/* $NetBSD: nfs_clvfsops.c,v 1.3 2018/09/03 16:29:34 riastradh Exp $ */
/*-
* Copyright (c) 1989, 1993, 1995
* 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.
*
* from nfs_vfsops.c 8.12 (Berkeley) 5/20/95
*/
/* So that loader and kldload(2) can find us, wherever we are.. */
MODULE_VERSION(nfs, 1);
MODULE_DEPEND(nfs, nfscommon, 1, 1, 1);
MODULE_DEPEND(nfs, krpc, 1, 1, 1);
MODULE_DEPEND(nfs, nfssvc, 1, 1, 1);
MODULE_DEPEND(nfs, nfslock, 1, 1, 1);
/*
* This structure is now defined in sys/nfs/nfs_diskless.c so that it
* can be shared by both NFS clients. It is declared here so that it
* will be defined for kernels built without NFS_ROOT, although it
* isn't used in that case.
*/
#if !defined(NEW_NFS_BOOT)
struct nfs_diskless nfs_diskless = { { { 0 } } };
struct nfsv3_diskless nfsv3_diskless = { { { 0 } } };
int nfs_diskless_valid = 0;
#endif
SYSCTL_INT(_vfs_nfs, OID_AUTO, diskless_valid, CTLFLAG_RD,
&nfs_diskless_valid, 0,
"Has the diskless struct been filled correctly");
SYSCTL_STRING(_vfs_nfs, OID_AUTO, diskless_rootpath, CTLFLAG_RD,
nfsv3_diskless.root_hostnam, 0, "Path to nfs root");
int
newnfs_iosize(struct nfsmount *nmp)
{
int iosize, maxio;
/* First, set the upper limit for iosize */
if (nmp->nm_flag & NFSMNT_NFSV4) {
maxio = NFS_MAXBSIZE;
} else if (nmp->nm_flag & NFSMNT_NFSV3) {
if (nmp->nm_sotype == SOCK_DGRAM)
maxio = NFS_MAXDGRAMDATA;
else
maxio = NFS_MAXBSIZE;
} else {
maxio = NFS_V2MAXDATA;
}
if (nmp->nm_rsize > maxio || nmp->nm_rsize == 0)
nmp->nm_rsize = maxio;
if (nmp->nm_rsize > NFS_MAXBSIZE)
nmp->nm_rsize = NFS_MAXBSIZE;
if (nmp->nm_readdirsize > maxio || nmp->nm_readdirsize == 0)
nmp->nm_readdirsize = maxio;
if (nmp->nm_readdirsize > nmp->nm_rsize)
nmp->nm_readdirsize = nmp->nm_rsize;
if (nmp->nm_wsize > maxio || nmp->nm_wsize == 0)
nmp->nm_wsize = maxio;
if (nmp->nm_wsize > NFS_MAXBSIZE)
nmp->nm_wsize = NFS_MAXBSIZE;
/*
* Calculate the size used for io buffers. Use the larger
* of the two sizes to minimise nfs requests but make sure
* that it is at least one VM page to avoid wasting buffer
* space. It must also be at least NFS_DIRBLKSIZ, since
* that is the buffer size used for directories.
*/
iosize = imax(nmp->nm_rsize, nmp->nm_wsize);
iosize = imax(iosize, PAGE_SIZE);
iosize = imax(iosize, NFS_DIRBLKSIZ);
nmp->nm_mountp->mnt_stat.f_iosize = iosize;
return (iosize);
}
/*
* Mount a remote root fs via. nfs. This depends on the info in the
* nfs_diskless structure that has been filled in properly by some primary
* bootstrap.
* It goes something like this:
* - do enough of "ifconfig" by calling ifioctl() so that the system
* can talk to the server
* - If nfs_diskless.mygateway is filled in, use that address as
* a default gateway.
* - build the rootfs mount point and call mountnfs() to do the rest.
*
* It is assumed to be safe to read, modify, and write the nfsv3_diskless
* structure, as well as other global NFS client variables here, as
* nfs_mountroot() will be called once in the boot before any other NFS
* client activity occurs.
*/
static int
nfs_mountroot(struct mount *mp)
{
struct thread *td = curthread;
struct nfsv3_diskless *nd = &nfsv3_diskless;
struct socket *so;
struct vnode *vp;
struct ifreq ir;
int error;
u_long l;
char buf[128];
char *cp;
#if defined(NEW_NFS_BOOT) && defined(NEW_NFS_BOOT_BOOTP)
bootpc_init(); /* use bootp to get nfs_diskless filled in */
#elif defined(NEW_NFS_BOOT)
nfs_setup_diskless();
#endif
if (nfs_diskless_valid == 0)
return (-1);
if (nfs_diskless_valid == 1)
nfs_convert_diskless();
/*
* XXX splnet, so networks will receive...
*/
splnet();
/*
* Do enough of ifconfig(8) so that the critical net interface can
* talk to the server.
*/
error = socreate(nd->myif.ifra_addr.sa_family, &so, nd->root_args.sotype, 0,
td->td_ucred, td);
if (error)
panic("nfs_mountroot: socreate(%04x): %d",
nd->myif.ifra_addr.sa_family, error);
#if 0 /* XXX Bad idea */
/*
* We might not have been told the right interface, so we pass
* over the first ten interfaces of the same kind, until we get
* one of them configured.
*/
/*
* If the gateway field is filled in, set it as the default route.
* Note that pxeboot will set a default route of 0 if the route
* is not set by the DHCP server. Check also for a value of 0
* to avoid panicking inappropriately in that situation.
*/
if (nd->mygateway.sin_len != 0 &&
nd->mygateway.sin_addr.s_addr != 0) {
struct sockaddr_in mask, sin;
bzero((caddr_t)&mask, sizeof(mask));
sin = mask;
sin.sin_family = AF_INET;
sin.sin_len = sizeof(sin);
/* XXX MRT use table 0 for this sort of thing */
CURVNET_SET(TD_TO_VNET(td));
error = rtrequest_fib(RTM_ADD, (struct sockaddr *)&sin,
(struct sockaddr *)&nd->mygateway,
(struct sockaddr *)&mask,
RTF_UP | RTF_GATEWAY, NULL, RT_DEFAULT_FIB);
CURVNET_RESTORE();
if (error)
panic("nfs_mountroot: RTM_ADD: %d", error);
}
/*
* This is not really an nfs issue, but it is much easier to
* set hostname here and then let the "/etc/rc.xxx" files
* mount the right /var based upon its preset value.
*/
mtx_lock(&prison0.pr_mtx);
strlcpy(prison0.pr_hostname, nd->my_hostnam,
sizeof(prison0.pr_hostname));
mtx_unlock(&prison0.pr_mtx);
inittodr(ntohl(nd->root_time));
return (0);
}
/*
* Internal version of mount system call for diskless setup.
*/
static int
nfs_mountdiskless(char *path,
struct sockaddr_in *sin, struct nfs_args *args, struct thread *td,
struct vnode **vpp, struct mount *mp)
{
struct sockaddr *nam;
int dirlen, error;
char *dirpath;
/*
* Find the directory path in "path", which also has the server's
* name/ip address in it.
*/
dirpath = strchr(path, ':');
if (dirpath != NULL)
dirlen = strlen(++dirpath);
else
dirlen = 0;
nam = sodupsockaddr((struct sockaddr *)sin, M_WAITOK);
if ((error = mountnfs(args, mp, nam, path, NULL, 0, dirpath, dirlen,
NULL, 0, vpp, td->td_ucred, td, NFS_DEFAULT_NAMETIMEO,
NFS_DEFAULT_NEGNAMETIMEO, 0)) != 0) {
printf("nfs_mountroot: mount %s on /: %d\n", path, error);
return (error);
}
return (0);
}
static void
nfs_sec_name(char *sec, int *flagsp)
{
if (!strcmp(sec, "krb5"))
*flagsp |= NFSMNT_KERB;
else if (!strcmp(sec, "krb5i"))
*flagsp |= (NFSMNT_KERB | NFSMNT_INTEGRITY);
else if (!strcmp(sec, "krb5p"))
*flagsp |= (NFSMNT_KERB | NFSMNT_PRIVACY);
}
static void
nfs_decode_args(struct mount *mp, struct nfsmount *nmp, struct nfs_args *argp,
const char *hostname, struct ucred *cred, struct thread *td)
{
int s;
int adjsock;
char *p;
s = splnet();
/*
* Set read-only flag if requested; otherwise, clear it if this is
* an update. If this is not an update, then either the read-only
* flag is already clear, or this is a root mount and it was set
* intentionally at some previous point.
*/
if (vfs_getopt(mp->mnt_optnew, "ro", NULL, NULL) == 0) {
MNT_ILOCK(mp);
mp->mnt_flag |= MNT_RDONLY;
MNT_IUNLOCK(mp);
} else if (mp->mnt_flag & MNT_UPDATE) {
MNT_ILOCK(mp);
mp->mnt_flag &= ~MNT_RDONLY;
MNT_IUNLOCK(mp);
}
/*
* Silently clear NFSMNT_NOCONN if it's a TCP mount, it makes
* no sense in that context. Also, set up appropriate retransmit
* and soft timeout behavior.
*/
if (argp->sotype == SOCK_STREAM) {
nmp->nm_flag &= ~NFSMNT_NOCONN;
nmp->nm_timeo = NFS_MAXTIMEO;
if ((argp->flags & NFSMNT_NFSV4) != 0)
nmp->nm_retry = INT_MAX;
else
nmp->nm_retry = NFS_RETRANS_TCP;
}
/* Also clear RDIRPLUS if NFSv2, it crashes some servers */
if ((argp->flags & (NFSMNT_NFSV3 | NFSMNT_NFSV4)) == 0) {
argp->flags &= ~NFSMNT_RDIRPLUS;
nmp->nm_flag &= ~NFSMNT_RDIRPLUS;
}
/* Re-bind if rsrvd port requested and wasn't on one */
adjsock = !(nmp->nm_flag & NFSMNT_RESVPORT)
&& (argp->flags & NFSMNT_RESVPORT);
/* Also re-bind if we're switching to/from a connected UDP socket */
adjsock |= ((nmp->nm_flag & NFSMNT_NOCONN) !=
(argp->flags & NFSMNT_NOCONN));
if ((argp->flags & NFSMNT_RETRANS) && argp->retrans > 1) {
nmp->nm_retry = argp->retrans;
if (nmp->nm_retry > NFS_MAXREXMIT)
nmp->nm_retry = NFS_MAXREXMIT;
}
if ((argp->flags & NFSMNT_WSIZE) && argp->wsize > 0) {
nmp->nm_wsize = argp->wsize;
/*
* Clip at the power of 2 below the size. There is an
* issue (not isolated) that causes intermittent page
* faults if this is not done.
*/
if (nmp->nm_wsize > NFS_FABLKSIZE)
nmp->nm_wsize = 1 << (fls(nmp->nm_wsize) - 1);
else
nmp->nm_wsize = NFS_FABLKSIZE;
}
if ((argp->flags & NFSMNT_RSIZE) && argp->rsize > 0) {
nmp->nm_rsize = argp->rsize;
/*
* Clip at the power of 2 below the size. There is an
* issue (not isolated) that causes intermittent page
* faults if this is not done.
*/
if (nmp->nm_rsize > NFS_FABLKSIZE)
nmp->nm_rsize = 1 << (fls(nmp->nm_rsize) - 1);
else
nmp->nm_rsize = NFS_FABLKSIZE;
}
/*
* Parse the "from" mountarg, passed by the generic mount(8) program
* or the mountroot code. This is used when rerooting into NFS.
*
* Note that the "hostname" is actually a "hostname:/share/path" string.
*/
static int
nfs_mount_parse_from(struct vfsoptlist *opts, char **hostnamep,
struct sockaddr_in **sinp, char *dirpath, size_t dirpathsize, int *dirlenp)
{
char nam[MNAMELEN + 1];
char *delimp, *hostp, *spec;
int error, have_bracket = 0, offset, rv, speclen;
struct sockaddr_in *sin;
size_t len;
/*
* If there has been a trailing slash at mounttime it seems
* that some mountd implementations fail to remove the mount
* entries from their mountlist while unmounting.
*/
for (speclen = strlen(spec);
speclen > 1 && spec[speclen - 1] == '/';
speclen--)
spec[speclen - 1] = '\0';
if (strlen(hostp) + strlen(spec) + 1 > MNAMELEN) {
printf("%s: %s:%s: name too long", __func__, hostp, spec);
return (EINVAL);
}
/* Make both '@' and ':' notations equal */
if (*hostp != '\0') {
len = strlen(hostp);
offset = 0;
if (have_bracket)
nam[offset++] = '[';
memmove(nam + offset, hostp, len);
if (have_bracket)
nam[len + offset++] = ']';
nam[len + offset++] = ':';
memmove(nam + len + offset, spec, speclen);
nam[len + speclen + offset] = '\0';
} else
nam[0] = '\0';
/*
* The old mount_nfs program passed the struct nfs_args
* from userspace to kernel. The new mount_nfs program
* passes string options via nmount() from userspace to kernel
* and we populate the struct nfs_args in the kernel.
*/
if (vfs_getopt(mp->mnt_optnew, "nfs_args", NULL, NULL) == 0) {
error = vfs_copyopt(mp->mnt_optnew, "nfs_args", &args,
sizeof(args));
if (error != 0)
goto out;
if (mp->mnt_flag & MNT_UPDATE) {
struct nfsmount *nmp = VFSTONFS(mp);
if (nmp == NULL) {
error = EIO;
goto out;
}
/*
* If a change from TCP->UDP is done and there are thread(s)
* that have I/O RPC(s) in progress with a transfer size
* greater than NFS_MAXDGRAMDATA, those thread(s) will be
* hung, retrying the RPC(s) forever. Usually these threads
* will be seen doing an uninterruptible sleep on wait channel
* "nfsreq".
*/
if (args.sotype == SOCK_DGRAM && nmp->nm_sotype == SOCK_STREAM)
tprintf(td->td_proc, LOG_WARNING,
"Warning: mount -u that changes TCP->UDP can result in hung threads\n");
/*
* Make the nfs_ip_paranoia sysctl serve as the default connection
* or no-connection mode for those protocols that support
* no-connection mode (the flag will be cleared later for protocols
* that do not support no-connection mode). This will allow a client
* to receive replies from a different IP then the request was
* sent to. Note: default value for nfs_ip_paranoia is 1 (paranoid),
* not 0.
*/
if (nfs_ip_paranoia == 0)
args.flags |= NFSMNT_NOCONN;
if (has_nfs_args_opt != 0) {
/*
* In the 'nfs_args' case, the pointers in the args
* structure are in userland - we copy them in here.
*/
if (args.fhsize < 0 || args.fhsize > NFSX_V3FHMAX) {
vfs_mount_error(mp, "Bad file handle");
error = EINVAL;
goto out;
}
error = copyin((caddr_t)args.fh, (caddr_t)nfh,
args.fhsize);
if (error != 0)
goto out;
error = copyinstr(args.hostname, hst, MNAMELEN - 1, &hstlen);
if (error != 0)
goto out;
bzero(&hst[hstlen], MNAMELEN - hstlen);
args.hostname = hst;
/* getsockaddr() call must be after above copyin() calls */
error = getsockaddr(&nam, (caddr_t)args.addr,
args.addrlen);
if (error != 0)
goto out;
} else if (nfs_mount_parse_from(mp->mnt_optnew,
&args.hostname, (struct sockaddr_in **)&nam, dirpath,
sizeof(dirpath), &dirlen) == 0) {
has_nfs_from_opt = 1;
bcopy(args.hostname, hst, MNAMELEN);
hst[MNAMELEN - 1] = '\0';
/*
* VFS Operations.
*
* mount system call
* It seems a bit dumb to copyinstr() the host and path here and then
* bcopy() them in mountnfs(), but I wanted to detect errors before
* doing the getsockaddr() call because getsockaddr() allocates an mbuf and
* an error after that means that I have to release the mbuf.
*/
/* ARGSUSED */
static int
nfs_cmount(struct mntarg *ma, void *data, uint64_t flags)
{
int error;
struct nfs_args args;
/*
* Common code for mount and mountroot
*/
static int
mountnfs(struct nfs_args *argp, struct mount *mp, struct sockaddr *nam,
char *hst, u_char *krbname, int krbnamelen, u_char *dirpath, int dirlen,
u_char *srvkrbname, int srvkrbnamelen, struct vnode **vpp,
struct ucred *cred, struct thread *td, int nametimeo, int negnametimeo,
int minvers)
{
struct nfsmount *nmp;
struct nfsnode *np;
int error, trycnt, ret;
struct nfsvattr nfsva;
struct nfsclclient *clp;
struct nfsclds *dsp, *tdsp;
uint32_t lease;
static u_int64_t clval = 0;
NFSCL_DEBUG(3, "in mnt\n");
clp = NULL;
if (mp->mnt_flag & MNT_UPDATE) {
nmp = VFSTONFS(mp);
printf("%s: MNT_UPDATE is no longer handled here\n", __func__);
FREE(nam, M_SONAME);
return (0);
} else {
MALLOC(nmp, struct nfsmount *, sizeof (struct nfsmount) +
krbnamelen + dirlen + srvkrbnamelen + 2,
M_NEWNFSMNT, M_WAITOK | M_ZERO);
TAILQ_INIT(&nmp->nm_bufq);
if (clval == 0)
clval = (u_int64_t)nfsboottime.tv_sec;
nmp->nm_clval = clval++;
nmp->nm_krbnamelen = krbnamelen;
nmp->nm_dirpathlen = dirlen;
nmp->nm_srvkrbnamelen = srvkrbnamelen;
if (td->td_ucred->cr_uid != (uid_t)0) {
/*
* nm_uid is used to get KerberosV credentials for
* the nfsv4 state handling operations if there is
* no host based principal set. Use the uid of
* this user if not root, since they are doing the
* mount. I don't think setting this for root will
* work, since root normally does not have user
* credentials in a credentials cache.
*/
nmp->nm_uid = td->td_ucred->cr_uid;
} else {
/*
* Just set to -1, so it won't be used.
*/
nmp->nm_uid = (uid_t)-1;
}
/*
* Since nfs_decode_args() might optionally set them, these
* need to be set to defaults before the call, so that the
* optional settings aren't overwritten.
*/
nmp->nm_nametimeo = nametimeo;
nmp->nm_negnametimeo = negnametimeo;
nmp->nm_timeo = NFS_TIMEO;
nmp->nm_retry = NFS_RETRANS;
nmp->nm_readahead = NFS_DEFRAHEAD;
/* This is empirical approximation of sqrt(hibufspace) * 256. */
nmp->nm_wcommitsize = NFS_MAXBSIZE / 256;
while ((long)nmp->nm_wcommitsize * nmp->nm_wcommitsize < hibufspace)
nmp->nm_wcommitsize *= 2;
nmp->nm_wcommitsize *= 256;
/*
* V2 can only handle 32 bit filesizes. A 4GB-1 limit may be too
* high, depending on whether we end up with negative offsets in
* the client or server somewhere. 2GB-1 may be safer.
*
* For V3, ncl_fsinfo will adjust this as necessary. Assume maximum
* that we can handle until we find out otherwise.
*/
if ((argp->flags & (NFSMNT_NFSV3 | NFSMNT_NFSV4)) == 0)
nmp->nm_maxfilesize = 0xffffffffLL;
else
nmp->nm_maxfilesize = OFF_MAX;
if ((argp->flags & (NFSMNT_NFSV3 | NFSMNT_NFSV4)) == 0) {
nmp->nm_wsize = NFS_WSIZE;
nmp->nm_rsize = NFS_RSIZE;
nmp->nm_readdirsize = NFS_READDIRSIZE;
}
nmp->nm_numgrps = NFS_MAXGRPS;
nmp->nm_tprintf_delay = nfs_tprintf_delay;
if (nmp->nm_tprintf_delay < 0)
nmp->nm_tprintf_delay = 0;
nmp->nm_tprintf_initial_delay = nfs_tprintf_initial_delay;
if (nmp->nm_tprintf_initial_delay < 0)
nmp->nm_tprintf_initial_delay = 0;
nmp->nm_fhsize = argp->fhsize;
if (nmp->nm_fhsize > 0)
bcopy((caddr_t)argp->fh, (caddr_t)nmp->nm_fh, argp->fhsize);
bcopy(hst, mp->mnt_stat.f_mntfromname, MNAMELEN);
nmp->nm_nam = nam;
/* Set up the sockets and per-host congestion */
nmp->nm_sotype = argp->sotype;
nmp->nm_soproto = argp->proto;
nmp->nm_sockreq.nr_prog = NFS_PROG;
if ((argp->flags & NFSMNT_NFSV4))
nmp->nm_sockreq.nr_vers = NFS_VER4;
else if ((argp->flags & NFSMNT_NFSV3))
nmp->nm_sockreq.nr_vers = NFS_VER3;
else
nmp->nm_sockreq.nr_vers = NFS_VER2;
if ((error = newnfs_connect(nmp, &nmp->nm_sockreq, cred, td, 0)))
goto bad;
/* For NFSv4.1, get the clientid now. */
if (nmp->nm_minorvers > 0) {
NFSCL_DEBUG(3, "at getcl\n");
error = nfscl_getcl(mp, cred, td, 0, &clp);
NFSCL_DEBUG(3, "aft getcl=%d\n", error);
if (error != 0)
goto bad;
}
if (nmp->nm_fhsize == 0 && (nmp->nm_flag & NFSMNT_NFSV4) &&
nmp->nm_dirpathlen > 0) {
NFSCL_DEBUG(3, "in dirp\n");
/*
* If the fhsize on the mount point == 0 for V4, the mount
* path needs to be looked up.
*/
trycnt = 3;
do {
error = nfsrpc_getdirpath(nmp, NFSMNT_DIRPATH(nmp),
cred, td);
NFSCL_DEBUG(3, "aft dirp=%d\n", error);
if (error)
(void) nfs_catnap(PZERO, error, "nfsgetdirp");
} while (error && --trycnt > 0);
if (error) {
error = nfscl_maperr(td, error, (uid_t)0, (gid_t)0);
goto bad;
}
}
/*
* A reference count is needed on the nfsnode representing the
* remote root. If this object is not persistent, then backward
* traversals of the mount point (i.e. "..") will not work if
* the nfsnode gets flushed out of the cache. Ufs does not have
* this problem, because one can identify root inodes by their
* number == ROOTINO (2).
*/
if (nmp->nm_fhsize > 0) {
/*
* Set f_iosize to NFS_DIRBLKSIZ so that bo_bsize gets set
* non-zero for the root vnode. f_iosize will be set correctly
* by nfs_statfs() before any I/O occurs.
*/
mp->mnt_stat.f_iosize = NFS_DIRBLKSIZ;
error = ncl_nget(mp, nmp->nm_fh, nmp->nm_fhsize, &np,
LK_EXCLUSIVE);
if (error)
goto bad;
*vpp = NFSTOV(np);
/*
* Get file attributes and transfer parameters for the
* mountpoint. This has the side effect of filling in
* (*vpp)->v_type with the correct value.
*/
ret = nfsrpc_getattrnovp(nmp, nmp->nm_fh, nmp->nm_fhsize, 1,
cred, td, &nfsva, NULL, &lease);
if (ret) {
/*
* Just set default values to get things going.
*/
NFSBZERO((caddr_t)&nfsva, sizeof (struct nfsvattr));
nfsva.na_vattr.va_type = VDIR;
nfsva.na_vattr.va_mode = 0777;
nfsva.na_vattr.va_nlink = 100;
nfsva.na_vattr.va_uid = (uid_t)0;
nfsva.na_vattr.va_gid = (gid_t)0;
nfsva.na_vattr.va_fileid = 2;
nfsva.na_vattr.va_gen = 1;
nfsva.na_vattr.va_blocksize = NFS_FABLKSIZE;
nfsva.na_vattr.va_size = 512 * 1024;
lease = 60;
}
(void) nfscl_loadattrcache(vpp, &nfsva, NULL, NULL, 0, 1);
if (nmp->nm_minorvers > 0) {
NFSCL_DEBUG(3, "lease=%d\n", (int)lease);
NFSLOCKCLSTATE();
clp->nfsc_renew = NFSCL_RENEW(lease);
clp->nfsc_expire = NFSD_MONOSEC + clp->nfsc_renew;
clp->nfsc_clientidrev++;
if (clp->nfsc_clientidrev == 0)
clp->nfsc_clientidrev++;
NFSUNLOCKCLSTATE();
/*
* Mount will succeed, so the renew thread can be
* started now.
*/
nfscl_start_renewthread(clp);
nfscl_clientrelease(clp);
}
if (argp->flags & NFSMNT_NFSV3)
ncl_fsinfo(nmp, *vpp, cred, td);
/* Mark if the mount point supports NFSv4 ACLs. */
if ((argp->flags & NFSMNT_NFSV4) != 0 && nfsrv_useacl != 0 &&
ret == 0 &&
NFSISSET_ATTRBIT(&nfsva.na_suppattr, NFSATTRBIT_ACL)) {
MNT_ILOCK(mp);
mp->mnt_flag |= MNT_NFS4ACLS;
MNT_IUNLOCK(mp);
}
/*
* Lose the lock but keep the ref.
*/
NFSVOPUNLOCK(*vpp, 0);
return (0);
}
error = EIO;
/*
* unmount system call
*/
static int
nfs_unmount(struct mount *mp, int mntflags)
{
struct thread *td;
struct nfsmount *nmp;
int error, flags = 0, i, trycnt = 0;
struct nfsclds *dsp, *tdsp;
td = curthread;
if (mntflags & MNT_FORCE)
flags |= FORCECLOSE;
nmp = VFSTONFS(mp);
/*
* Goes something like this..
* - Call vflush() to clear out vnodes for this filesystem
* - Close the socket
* - Free up the data structures
*/
/* In the forced case, cancel any outstanding requests. */
if (mntflags & MNT_FORCE) {
error = newnfs_nmcancelreqs(nmp);
if (error)
goto out;
/* For a forced close, get rid of the renew thread now */
nfscl_umount(nmp, td);
}
/* We hold 1 extra ref on the root vnode; see comment in mountnfs(). */
do {
error = vflush(mp, 1, flags, td);
if ((mntflags & MNT_FORCE) && error != 0 && ++trycnt < 30)
(void) nfs_catnap(PSOCK, error, "newndm");
} while ((mntflags & MNT_FORCE) && error != 0 && trycnt < 30);
if (error)
goto out;
/*
* We are now committed to the unmount.
*/
if ((mntflags & MNT_FORCE) == 0)
nfscl_umount(nmp, td);
/* Make sure no nfsiods are assigned to this mount. */
mtx_lock(&ncl_iod_mutex);
for (i = 0; i < NFS_MAXASYNCDAEMON; i++)
if (ncl_iodmount[i] == nmp) {
ncl_iodwant[i] = NFSIOD_AVAILABLE;
ncl_iodmount[i] = NULL;
}
mtx_unlock(&ncl_iod_mutex);
newnfs_disconnect(&nmp->nm_sockreq);
crfree(nmp->nm_sockreq.nr_cred);
FREE(nmp->nm_nam, M_SONAME);
if (nmp->nm_sockreq.nr_auth != NULL)
AUTH_DESTROY(nmp->nm_sockreq.nr_auth);
mtx_destroy(&nmp->nm_sockreq.nr_mtx);
mtx_destroy(&nmp->nm_mtx);
TAILQ_FOREACH_SAFE(dsp, &nmp->nm_sess, nfsclds_list, tdsp)
nfscl_freenfsclds(dsp);
FREE(nmp, M_NEWNFSMNT);
out:
return (error);
}
/*
* Return root of a filesystem
*/
static int
nfs_root(struct mount *mp, int flags, struct vnode **vpp)
{
struct vnode *vp;
struct nfsmount *nmp;
struct nfsnode *np;
int error;
nmp = VFSTONFS(mp);
error = ncl_nget(mp, nmp->nm_fh, nmp->nm_fhsize, &np, flags);
if (error)
return error;
vp = NFSTOV(np);
/*
* Get transfer parameters and attributes for root vnode once.
*/
mtx_lock(&nmp->nm_mtx);
if (NFSHASNFSV3(nmp) && !NFSHASGOTFSINFO(nmp)) {
mtx_unlock(&nmp->nm_mtx);
ncl_fsinfo(nmp, vp, curthread->td_ucred, curthread);
} else
mtx_unlock(&nmp->nm_mtx);
if (vp->v_type == VNON)
vp->v_type = VDIR;
vp->v_vflag |= VV_ROOT;
*vpp = vp;
return (0);
}
/*
* Flush out the buffer cache
*/
/* ARGSUSED */
static int
nfs_sync(struct mount *mp, int waitfor)
{
struct vnode *vp, *mvp;
struct thread *td;
int error, allerror = 0;
td = curthread;
MNT_ILOCK(mp);
/*
* If a forced dismount is in progress, return from here so that
* the umount(2) syscall doesn't get stuck in VFS_SYNC() before
* calling VFS_UNMOUNT().
*/
if ((mp->mnt_kern_flag & MNTK_UNMOUNTF) != 0) {
MNT_IUNLOCK(mp);
return (EBADF);
}
MNT_IUNLOCK(mp);
/*
* Force stale buffer cache information to be flushed.
*/
loop:
MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
/* XXX Racy bv_cnt check. */
if (NFSVOPISLOCKED(vp) || vp->v_bufobj.bo_dirty.bv_cnt == 0 ||
waitfor == MNT_LAZY) {
VI_UNLOCK(vp);
continue;
}
if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td)) {
MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
goto loop;
}
error = VOP_FSYNC(vp, waitfor, td);
if (error)
allerror = error;
NFSVOPUNLOCK(vp, 0);
vrele(vp);
}
return (allerror);
}
static int
nfs_sysctl(struct mount *mp, fsctlop_t op, struct sysctl_req *req)
{
struct nfsmount *nmp = VFSTONFS(mp);
struct vfsquery vq;
int error;
bzero(&vq, sizeof(vq));
switch (op) {
#if 0
case VFS_CTL_NOLOCKS:
val = (nmp->nm_flag & NFSMNT_NOLOCKS) ? 1 : 0;
if (req->oldptr != NULL) {
error = SYSCTL_OUT(req, &val, sizeof(val));
if (error)
return (error);
}
if (req->newptr != NULL) {
error = SYSCTL_IN(req, &val, sizeof(val));
if (error)
return (error);
if (val)
nmp->nm_flag |= NFSMNT_NOLOCKS;
else
nmp->nm_flag &= ~NFSMNT_NOLOCKS;
}
break;
#endif
case VFS_CTL_QUERY:
mtx_lock(&nmp->nm_mtx);
if (nmp->nm_state & NFSSTA_TIMEO)
vq.vq_flags |= VQ_NOTRESP;
mtx_unlock(&nmp->nm_mtx);
#if 0
if (!(nmp->nm_flag & NFSMNT_NOLOCKS) &&
(nmp->nm_state & NFSSTA_LOCKTIMEO))
vq.vq_flags |= VQ_NOTRESPLOCK;
#endif
error = SYSCTL_OUT(req, &vq, sizeof(vq));
break;
case VFS_CTL_TIMEO:
if (req->oldptr != NULL) {
error = SYSCTL_OUT(req, &nmp->nm_tprintf_initial_delay,
sizeof(nmp->nm_tprintf_initial_delay));
if (error)
return (error);
}
if (req->newptr != NULL) {
error = vfs_suser(mp, req->td);
if (error)
return (error);
error = SYSCTL_IN(req, &nmp->nm_tprintf_initial_delay,
sizeof(nmp->nm_tprintf_initial_delay));
if (error)
return (error);
if (nmp->nm_tprintf_initial_delay < 0)
nmp->nm_tprintf_initial_delay = 0;
}
break;
default:
return (ENOTSUP);
}
return (0);
}
/*
* Purge any RPCs in progress, so that they will all return errors.
* This allows dounmount() to continue as far as VFS_UNMOUNT() for a
* forced dismount.
*/
static void
nfs_purge(struct mount *mp)
{
struct nfsmount *nmp = VFSTONFS(mp);
newnfs_nmcancelreqs(nmp);
}
/*
* Extract the information needed by the nlm from the nfs vnode.
*/
static void
nfs_getnlminfo(struct vnode *vp, uint8_t *fhp, size_t *fhlenp,
struct sockaddr_storage *sp, int *is_v3p, off_t *sizep,
struct timeval *timeop)
{
struct nfsmount *nmp;
struct nfsnode *np = VTONFS(vp);
nmp = VFSTONFS(vp->v_mount);
if (fhlenp != NULL)
*fhlenp = (size_t)np->n_fhp->nfh_len;
if (fhp != NULL)
bcopy(np->n_fhp->nfh_fh, fhp, np->n_fhp->nfh_len);
if (sp != NULL)
bcopy(nmp->nm_nam, sp, uimin(nmp->nm_nam->sa_len, sizeof(*sp)));
if (is_v3p != NULL)
*is_v3p = NFS_ISV3(vp);
if (sizep != NULL)
*sizep = np->n_size;
if (timeop != NULL) {
timeop->tv_sec = nmp->nm_timeo / NFS_HZ;
timeop->tv_usec = (nmp->nm_timeo % NFS_HZ) * (1000000 / NFS_HZ);
}
}
/*
* This function prints out an option name, based on the conditional
* argument.
*/
static __inline void nfscl_printopt(struct nfsmount *nmp, int testval,
char *opt, char **buf, size_t *blen)
{
int len;
if (testval != 0 && *blen > strlen(opt)) {
len = snprintf(*buf, *blen, "%s", opt);
if (len != strlen(opt))
printf("EEK!!\n");
*buf += len;
*blen -= len;
}
}
/*
* This function printf out an options integer value.
*/
static __inline void nfscl_printoptval(struct nfsmount *nmp, int optval,
char *opt, char **buf, size_t *blen)
{
int len;
if (*blen > strlen(opt) + 1) {
/* Could result in truncated output string. */
len = snprintf(*buf, *blen, "%s=%d", opt, optval);
if (len < *blen) {
*buf += len;
*blen -= len;
}
}
}
/*
* Load the option flags and values into the buffer.
*/
void nfscl_retopts(struct nfsmount *nmp, char *buffer, size_t buflen)
{
char *buf;
size_t blen;
