static int
unionfs_statfs(struct mount *mp, struct statfs *sbp)
{
struct unionfs_mount *ump;
int error;
struct statfs *mstat;
uint64_t lbsize;
ump = MOUNTTOUNIONFSMOUNT(mp);
UNIONFSDEBUG("unionfs_statfs(mp = %p, lvp = %p, uvp = %p)\n",
(void *)mp, (void *)ump->um_lowervp, (void *)ump->um_uppervp);
mstat = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK | M_ZERO);
error = VFS_STATFS(ump->um_lowervp->v_mount, mstat);
if (error) {
free(mstat, M_STATFS);
return (error);
}
/* now copy across the "interesting" information and fake the rest */
sbp->f_blocks = mstat->f_blocks;
sbp->f_files = mstat->f_files;
lbsize = mstat->f_bsize;
error = VFS_STATFS(ump->um_uppervp->v_mount, mstat);
if (error) {
free(mstat, M_STATFS);
return (error);
}
/*
* The FS type etc is copy from upper vfs.
* (write able vfs have priority)
*/
sbp->f_type = mstat->f_type;
sbp->f_flags = mstat->f_flags;
sbp->f_bsize = mstat->f_bsize;
sbp->f_iosize = mstat->f_iosize;
if (mstat->f_bsize != lbsize)
sbp->f_blocks = ((off_t)sbp->f_blocks * lbsize) /
mstat->f_bsize;
sbp->f_blocks += mstat->f_blocks;
sbp->f_bfree = mstat->f_bfree;
sbp->f_bavail = mstat->f_bavail;
sbp->f_files += mstat->f_files;
sbp->f_ffree = mstat->f_ffree;
free(mstat, M_STATFS);
return (0);
}
/*
* If the VFS does not implement statvfs, then call statfs and convert
* the values. This code was taken from libc's __cvtstatvfs() function,
* contributed by Joerg Sonnenberger.
*/
int
vfs_stdstatvfs(struct mount *mp, struct statvfs *sbp, struct ucred *cred)
{
struct statfs *in;
int error;
in = &mp->mnt_stat;
error = VFS_STATFS(mp, in, cred);
if (error == 0) {
bzero(sbp, sizeof(*sbp));
sbp->f_bsize = in->f_bsize;
sbp->f_frsize = in->f_bsize;
sbp->f_blocks = in->f_blocks;
sbp->f_bfree = in->f_bfree;
sbp->f_bavail = in->f_bavail;
sbp->f_files = in->f_files;
sbp->f_ffree = in->f_ffree;
/*
* XXX
* This field counts the number of available inodes to non-root
* users, but this information is not available via statfs.
* Just ignore this issue by returning the total number
* instead.
*/
sbp->f_favail = in->f_ffree;
/*
* XXX
* This field has a different meaning for statfs and statvfs.
* For the former it is the cookie exported for NFS and not
* intended for normal userland use.
*/
sbp->f_fsid = 0;
sbp->f_flag = 0;
if (in->f_flags & MNT_RDONLY)
sbp->f_flag |= ST_RDONLY;
if (in->f_flags & MNT_NOSUID)
sbp->f_flag |= ST_NOSUID;
sbp->f_namemax = 0;
sbp->f_owner = in->f_owner;
/*
* XXX
* statfs contains the type as string, statvfs expects it as
* enumeration.
*/
sbp->f_type = 0;
sbp->f_syncreads = in->f_syncreads;
sbp->f_syncwrites = in->f_syncwrites;
sbp->f_asyncreads = in->f_asyncreads;
sbp->f_asyncwrites = in->f_asyncwrites;
}
return (error);
}
static OSKIT_COMDECL filesystem_statfs(oskit_filesystem_t *f,
oskit_statfs_t *out_stats)
{
struct gfilesystem *fs = (struct gfilesystem *) f;
struct statfs *sp;
struct mount *mp;
struct proc *p;
oskit_error_t ferror;
int error;
if (!fs || !fs->count || !fs->mp)
return OSKIT_E_INVALIDARG;
ferror = getproc(&p);
if (ferror)
return ferror;
mp = fs->mp;
sp = &mp->mnt_stat;
error = VFS_STATFS(fs->mp, sp, p);
prfree(p);
if (error)
return errno_to_oskit_error(error);
sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
out_stats->flag = 0;
if (sp->f_flags & MNT_RDONLY)
out_stats->flag |= OSKIT_FS_RDONLY;
if (sp->f_flags & MNT_NOEXEC)
out_stats->flag |= OSKIT_FS_NOEXEC;
if (sp->f_flags & MNT_NOSUID)
out_stats->flag |= OSKIT_FS_NOSUID;
if (sp->f_flags & MNT_NODEV)
out_stats->flag |= OSKIT_FS_NODEV;
out_stats->bsize = sp->f_bsize;
out_stats->frsize = sp->f_bsize;
out_stats->blocks = sp->f_blocks;
out_stats->bfree = sp->f_bfree;
out_stats->bavail = sp->f_bavail;
out_stats->files = sp->f_files;
out_stats->ffree = sp->f_ffree;
out_stats->favail = sp->f_ffree;
out_stats->fsid = sp->f_fsid.val[0]; /* device number */
out_stats->namemax = NAME_MAX;
return 0;
}
static int
hammer_vfs_mount(struct mount *mp, char *mntpt, caddr_t data,
struct ucred *cred)
{
struct hammer_mount_info info;
hammer_mount_t hmp;
hammer_volume_t rootvol;
struct vnode *rootvp;
struct vnode *devvp = NULL;
const char *upath; /* volume name in userspace */
char *path; /* volume name in system space */
int error;
int i;
int master_id;
char *next_volume_ptr = NULL;
/*
* Accept hammer_mount_info. mntpt is NULL for root mounts at boot.
*/
if (mntpt == NULL) {
bzero(&info, sizeof(info));
info.asof = 0;
info.hflags = 0;
info.nvolumes = 1;
next_volume_ptr = mp->mnt_stat.f_mntfromname;
/* Count number of volumes separated by ':' */
for (char *p = next_volume_ptr; *p != '\0'; ++p) {
if (*p == ':') {
++info.nvolumes;
}
}
mp->mnt_flag &= ~MNT_RDONLY; /* mount R/W */
} else {
if ((error = copyin(data, &info, sizeof(info))) != 0)
return (error);
}
/*
* updating or new mount
*/
if (mp->mnt_flag & MNT_UPDATE) {
hmp = (void *)mp->mnt_data;
KKASSERT(hmp != NULL);
} else {
if (info.nvolumes <= 0 || info.nvolumes > HAMMER_MAX_VOLUMES)
return (EINVAL);
hmp = NULL;
}
/*
* master-id validation. The master id may not be changed by a
* mount update.
*/
if (info.hflags & HMNT_MASTERID) {
if (hmp && hmp->master_id != info.master_id) {
kprintf("hammer: cannot change master id "
"with mount update\n");
return(EINVAL);
}
master_id = info.master_id;
if (master_id < -1 || master_id >= HAMMER_MAX_MASTERS)
return (EINVAL);
} else {
if (hmp)
master_id = hmp->master_id;
else
master_id = 0;
}
/*
* Internal mount data structure
*/
if (hmp == NULL) {
hmp = kmalloc(sizeof(*hmp), M_HAMMER, M_WAITOK | M_ZERO);
mp->mnt_data = (qaddr_t)hmp;
hmp->mp = mp;
/*TAILQ_INIT(&hmp->recycle_list);*/
/*
* Make sure kmalloc type limits are set appropriately.
*
* Our inode kmalloc group is sized based on maxvnodes
* (controlled by the system, not us).
*/
kmalloc_create(&hmp->m_misc, "HAMMER-others");
kmalloc_create(&hmp->m_inodes, "HAMMER-inodes");
kmalloc_raise_limit(hmp->m_inodes, 0); /* unlimited */
hmp->root_btree_beg.localization = 0x00000000U;
hmp->root_btree_beg.obj_id = -0x8000000000000000LL;
hmp->root_btree_beg.key = -0x8000000000000000LL;
hmp->root_btree_beg.create_tid = 1;
hmp->root_btree_beg.delete_tid = 1;
hmp->root_btree_beg.rec_type = 0;
hmp->root_btree_beg.obj_type = 0;
hmp->root_btree_end.localization = 0xFFFFFFFFU;
hmp->root_btree_end.obj_id = 0x7FFFFFFFFFFFFFFFLL;
hmp->root_btree_end.key = 0x7FFFFFFFFFFFFFFFLL;
hmp->root_btree_end.create_tid = 0xFFFFFFFFFFFFFFFFULL;
hmp->root_btree_end.delete_tid = 0; /* special case */
hmp->root_btree_end.rec_type = 0xFFFFU;
hmp->root_btree_end.obj_type = 0;
hmp->krate.freq = 1; /* maximum reporting rate (hz) */
hmp->krate.count = -16; /* initial burst */
hmp->sync_lock.refs = 1;
hmp->free_lock.refs = 1;
hmp->undo_lock.refs = 1;
hmp->blkmap_lock.refs = 1;
hmp->snapshot_lock.refs = 1;
hmp->volume_lock.refs = 1;
TAILQ_INIT(&hmp->delay_list);
TAILQ_INIT(&hmp->flush_group_list);
TAILQ_INIT(&hmp->objid_cache_list);
TAILQ_INIT(&hmp->undo_lru_list);
TAILQ_INIT(&hmp->reclaim_list);
RB_INIT(&hmp->rb_dedup_crc_root);
RB_INIT(&hmp->rb_dedup_off_root);
TAILQ_INIT(&hmp->dedup_lru_list);
}
hmp->hflags &= ~HMNT_USERFLAGS;
hmp->hflags |= info.hflags & HMNT_USERFLAGS;
hmp->master_id = master_id;
if (info.asof) {
mp->mnt_flag |= MNT_RDONLY;
hmp->asof = info.asof;
} else {
hmp->asof = HAMMER_MAX_TID;
}
hmp->volume_to_remove = -1;
/*
* Re-open read-write if originally read-only, or vise-versa.
*
* When going from read-only to read-write execute the stage2
* recovery if it has not already been run.
*/
if (mp->mnt_flag & MNT_UPDATE) {
lwkt_gettoken(&hmp->fs_token);
error = 0;
if (hmp->ronly && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
kprintf("HAMMER read-only -> read-write\n");
hmp->ronly = 0;
RB_SCAN(hammer_vol_rb_tree, &hmp->rb_vols_root, NULL,
hammer_adjust_volume_mode, NULL);
rootvol = hammer_get_root_volume(hmp, &error);
if (rootvol) {
hammer_recover_flush_buffers(hmp, rootvol, 1);
error = hammer_recover_stage2(hmp, rootvol);
bcopy(rootvol->ondisk->vol0_blockmap,
hmp->blockmap,
sizeof(hmp->blockmap));
hammer_rel_volume(rootvol, 0);
}
RB_SCAN(hammer_ino_rb_tree, &hmp->rb_inos_root, NULL,
hammer_reload_inode, NULL);
/* kernel clears MNT_RDONLY */
} else if (hmp->ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
kprintf("HAMMER read-write -> read-only\n");
hmp->ronly = 1; /* messy */
RB_SCAN(hammer_ino_rb_tree, &hmp->rb_inos_root, NULL,
hammer_reload_inode, NULL);
hmp->ronly = 0;
hammer_flusher_sync(hmp);
hammer_flusher_sync(hmp);
hammer_flusher_sync(hmp);
hmp->ronly = 1;
RB_SCAN(hammer_vol_rb_tree, &hmp->rb_vols_root, NULL,
hammer_adjust_volume_mode, NULL);
}
lwkt_reltoken(&hmp->fs_token);
return(error);
}
RB_INIT(&hmp->rb_vols_root);
RB_INIT(&hmp->rb_inos_root);
RB_INIT(&hmp->rb_redo_root);
RB_INIT(&hmp->rb_nods_root);
RB_INIT(&hmp->rb_undo_root);
RB_INIT(&hmp->rb_resv_root);
RB_INIT(&hmp->rb_bufs_root);
RB_INIT(&hmp->rb_pfsm_root);
hmp->ronly = ((mp->mnt_flag & MNT_RDONLY) != 0);
RB_INIT(&hmp->volu_root);
RB_INIT(&hmp->undo_root);
RB_INIT(&hmp->data_root);
RB_INIT(&hmp->meta_root);
RB_INIT(&hmp->lose_root);
TAILQ_INIT(&hmp->iorun_list);
lwkt_token_init(&hmp->fs_token, "hammerfs");
lwkt_token_init(&hmp->io_token, "hammerio");
lwkt_gettoken(&hmp->fs_token);
/*
* Load volumes
*/
path = objcache_get(namei_oc, M_WAITOK);
hmp->nvolumes = -1;
for (i = 0; i < info.nvolumes; ++i) {
if (mntpt == NULL) {
/*
* Root mount.
*/
KKASSERT(next_volume_ptr != NULL);
strcpy(path, "");
if (*next_volume_ptr != '/') {
/* relative path */
strcpy(path, "/dev/");
}
int k;
for (k = strlen(path); k < MAXPATHLEN-1; ++k) {
if (*next_volume_ptr == '\0') {
break;
} else if (*next_volume_ptr == ':') {
++next_volume_ptr;
break;
} else {
path[k] = *next_volume_ptr;
++next_volume_ptr;
}
}
path[k] = '\0';
error = 0;
cdev_t dev = kgetdiskbyname(path);
error = bdevvp(dev, &devvp);
if (error) {
kprintf("hammer_mountroot: can't find devvp\n");
}
} else {
error = copyin(&info.volumes[i], &upath,
sizeof(char *));
if (error == 0)
error = copyinstr(upath, path,
MAXPATHLEN, NULL);
}
if (error == 0)
error = hammer_install_volume(hmp, path, devvp);
if (error)
break;
}
objcache_put(namei_oc, path);
/*
* Make sure we found a root volume
*/
if (error == 0 && hmp->rootvol == NULL) {
kprintf("hammer_mount: No root volume found!\n");
error = EINVAL;
}
/*
* Check that all required volumes are available
*/
if (error == 0 && hammer_mountcheck_volumes(hmp)) {
kprintf("hammer_mount: Missing volumes, cannot mount!\n");
error = EINVAL;
}
if (error) {
/* called with fs_token held */
hammer_free_hmp(mp);
return (error);
}
/*
* No errors, setup enough of the mount point so we can lookup the
* root vnode.
*/
mp->mnt_iosize_max = MAXPHYS;
mp->mnt_kern_flag |= MNTK_FSMID;
mp->mnt_kern_flag |= MNTK_THR_SYNC; /* new vsyncscan semantics */
/*
* MPSAFE code. Note that VOPs and VFSops which are not MPSAFE
* will acquire a per-mount token prior to entry and release it
* on return, so even if we do not specify it we no longer get
* the BGL regardlless of how we are flagged.
*/
mp->mnt_kern_flag |= MNTK_ALL_MPSAFE;
/*MNTK_RD_MPSAFE | MNTK_GA_MPSAFE | MNTK_IN_MPSAFE;*/
/*
* note: f_iosize is used by vnode_pager_haspage() when constructing
* its VOP_BMAP call.
*/
mp->mnt_stat.f_iosize = HAMMER_BUFSIZE;
mp->mnt_stat.f_bsize = HAMMER_BUFSIZE;
mp->mnt_vstat.f_frsize = HAMMER_BUFSIZE;
mp->mnt_vstat.f_bsize = HAMMER_BUFSIZE;
mp->mnt_maxsymlinklen = 255;
mp->mnt_flag |= MNT_LOCAL;
vfs_add_vnodeops(mp, &hammer_vnode_vops, &mp->mnt_vn_norm_ops);
vfs_add_vnodeops(mp, &hammer_spec_vops, &mp->mnt_vn_spec_ops);
vfs_add_vnodeops(mp, &hammer_fifo_vops, &mp->mnt_vn_fifo_ops);
/*
* The root volume's ondisk pointer is only valid if we hold a
* reference to it.
*/
rootvol = hammer_get_root_volume(hmp, &error);
if (error)
goto failed;
/*
* Perform any necessary UNDO operations. The recovery code does
* call hammer_undo_lookup() so we have to pre-cache the blockmap,
* and then re-copy it again after recovery is complete.
*
* If this is a read-only mount the UNDO information is retained
* in memory in the form of dirty buffer cache buffers, and not
* written back to the media.
*/
bcopy(rootvol->ondisk->vol0_blockmap, hmp->blockmap,
sizeof(hmp->blockmap));
/*
* Check filesystem version
*/
hmp->version = rootvol->ondisk->vol_version;
if (hmp->version < HAMMER_VOL_VERSION_MIN ||
hmp->version > HAMMER_VOL_VERSION_MAX) {
kprintf("HAMMER: mount unsupported fs version %d\n",
hmp->version);
error = ERANGE;
goto done;
}
/*
* The undo_rec_limit limits the size of flush groups to avoid
* blowing out the UNDO FIFO. This calculation is typically in
* the tens of thousands and is designed primarily when small
* HAMMER filesystems are created.
*/
hmp->undo_rec_limit = hammer_undo_max(hmp) / 8192 + 100;
if (hammer_debug_general & 0x0001)
kprintf("HAMMER: undo_rec_limit %d\n", hmp->undo_rec_limit);
/*
* NOTE: Recover stage1 not only handles meta-data recovery, it
* also sets hmp->undo_seqno for HAMMER VERSION 4+ filesystems.
*/
error = hammer_recover_stage1(hmp, rootvol);
if (error) {
kprintf("Failed to recover HAMMER filesystem on mount\n");
goto done;
}
/*
* Finish setup now that we have a good root volume.
*
* The top 16 bits of fsid.val[1] is a pfs id.
*/
ksnprintf(mp->mnt_stat.f_mntfromname,
sizeof(mp->mnt_stat.f_mntfromname), "%s",
rootvol->ondisk->vol_name);
mp->mnt_stat.f_fsid.val[0] =
crc32((char *)&rootvol->ondisk->vol_fsid + 0, 8);
mp->mnt_stat.f_fsid.val[1] =
crc32((char *)&rootvol->ondisk->vol_fsid + 8, 8);
mp->mnt_stat.f_fsid.val[1] &= 0x0000FFFF;
mp->mnt_vstat.f_fsid_uuid = rootvol->ondisk->vol_fsid;
mp->mnt_vstat.f_fsid = crc32(&mp->mnt_vstat.f_fsid_uuid,
sizeof(mp->mnt_vstat.f_fsid_uuid));
/*
* Certain often-modified fields in the root volume are cached in
* the hammer_mount structure so we do not have to generate lots
* of little UNDO structures for them.
*
* Recopy after recovery. This also has the side effect of
* setting our cached undo FIFO's first_offset, which serves to
* placemark the FIFO start for the NEXT flush cycle while the
* on-disk first_offset represents the LAST flush cycle.
*/
hmp->next_tid = rootvol->ondisk->vol0_next_tid;
hmp->flush_tid1 = hmp->next_tid;
hmp->flush_tid2 = hmp->next_tid;
bcopy(rootvol->ondisk->vol0_blockmap, hmp->blockmap,
sizeof(hmp->blockmap));
hmp->copy_stat_freebigblocks = rootvol->ondisk->vol0_stat_freebigblocks;
hammer_flusher_create(hmp);
/*
* Locate the root directory using the root cluster's B-Tree as a
* starting point. The root directory uses an obj_id of 1.
*
* FUTURE: Leave the root directory cached referenced but unlocked
* in hmp->rootvp (need to flush it on unmount).
*/
error = hammer_vfs_vget(mp, NULL, 1, &rootvp);
if (error)
goto done;
vput(rootvp);
/*vn_unlock(hmp->rootvp);*/
if (hmp->ronly == 0)
error = hammer_recover_stage2(hmp, rootvol);
/*
* If the stage2 recovery fails be sure to clean out all cached
* vnodes before throwing away the mount structure or bad things
* will happen.
*/
if (error)
vflush(mp, 0, 0);
done:
if ((mp->mnt_flag & MNT_UPDATE) == 0) {
/* New mount */
/* Populate info for mount point (NULL pad)*/
bzero(mp->mnt_stat.f_mntonname, MNAMELEN);
size_t size;
if (mntpt) {
copyinstr(mntpt, mp->mnt_stat.f_mntonname,
MNAMELEN -1, &size);
} else { /* Root mount */
mp->mnt_stat.f_mntonname[0] = '/';
}
}
(void)VFS_STATFS(mp, &mp->mnt_stat, cred);
hammer_rel_volume(rootvol, 0);
failed:
/*
* Cleanup and return.
*/
if (error) {
/* called with fs_token held */
hammer_free_hmp(mp);
} else {
lwkt_reltoken(&hmp->fs_token);
}
return (error);
}
/*
* VFS Operations.
*
* mount system call
*/
int
ffs_mount(struct mount *mp, const char *path, void *data,
struct nameidata *ndp, struct proc *p)
{
struct vnode *devvp;
struct ufs_args args;
struct ufsmount *ump = NULL;
struct fs *fs;
int error = 0, flags;
int ronly;
mode_t accessmode;
size_t size;
error = copyin(data, &args, sizeof (struct ufs_args));
if (error)
return (error);
#ifndef FFS_SOFTUPDATES
if (mp->mnt_flag & MNT_SOFTDEP) {
printf("WARNING: soft updates isn't compiled in\n");
mp->mnt_flag &= ~MNT_SOFTDEP;
}
#endif
/*
* Soft updates is incompatible with "async",
* so if we are doing softupdates stop the user
* from setting the async flag.
*/
if ((mp->mnt_flag & (MNT_SOFTDEP | MNT_ASYNC)) ==
(MNT_SOFTDEP | MNT_ASYNC)) {
return (EINVAL);
}
/*
* If updating, check whether changing from read-only to
* read/write; if there is no device name, that's all we do.
*/
if (mp->mnt_flag & MNT_UPDATE) {
ump = VFSTOUFS(mp);
fs = ump->um_fs;
devvp = ump->um_devvp;
error = 0;
ronly = fs->fs_ronly;
if (ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
/* Flush any dirty data */
mp->mnt_flag &= ~MNT_RDONLY;
VFS_SYNC(mp, MNT_WAIT, p->p_ucred, p);
mp->mnt_flag |= MNT_RDONLY;
/*
* Get rid of files open for writing.
*/
flags = WRITECLOSE;
if (mp->mnt_flag & MNT_FORCE)
flags |= FORCECLOSE;
if (fs->fs_flags & FS_DOSOFTDEP) {
error = softdep_flushfiles(mp, flags, p);
mp->mnt_flag &= ~MNT_SOFTDEP;
} else
error = ffs_flushfiles(mp, flags, p);
ronly = 1;
}
/*
* Flush soft dependencies if disabling it via an update
* mount. This may leave some items to be processed,
* so don't do this yet XXX.
*/
if ((fs->fs_flags & FS_DOSOFTDEP) &&
!(mp->mnt_flag & MNT_SOFTDEP) &&
!(mp->mnt_flag & MNT_RDONLY) && fs->fs_ronly == 0) {
#if 0
flags = WRITECLOSE;
if (mp->mnt_flag & MNT_FORCE)
flags |= FORCECLOSE;
error = softdep_flushfiles(mp, flags, p);
#elif FFS_SOFTUPDATES
mp->mnt_flag |= MNT_SOFTDEP;
#endif
}
/*
* When upgrading to a softdep mount, we must first flush
* all vnodes. (not done yet -- see above)
*/
if (!(fs->fs_flags & FS_DOSOFTDEP) &&
(mp->mnt_flag & MNT_SOFTDEP) && fs->fs_ronly == 0) {
#if 0
flags = WRITECLOSE;
if (mp->mnt_flag & MNT_FORCE)
flags |= FORCECLOSE;
error = ffs_flushfiles(mp, flags, p);
#else
mp->mnt_flag &= ~MNT_SOFTDEP;
#endif
}
if (!error && (mp->mnt_flag & MNT_RELOAD))
error = ffs_reload(mp, ndp->ni_cnd.cn_cred, p);
if (error)
goto error_1;
if (ronly && (mp->mnt_flag & MNT_WANTRDWR)) {
/*
* If upgrade to read-write by non-root, then verify
* that user has necessary permissions on the device.
*/
if (suser(p, 0)) {
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p);
error = VOP_ACCESS(devvp, VREAD | VWRITE,
p->p_ucred, p);
VOP_UNLOCK(devvp, 0, p);
if (error)
goto error_1;
}
if (fs->fs_clean == 0) {
#if 0
/*
* It is safe mount unclean file system
* if it was previously mounted with softdep
* but we may loss space and must
* sometimes run fsck manually.
*/
if (fs->fs_flags & FS_DOSOFTDEP)
printf(
"WARNING: %s was not properly unmounted\n",
fs->fs_fsmnt);
else
#endif
if (mp->mnt_flag & MNT_FORCE) {
printf(
"WARNING: %s was not properly unmounted\n",
fs->fs_fsmnt);
} else {
printf(
"WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
fs->fs_fsmnt);
error = EROFS;
goto error_1;
}
}
if ((fs->fs_flags & FS_DOSOFTDEP)) {
error = softdep_mount(devvp, mp, fs,
p->p_ucred);
if (error)
goto error_1;
}
fs->fs_contigdirs=(u_int8_t*)malloc((u_long)fs->fs_ncg,
M_UFSMNT, M_WAITOK);
bzero(fs->fs_contigdirs, fs->fs_ncg);
ronly = 0;
}
if (args.fspec == 0) {
/*
* Process export requests.
*/
error = vfs_export(mp, &ump->um_export,
&args.export_info);
if (error)
goto error_1;
else
goto success;
}
}
/*
* Not an update, or updating the name: look up the name
* and verify that it refers to a sensible block device.
*/
NDINIT(ndp, LOOKUP, FOLLOW, UIO_USERSPACE, args.fspec, p);
if ((error = namei(ndp)) != 0)
goto error_1;
devvp = ndp->ni_vp;
if (devvp->v_type != VBLK) {
error = ENOTBLK;
goto error_2;
}
if (major(devvp->v_rdev) >= nblkdev) {
error = ENXIO;
goto error_2;
}
/*
* If mount by non-root, then verify that user has necessary
* permissions on the device.
*/
if (suser(p, 0)) {
accessmode = VREAD;
if ((mp->mnt_flag & MNT_RDONLY) == 0)
accessmode |= VWRITE;
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p);
error = VOP_ACCESS(devvp, accessmode, p->p_ucred, p);
VOP_UNLOCK(devvp, 0, p);
if (error)
goto error_2;
}
if (mp->mnt_flag & MNT_UPDATE) {
/*
* UPDATE
* If it's not the same vnode, or at least the same device
* then it's not correct.
*/
if (devvp != ump->um_devvp) {
if (devvp->v_rdev == ump->um_devvp->v_rdev) {
vrele(devvp);
} else {
error = EINVAL; /* needs translation */
}
} else
vrele(devvp);
/*
* Update device name only on success
*/
if (!error) {
/*
* Save "mounted from" info for mount point (NULL pad)
*/
copyinstr(args.fspec,
mp->mnt_stat.f_mntfromname,
MNAMELEN - 1,
&size);
bzero(mp->mnt_stat.f_mntfromname + size,
MNAMELEN - size);
}
} else {
/*
* Since this is a new mount, we want the names for
* the device and the mount point copied in. If an
* error occurs, the mountpoint is discarded by the
* upper level code.
*/
/* Save "last mounted on" info for mount point (NULL pad)*/
copyinstr(path, /* mount point*/
mp->mnt_stat.f_mntonname, /* save area*/
MNAMELEN - 1, /* max size*/
&size); /* real size*/
bzero(mp->mnt_stat.f_mntonname + size, MNAMELEN - size);
/* Save "mounted from" info for mount point (NULL pad)*/
copyinstr(args.fspec, /* device name*/
mp->mnt_stat.f_mntfromname, /* save area*/
MNAMELEN - 1, /* max size*/
&size); /* real size*/
bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
error = ffs_mountfs(devvp, mp, p);
}
if (error)
goto error_2;
/*
* Initialize FS stat information in mount struct; uses both
* mp->mnt_stat.f_mntonname and mp->mnt_stat.f_mntfromname
*
* This code is common to root and non-root mounts
*/
bcopy(&args, &mp->mnt_stat.mount_info.ufs_args, sizeof(args));
(void)VFS_STATFS(mp, &mp->mnt_stat, p);
success:
if (path && (mp->mnt_flag & MNT_UPDATE)) {
/* Update clean flag after changing read-onlyness. */
fs = ump->um_fs;
if (ronly != fs->fs_ronly) {
fs->fs_ronly = ronly;
fs->fs_clean = ronly &&
(fs->fs_flags & FS_UNCLEAN) == 0 ? 1 : 0;
if (ronly)
free(fs->fs_contigdirs, M_UFSMNT);
}
if (!ronly) {
if (mp->mnt_flag & MNT_SOFTDEP)
fs->fs_flags |= FS_DOSOFTDEP;
else
fs->fs_flags &= ~FS_DOSOFTDEP;
}
ffs_sbupdate(ump, MNT_WAIT);
}
return (0);
error_2: /* error with devvp held */
vrele (devvp);
error_1: /* no state to back out */
return (error);
}
static int
ntfs_mount(struct mount *mp, char *path, caddr_t data, struct ucred *cred)
{
size_t size;
int error;
struct vnode *devvp;
struct ntfs_args args;
struct nlookupdata nd;
struct vnode *rootvp;
error = 0;
/*
* Use NULL path to flag a root mount
*/
if( path == NULL) {
/*
***
* Mounting root file system
***
*/
/* Get vnode for root device*/
if( bdevvp( rootdev, &rootvp))
panic("ffs_mountroot: can't setup bdevvp for root");
/*
* FS specific handling
*/
mp->mnt_flag |= MNT_RDONLY; /* XXX globally applicable?*/
/*
* Attempt mount
*/
if( ( error = ntfs_mountfs(rootvp, mp, &args, cred)) != 0) {
/* fs specific cleanup (if any)*/
goto error_1;
}
goto dostatfs; /* success*/
}
/*
***
* Mounting non-root file system or updating a file system
***
*/
/* copy in user arguments*/
error = copyin(data, (caddr_t)&args, sizeof (struct ntfs_args));
if (error)
goto error_1; /* can't get arguments*/
/*
* If updating, check whether changing from read-only to
* read/write; if there is no device name, that's all we do.
*/
if (mp->mnt_flag & MNT_UPDATE) {
/* if not updating name...*/
if (args.fspec == NULL) {
/*
* Process export requests. Jumping to "success"
* will return the vfs_export() error code.
*/
struct ntfsmount *ntm = VFSTONTFS(mp);
error = vfs_export(mp, &ntm->ntm_export, &args.export);
goto success;
}
kprintf("ntfs_mount(): MNT_UPDATE not supported\n");
error = EINVAL;
goto error_1;
}
/*
* Not an update, or updating the name: look up the name
* and verify that it refers to a sensible block device.
*/
devvp = NULL;
error = nlookup_init(&nd, args.fspec, UIO_USERSPACE, NLC_FOLLOW);
if (error == 0)
error = nlookup(&nd);
if (error == 0)
error = cache_vref(&nd.nl_nch, nd.nl_cred, &devvp);
nlookup_done(&nd);
if (error)
goto error_1;
if (!vn_isdisk(devvp, &error))
goto error_2;
if (mp->mnt_flag & MNT_UPDATE) {
#if 0
/*
********************
* UPDATE
********************
*/
if (devvp != ntmp->um_devvp)
error = EINVAL; /* needs translation */
else
vrele(devvp);
/*
* Update device name only on success
*/
if( !error) {
/* Save "mounted from" info for mount point (NULL pad)*/
copyinstr( args.fspec,
mp->mnt_stat.f_mntfromname,
MNAMELEN - 1,
&size);
bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
}
#endif
} else {
/*
********************
* NEW MOUNT
********************
*/
/* Save "mounted from" info for mount point (NULL pad)*/
copyinstr( args.fspec, /* device name*/
mp->mnt_stat.f_mntfromname, /* save area*/
MNAMELEN - 1, /* max size*/
&size); /* real size*/
bzero( mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
error = ntfs_mountfs(devvp, mp, &args, cred);
}
if (error) {
goto error_2;
}
dostatfs:
/*
* Initialize FS stat information in mount struct; uses
* mp->mnt_stat.f_mntfromname.
*
* This code is common to root and non-root mounts
*/
(void)VFS_STATFS(mp, &mp->mnt_stat, cred);
goto success;
error_2: /* error with devvp held*/
/* release devvp before failing*/
vrele(devvp);
error_1: /* no state to back out*/
success:
return(error);
}
static int
_xfs_mount(struct mount *mp)
{
struct xfsmount *xmp;
struct xfs_vnode *rootvp;
struct ucred *curcred;
struct vnode *rvp, *devvp;
struct cdev *ddev;
struct g_consumer *cp;
struct thread *td;
int error;
td = curthread;
ddev = NULL;
cp = NULL;
if (vfs_filteropt(mp->mnt_optnew, xfs_opts))
return (EINVAL);
if (mp->mnt_flag & MNT_UPDATE)
return (0);
if ((mp->mnt_flag & MNT_RDONLY) == 0)
return (EPERM);
xmp = xfsmount_allocate(mp);
if (xmp == NULL)
return (ENOMEM);
if((error = _xfs_param_copyin(mp, td)) != 0)
goto fail;
curcred = td->td_ucred;
XVFS_MOUNT(XFSTOVFS(xmp), &xmp->m_args, curcred, error);
if (error)
goto fail;
XVFS_ROOT(XFSTOVFS(xmp), &rootvp, error);
ddev = XFS_VFSTOM(XFSTOVFS(xmp))->m_ddev_targp->dev;
devvp = XFS_VFSTOM(XFSTOVFS(xmp))->m_ddev_targp->specvp;
if (error)
goto fail_unmount;
if (ddev->si_iosize_max != 0)
mp->mnt_iosize_max = ddev->si_iosize_max;
if (mp->mnt_iosize_max > MAXPHYS)
mp->mnt_iosize_max = MAXPHYS;
mp->mnt_flag |= MNT_LOCAL;
mp->mnt_stat.f_fsid.val[0] = dev2udev(ddev);
mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
if ((error = VFS_STATFS(mp, &mp->mnt_stat)) != 0)
goto fail_unmount;
rvp = rootvp->v_vnode;
rvp->v_vflag |= VV_ROOT;
VN_RELE(rootvp);
return (0);
fail_unmount:
XVFS_UNMOUNT(XFSTOVFS(xmp), 0, curcred, error);
if (devvp != NULL) {
cp = devvp->v_bufobj.bo_private;
if (cp != NULL) {
DROP_GIANT();
g_topology_lock();
g_vfs_close(cp);
g_topology_unlock();
PICKUP_GIANT();
}
}
fail:
if (xmp != NULL)
xfsmount_deallocate(xmp);
return (error);
}
int
domount(kthread_t *td, vnode_t *vp, const char *fstype, char *fspath,
char *fspec, int fsflags)
{
struct mount *mp;
struct vfsconf *vfsp;
struct ucred *newcr, *oldcr;
int error;
/*
* Be ultra-paranoid about making sure the type and fspath
* variables will fit in our mp buffers, including the
* terminating NUL.
*/
if (strlen(fstype) >= MFSNAMELEN || strlen(fspath) >= MNAMELEN)
return (ENAMETOOLONG);
vfsp = vfs_byname_kld(fstype, td, &error);
if (vfsp == NULL)
return (ENODEV);
if (vp->v_type != VDIR)
return (ENOTDIR);
simple_lock(&vp->v_interlock);
if ((vp->v_iflag & VI_MOUNT) != 0 ||
vp->v_mountedhere != NULL) {
simple_unlock(&vp->v_interlock);
return (EBUSY);
}
vp->v_iflag |= VI_MOUNT;
simple_unlock(&vp->v_interlock);
/*
* Allocate and initialize the filesystem.
*/
vn_lock(vp, LK_SHARED | LK_RETRY);
mp = vfs_mount_alloc(vp, vfsp, fspath, td);
VOP_UNLOCK(vp);
mp->mnt_optnew = NULL;
vfs_setmntopt(mp, "from", fspec, 0);
mp->mnt_optnew = mp->mnt_opt;
mp->mnt_opt = NULL;
/*
* Set the mount level flags.
* crdup() can sleep, so do it before acquiring a mutex.
*/
newcr = crdup(kcred);
MNT_ILOCK(mp);
if (fsflags & MNT_RDONLY)
mp->mnt_flag |= MNT_RDONLY;
mp->mnt_flag &=~ MNT_UPDATEMASK;
mp->mnt_flag |= fsflags & (MNT_UPDATEMASK | MNT_FORCE | MNT_ROOTFS);
/*
* Unprivileged user can trigger mounting a snapshot, but we don't want
* him to unmount it, so we switch to privileged credentials.
*/
oldcr = mp->mnt_cred;
mp->mnt_cred = newcr;
mp->mnt_stat.f_owner = mp->mnt_cred->cr_uid;
MNT_IUNLOCK(mp);
crfree(oldcr);
/*
* Mount the filesystem.
* XXX The final recipients of VFS_MOUNT just overwrite the ndp they
* get. No freeing of cn_pnbuf.
*/
error = VFS_MOUNT(mp, td);
if (!error) {
if (mp->mnt_opt != NULL)
vfs_freeopts(mp->mnt_opt);
mp->mnt_opt = mp->mnt_optnew;
(void)VFS_STATFS(mp, &mp->mnt_stat, td);
}
/*
* Prevent external consumers of mount options from reading
* mnt_optnew.
*/
mp->mnt_optnew = NULL;
vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
/*
* Put the new filesystem on the mount list after root.
*/
#ifdef FREEBSD_NAMECACHE
cache_purge(vp);
#endif
if (!error) {
vnode_t *mvp;
simple_lock(&vp->v_interlock);
vp->v_iflag &= ~VI_MOUNT;
simple_unlock(&vp->v_interlock);
vp->v_mountedhere = mp;
mountlist_append(mp);
vfs_event_signal(NULL, VQ_MOUNT, 0);
if (VFS_ROOT(mp, LK_EXCLUSIVE, &mvp, td))
panic("mount: lost mount");
mountcheckdirs(vp, mvp);
vput(mvp);
VOP_UNLOCK(vp);
if ((mp->mnt_flag & MNT_RDONLY) == 0)
vfs_syncer_add_to_worklist(mp);
vfs_unbusy(mp, td);
vfs_mountedfrom(mp, fspec);
} else {
simple_lock(&vp->v_interlock);
vp->v_iflag &= ~VI_MOUNT;
simple_unlock(&vp->v_interlock);
VOP_UNLOCK(vp);
vfs_unbusy(mp, td);
vfs_mount_destroy(mp);
}
return (error);
}
int
mount_snapshot(kthread_t *td, vnode_t **vpp, const char *fstype, char *fspath,
char *fspec, int fsflags)
{
struct vfsconf *vfsp;
struct mount *mp;
vnode_t *vp, *mvp;
struct ucred *cr;
int error;
/*
* Be ultra-paranoid about making sure the type and fspath
* variables will fit in our mp buffers, including the
* terminating NUL.
*/
if (strlen(fstype) >= MFSNAMELEN || strlen(fspath) >= MNAMELEN)
return (ENAMETOOLONG);
vfsp = vfs_byname_kld(fstype, td, &error);
if (vfsp == NULL)
return (ENODEV);
vp = *vpp;
if (vp->v_type != VDIR)
return (ENOTDIR);
/*
* We need vnode lock to protect v_mountedhere and vnode interlock
* to protect v_iflag.
*/
vn_lock(vp, LK_SHARED | LK_RETRY);
VI_LOCK(vp);
if ((vp->v_iflag & VI_MOUNT) != 0 || vp->v_mountedhere != NULL) {
VI_UNLOCK(vp);
VOP_UNLOCK(vp, 0);
return (EBUSY);
}
vp->v_iflag |= VI_MOUNT;
VI_UNLOCK(vp);
VOP_UNLOCK(vp, 0);
/*
* Allocate and initialize the filesystem.
* We don't want regular user that triggered snapshot mount to be able
* to unmount it, so pass credentials of the parent mount.
*/
mp = vfs_mount_alloc(vp, vfsp, fspath, vp->v_mount->mnt_cred);
mp->mnt_optnew = NULL;
vfs_setmntopt(mp, "from", fspec, 0);
mp->mnt_optnew = mp->mnt_opt;
mp->mnt_opt = NULL;
/*
* Set the mount level flags.
*/
mp->mnt_flag = fsflags & MNT_UPDATEMASK;
/*
* Snapshots are always read-only.
*/
mp->mnt_flag |= MNT_RDONLY;
/*
* We don't want snapshots to allow access to vulnerable setuid
* programs, so we turn off setuid when mounting snapshots.
*/
mp->mnt_flag |= MNT_NOSUID;
/*
* We don't want snapshots to be visible in regular
* mount(8) and df(1) output.
*/
mp->mnt_flag |= MNT_IGNORE;
/*
* XXX: This is evil, but we can't mount a snapshot as a regular user.
* XXX: Is is safe when snapshot is mounted from within a jail?
*/
cr = td->td_ucred;
td->td_ucred = kcred;
error = VFS_MOUNT(mp);
td->td_ucred = cr;
if (error != 0) {
VI_LOCK(vp);
vp->v_iflag &= ~VI_MOUNT;
VI_UNLOCK(vp);
vrele(vp);
vfs_unbusy(mp);
vfs_mount_destroy(mp);
*vpp = NULL;
return (error);
}
if (mp->mnt_opt != NULL)
vfs_freeopts(mp->mnt_opt);
mp->mnt_opt = mp->mnt_optnew;
(void)VFS_STATFS(mp, &mp->mnt_stat);
/*
* Prevent external consumers of mount options from reading
* mnt_optnew.
*/
mp->mnt_optnew = NULL;
vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
#ifdef FREEBSD_NAMECACHE
cache_purge(vp);
#endif
VI_LOCK(vp);
vp->v_iflag &= ~VI_MOUNT;
VI_UNLOCK(vp);
vp->v_mountedhere = mp;
/* Put the new filesystem on the mount list. */
mtx_lock(&mountlist_mtx);
TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
mtx_unlock(&mountlist_mtx);
vfs_event_signal(NULL, VQ_MOUNT, 0);
if (VFS_ROOT(mp, LK_EXCLUSIVE, &mvp))
panic("mount: lost mount");
vput(vp);
vfs_unbusy(mp);
*vpp = mvp;
return (0);
}
