vp);
MNT_IUNLOCK(mp);
BO_LOCK(bo);
TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) {
if (!BUF_ISLOCKED(bp) &&
(bp->b_flags & (B_DELWRI | B_NEEDCOMMIT))
== (B_DELWRI | B_NEEDCOMMIT))
Synopsis of problem being fixed: Dan Nelson originally reported that blocks of zeros could wind up in a file written to over NFS by a client. The problem only occurs a few times per several gigabytes of data. This problem turned out to be bug #3 below. bug #1: B_CLUSTEROK must be cleared when an NFS buffer is reverted from stage 2 (ready for commit rpc) to stage 1 (ready for write). Reversions can occur when a dirty NFS buffer is redirtied with new data. Otherwise the VFS/BIO system may end up thinking that a stage 1 NFS buffer is clusterable. Stage 1 NFS buffers are not clusterable. bug #2: B_CLUSTEROK was inappropriately set for a 'short' NFS buffer (short buffers only occur near the EOF of the file). Change to only set when the buffer is a full biosize (usually 8K). This bug has no effect but should be fixed in -current anyway. It need not be backported. bug #3: B_NEEDCOMMIT was inappropriately set in nfs_flush() (which is typically only called by the update daemon). nfs_flush() does a multi-pass loop but due to the lack of vnode locking it is possible for new buffers to be added to the dirtyblkhd list while a flush operation is going on. This may result in nfs_flush() setting B_NEEDCOMMIT on a buffer which has *NOT* yet gone through its stage 1 write, causing only the commit rpc to be made and thus causing the contents of the buffer to be thrown away (never sent to the server). The patch also contains some cleanup, which only applies to the commit into -current. Reviewed by: dg, julian Originally Reported by: Dan Nelson <dnelson@emsphone.com>
1999-12-12 06:09:57 +00:00
bp->b_flags &= ~(B_NEEDCOMMIT | B_CLUSTEROK);
}
BO_UNLOCK(bo);
vdrop(vp);
MNT_ILOCK(mp);
}
MNT_IUNLOCK(mp);
}
/*
* Helper functions for former macros. Some of these should be
* moved to their callers.
*/
int
nfsm_mtofh_xx(struct vnode *d, struct vnode **v, int v3, int *f,
struct mbuf **md, caddr_t *dpos)
{
struct nfsnode *ttnp;
struct vnode *ttvp;
nfsfh_t *ttfhp;
u_int32_t *tl;
int ttfhsize;
int t1;
if (v3) {
tl = nfsm_dissect_xx(NFSX_UNSIGNED, md, dpos);
if (tl == NULL)
return EBADRPC;
*f = fxdr_unsigned(int, *tl);
} else
*f = 1;
if (*f) {
t1 = nfsm_getfh_xx(&ttfhp, &ttfhsize, (v3), md, dpos);
if (t1 != 0)
return t1;
t1 = nfs_nget(d->v_mount, ttfhp, ttfhsize, &ttnp, LK_EXCLUSIVE);
if (t1 != 0)
return t1;
*v = NFSTOV(ttnp);
}
if (v3) {
tl = nfsm_dissect_xx(NFSX_UNSIGNED, md, dpos);
if (tl == NULL)
return EBADRPC;
if (*f)
*f = fxdr_unsigned(int, *tl);
else if (fxdr_unsigned(int, *tl))
nfsm_adv_xx(NFSX_V3FATTR, md, dpos);
}
if (*f) {
ttvp = *v;
t1 = nfs_loadattrcache(&ttvp, md, dpos, NULL, 0);
if (t1)
return t1;
*v = ttvp;
}
return 0;
}
int
nfsm_getfh_xx(nfsfh_t **f, int *s, int v3, struct mbuf **md, caddr_t *dpos)
{
u_int32_t *tl;
if (v3) {
tl = nfsm_dissect_xx(NFSX_UNSIGNED, md, dpos);
if (tl == NULL)
return EBADRPC;
*s = fxdr_unsigned(int, *tl);
if (*s <= 0 || *s > NFSX_V3FHMAX)
return EBADRPC;
} else
*s = NFSX_V2FH;
*f = nfsm_dissect_xx(nfsm_rndup(*s), md, dpos);
if (*f == NULL)
return EBADRPC;
else
return 0;
}
int
nfsm_loadattr_xx(struct vnode **v, struct vattr *va, struct mbuf **md,
caddr_t *dpos)
{
int t1;
struct vnode *ttvp = *v;
t1 = nfs_loadattrcache(&ttvp, md, dpos, va, 0);
if (t1 != 0)
return t1;
*v = ttvp;
return 0;
}
int
nfsm_postop_attr_xx(struct vnode **v, int *f, struct mbuf **md,
caddr_t *dpos)
{
u_int32_t *tl;
int t1;
struct vnode *ttvp = *v;
tl = nfsm_dissect_xx(NFSX_UNSIGNED, md, dpos);
if (tl == NULL)
return EBADRPC;
*f = fxdr_unsigned(int, *tl);
if (*f != 0) {
t1 = nfs_loadattrcache(&ttvp, md, dpos, NULL, 1);
if (t1 != 0) {
*f = 0;
return t1;
}
*v = ttvp;
}
return 0;
}
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
int
nfsm_wcc_data_xx(struct vnode **v, int *f, struct mbuf **md, caddr_t *dpos)
{
u_int32_t *tl;
int ttattrf, ttretf = 0;
int t1;
tl = nfsm_dissect_xx(NFSX_UNSIGNED, md, dpos);
if (tl == NULL)
return EBADRPC;
if (*tl == nfs_true) {
tl = nfsm_dissect_xx(6 * NFSX_UNSIGNED, md, dpos);
if (tl == NULL)
return EBADRPC;
mtx_lock(&(VTONFS(*v))->n_mtx);
if (*f)
ttretf = (VTONFS(*v)->n_mtime.tv_sec == fxdr_unsigned(u_int32_t, *(tl + 2)) &&
VTONFS(*v)->n_mtime.tv_nsec == fxdr_unsigned(u_int32_t, *(tl + 3)));
mtx_unlock(&(VTONFS(*v))->n_mtx);
}
t1 = nfsm_postop_attr_xx(v, &ttattrf, md, dpos);
if (t1)
return t1;
if (*f)
*f = ttretf;
else
*f = ttattrf;
return 0;
}
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
int
nfsm_strtom_xx(const char *a, int s, int m, struct mbuf **mb, caddr_t *bpos)
{
u_int32_t *tl;
int t1;
if (s > m)
return ENAMETOOLONG;
t1 = nfsm_rndup(s) + NFSX_UNSIGNED;
if (t1 <= M_TRAILINGSPACE(*mb)) {
tl = nfsm_build_xx(t1, mb, bpos);
*tl++ = txdr_unsigned(s);
*(tl + ((t1 >> 2) - 2)) = 0;
bcopy(a, tl, s);
} else {
t1 = nfsm_strtmbuf(mb, bpos, a, s);
if (t1 != 0)
return t1;
}
return 0;
These changes embody the support of the fully coherent merged VM buffer cache, much higher filesystem I/O performance, and much better paging performance. It represents the culmination of over 6 months of R&D. The majority of the merged VM/cache work is by John Dyson. The following highlights the most significant changes. Additionally, there are (mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to support the new VM/buffer scheme. vfs_bio.c: Significant rewrite of most of vfs_bio to support the merged VM buffer cache scheme. The scheme is almost fully compatible with the old filesystem interface. Significant improvement in the number of opportunities for write clustering. vfs_cluster.c, vfs_subr.c Upgrade and performance enhancements in vfs layer code to support merged VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff. vm_object.c: Yet more improvements in the collapse code. Elimination of some windows that can cause list corruption. vm_pageout.c: Fixed it, it really works better now. Somehow in 2.0, some "enhancements" broke the code. This code has been reworked from the ground-up. vm_fault.c, vm_page.c, pmap.c, vm_object.c Support for small-block filesystems with merged VM/buffer cache scheme. pmap.c vm_map.c Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of kernel PTs. vm_glue.c Much simpler and more effective swapping code. No more gratuitous swapping. proc.h Fixed the problem that the p_lock flag was not being cleared on a fork. swap_pager.c, vnode_pager.c Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the code doesn't need it anymore. machdep.c Changes to better support the parameter values for the merged VM/buffer cache scheme. machdep.c, kern_exec.c, vm_glue.c Implemented a seperate submap for temporary exec string space and another one to contain process upages. This eliminates all map fragmentation problems that previously existed. ffs_inode.c, ufs_inode.c, ufs_readwrite.c Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on busy buffers. Submitted by: John Dyson and David Greenman
1995-01-09 16:06:02 +00:00
}
int
nfsm_fhtom_xx(struct vnode *v, int v3, struct mbuf **mb, caddr_t *bpos)
{
u_int32_t *tl;
int t1;
caddr_t cp;
if (v3) {
t1 = nfsm_rndup(VTONFS(v)->n_fhsize) + NFSX_UNSIGNED;
if (t1 < M_TRAILINGSPACE(*mb)) {
tl = nfsm_build_xx(t1, mb, bpos);
*tl++ = txdr_unsigned(VTONFS(v)->n_fhsize);
*(tl + ((t1 >> 2) - 2)) = 0;
bcopy(VTONFS(v)->n_fhp, tl, VTONFS(v)->n_fhsize);
} else {
t1 = nfsm_strtmbuf(mb, bpos,
(const char *)VTONFS(v)->n_fhp,
VTONFS(v)->n_fhsize);
if (t1 != 0)
return t1;
}
} else {
cp = nfsm_build_xx(NFSX_V2FH, mb, bpos);
bcopy(VTONFS(v)->n_fhp, cp, NFSX_V2FH);
}
return 0;
}
void
nfsm_v3attrbuild_xx(struct vattr *va, int full, struct mbuf **mb,
caddr_t *bpos)
{
u_int32_t *tl;
if (va->va_mode != (mode_t)VNOVAL) {
tl = nfsm_build_xx(2 * NFSX_UNSIGNED, mb, bpos);
*tl++ = nfs_true;
*tl = txdr_unsigned(va->va_mode);
} else {
tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
*tl = nfs_false;
}
if (full && va->va_uid != (uid_t)VNOVAL) {
tl = nfsm_build_xx(2 * NFSX_UNSIGNED, mb, bpos);
*tl++ = nfs_true;
*tl = txdr_unsigned(va->va_uid);
} else {
tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
*tl = nfs_false;
}
if (full && va->va_gid != (gid_t)VNOVAL) {
tl = nfsm_build_xx(2 * NFSX_UNSIGNED, mb, bpos);
*tl++ = nfs_true;
*tl = txdr_unsigned(va->va_gid);
} else {
tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
*tl = nfs_false;
}
if (full && va->va_size != VNOVAL) {
tl = nfsm_build_xx(3 * NFSX_UNSIGNED, mb, bpos);
*tl++ = nfs_true;
txdr_hyper(va->va_size, tl);
} else {
tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
*tl = nfs_false;
}
if (va->va_atime.tv_sec != VNOVAL) {
if (va->va_atime.tv_sec != time_second) {
tl = nfsm_build_xx(3 * NFSX_UNSIGNED, mb, bpos);
*tl++ = txdr_unsigned(NFSV3SATTRTIME_TOCLIENT);
txdr_nfsv3time(&va->va_atime, tl);
} else {
tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
*tl = txdr_unsigned(NFSV3SATTRTIME_TOSERVER);
}
} else {
tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
*tl = txdr_unsigned(NFSV3SATTRTIME_DONTCHANGE);
}
if (va->va_mtime.tv_sec != VNOVAL) {
if (va->va_mtime.tv_sec != time_second) {
tl = nfsm_build_xx(3 * NFSX_UNSIGNED, mb, bpos);
*tl++ = txdr_unsigned(NFSV3SATTRTIME_TOCLIENT);
txdr_nfsv3time(&va->va_mtime, tl);
} else {
tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
*tl = txdr_unsigned(NFSV3SATTRTIME_TOSERVER);
}
} else {
tl = nfsm_build_xx(NFSX_UNSIGNED, mb, bpos);
*tl = txdr_unsigned(NFSV3SATTRTIME_DONTCHANGE);
}
}