static int zfsfuse_access(fuse_req_t req, fuse_ino_t ino, int mask)
{
vfs_t *vfs = (vfs_t *) fuse_req_userdata(req);
zfsvfs_t *zfsvfs = vfs->vfs_data;
ZFS_ENTER(zfsvfs);
znode_t *znode;
int error = zfs_zget(zfsvfs, ino, &znode, B_TRUE);
if(error) {
ZFS_EXIT(zfsvfs);
/* If the inode we are trying to get was recently deleted
dnode_hold_impl will return EEXIST instead of ENOENT */
return error == EEXIST ? ENOENT : error;
}
ASSERT(znode != NULL);
vnode_t *vp = ZTOV(znode);
ASSERT(vp != NULL);
cred_t cred;
zfsfuse_getcred(req, &cred);
int mode = 0;
if(mask & R_OK)
mode |= VREAD;
if(mask & W_OK)
mode |= VWRITE;
if(mask & X_OK)
mode |= VEXEC;
error = VOP_ACCESS(vp, mode, 0, &cred, NULL);
VN_RELE(vp);
ZFS_EXIT(zfsvfs);
return error;
}
static int
zfs_vfs_sync(struct mount *mp, __unused int waitfor, __unused vfs_context_t context)
{
zfsvfs_t *zfsvfs = vfs_fsprivate(mp);
ZFS_ENTER(zfsvfs);
/*
* Mac OS X needs a file system modify time
*
* We use the mtime of the "com.apple.system.mtime"
* extended attribute, which is associated with the
* file system root directory.
*
* Here we sync any mtime changes to this attribute.
*/
if (zfsvfs->z_mtime_vp != NULL) {
timestruc_t mtime;
znode_t *zp;
top:
zp = VTOZ(zfsvfs->z_mtime_vp);
ZFS_TIME_DECODE(&mtime, zp->z_phys->zp_mtime);
if (zfsvfs->z_last_mtime_synced < mtime.tv_sec) {
dmu_tx_t *tx;
int error;
tx = dmu_tx_create(zfsvfs->z_os);
dmu_tx_hold_bonus(tx, zp->z_id);
error = dmu_tx_assign(tx, zfsvfs->z_assign);
if (error) {
if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) {
dmu_tx_wait(tx);
dmu_tx_abort(tx);
goto top;
}
dmu_tx_abort(tx);
} else {
dmu_buf_will_dirty(zp->z_dbuf, tx);
dmu_tx_commit(tx);
zfsvfs->z_last_mtime_synced = mtime.tv_sec;
}
}
}
if (zfsvfs->z_log != NULL)
zil_commit(zfsvfs->z_log, UINT64_MAX, 0);
else
txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), 0);
ZFS_EXIT(zfsvfs);
return (0);
}
/* ARGSUSED */
static int
zfsctl_shares_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
caller_context_t *ct)
{
zfsvfs_t *zfsvfs = vp->v_vfsp->vfs_data;
znode_t *dzp;
int error;
ZFS_ENTER(zfsvfs);
if (zfsvfs->z_shares_dir == 0) {
ZFS_EXIT(zfsvfs);
return (ENOTSUP);
}
if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &dzp)) == 0) {
error = VOP_GETATTR(ZTOV(dzp), vap, flags, cr, ct);
VN_RELE(ZTOV(dzp));
}
ZFS_EXIT(zfsvfs);
return (error);
}
/*ARGSUSED*/
static int
zfsctl_shares_fid(vnode_t *vp, fid_t *fidp, caller_context_t *ct)
{
zfsvfs_t *zfsvfs = vp->v_vfsp->vfs_data;
znode_t *dzp;
int error;
ZFS_ENTER(zfsvfs);
if (zfsvfs->z_shares_dir == 0) {
ZFS_EXIT(zfsvfs);
return (ENOTSUP);
}
if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &dzp)) == 0) {
error = VOP_FID(ZTOV(dzp), fidp, ct);
VN_RELE(ZTOV(dzp));
}
ZFS_EXIT(zfsvfs);
return (error);
}
/*
* File handle to vnode pointer
*/
static int
zfs_vfs_fhtovp(struct mount *mp, int fhlen, unsigned char *fhp,
struct vnode **vpp, __unused vfs_context_t context)
{
zfsvfs_t *zfsvfs = vfs_fsprivate(mp);
zfs_zfid_t *zfid = (zfs_zfid_t *)fhp;
znode_t *zp;
uint64_t obj_num = 0;
uint64_t fid_gen = 0;
uint64_t zp_gen;
int i;
int error;
*vpp = NULL;
ZFS_ENTER(zfsvfs);
if (fhlen < sizeof (zfs_zfid_t)) {
error = EINVAL;
goto out;
}
/*
* Grab the object and gen numbers in an endian neutral manner
*/
for (i = 0; i < sizeof (zfid->zf_object); i++)
obj_num |= ((uint64_t)zfid->zf_object[i]) << (8 * i);
for (i = 0; i < sizeof (zfid->zf_gen); i++)
fid_gen |= ((uint64_t)zfid->zf_gen[i]) << (8 * i);
if ((error = zfs_zget(zfsvfs, obj_num, &zp))) {
goto out;
}
zp_gen = zp->z_phys->zp_gen;
if (zp_gen == 0)
zp_gen = 1;
if (zp->z_unlinked || zp_gen != fid_gen) {
vnode_put(ZTOV(zp));
error = EINVAL;
goto out;
}
*vpp = ZTOV(zp);
out:
ZFS_EXIT(zfsvfs);
return (error);
}
/* ARGSUSED */
static int
zfsctl_root_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
caller_context_t *ct)
{
zfsvfs_t *zfsvfs = vp->v_vfsp->vfs_data;
ZFS_ENTER(zfsvfs);
vap->va_nodeid = ZFSCTL_INO_ROOT;
vap->va_nlink = vap->va_size = NROOT_ENTRIES;
zfsctl_common_getattr(vp, vap);
ZFS_EXIT(zfsvfs);
return (0);
}
/* ARGSUSED */
static int
zfsctl_snapdir_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
caller_context_t *ct)
{
zfsvfs_t *zfsvfs = vp->v_vfsp->vfs_data;
zfsctl_snapdir_t *sdp = vp->v_data;
ZFS_ENTER(zfsvfs);
zfsctl_common_getattr(vp, vap);
vap->va_nodeid = gfs_file_inode(vp);
vap->va_nlink = vap->va_size = avl_numnodes(&sdp->sd_snaps) + 2;
ZFS_EXIT(zfsvfs);
return (0);
}
int
zfs_root(zfs_sb_t *zsb, struct inode **ipp)
{
znode_t *rootzp;
int error;
ZFS_ENTER(zsb);
error = zfs_zget(zsb, zsb->z_root, &rootzp);
if (error == 0)
*ipp = ZTOI(rootzp);
ZFS_EXIT(zsb);
return (error);
}
/* ARGSUSED */
static int
zpl_snapdir_getattr(struct vfsmount *mnt, struct dentry *dentry,
struct kstat *stat)
{
zfs_sb_t *zsb = ITOZSB(dentry->d_inode);
int error;
ZFS_ENTER(zsb);
error = simple_getattr(mnt, dentry, stat);
stat->nlink = stat->size = 2;
stat->ctime = stat->mtime = dmu_objset_snap_cmtime(zsb->z_os);
stat->atime = CURRENT_TIME;
ZFS_EXIT(zsb);
return (error);
}
static int
zfs_vfs_root(struct mount *mp, struct vnode **vpp, __unused vfs_context_t context)
{
zfsvfs_t *zfsvfs = vfs_fsprivate(mp);
znode_t *rootzp;
int error;
ZFS_ENTER(zfsvfs);
error = zfs_zget(zfsvfs, zfsvfs->z_root, &rootzp);
if (error == 0)
*vpp = ZTOV(rootzp);
ZFS_EXIT(zfsvfs);
return (error);
}
int
zfs_sb_prune(struct super_block *sb, unsigned long nr_to_scan, int *objects)
{
zfs_sb_t *zsb = sb->s_fs_info;
struct shrinker *shrinker = &sb->s_shrink;
struct shrink_control sc = {
.nr_to_scan = nr_to_scan,
.gfp_mask = GFP_KERNEL,
};
ZFS_ENTER(zsb);
*objects = (*shrinker->shrink)(shrinker, &sc);
ZFS_EXIT(zsb);
return (0);
}
static int
zfs_root(vfs_t *vfsp, vnode_t **vpp)
{
zfsvfs_t *zfsvfs = vfsp->vfs_data;
znode_t *rootzp;
int error;
ZFS_ENTER(zfsvfs);
error = zfs_zget(zfsvfs, zfsvfs->z_root, &rootzp);
if (error == 0)
*vpp = ZTOV(rootzp);
ZFS_EXIT(zfsvfs);
return (error);
}
static int zfsfuse_read(fuse_req_t req, fuse_ino_t ino, size_t size, off_t off, struct fuse_file_info *fi)
{
file_info_t *info = (file_info_t *)(uintptr_t) fi->fh;
vnode_t *vp = info->vp;
ASSERT(vp != NULL);
ASSERT(VTOZ(vp) != NULL);
ASSERT(VTOZ(vp)->z_id == ino);
vfs_t *vfs = (vfs_t *) fuse_req_userdata(req);
zfsvfs_t *zfsvfs = vfs->vfs_data;
char *outbuf = kmem_alloc(size, KM_NOSLEEP);
if(outbuf == NULL)
return ENOMEM;
ZFS_ENTER(zfsvfs);
iovec_t iovec;
uio_t uio;
uio.uio_iov = &iovec;
uio.uio_iovcnt = 1;
uio.uio_segflg = UIO_SYSSPACE;
uio.uio_fmode = 0;
uio.uio_llimit = RLIM64_INFINITY;
iovec.iov_base = outbuf;
iovec.iov_len = size;
uio.uio_resid = iovec.iov_len;
uio.uio_loffset = off;
cred_t cred;
zfsfuse_getcred(req, &cred);
int error = VOP_READ(vp, &uio, info->flags, &cred, NULL);
ZFS_EXIT(zfsvfs);
if(!error)
fuse_reply_buf(req, outbuf, uio.uio_loffset - off);
kmem_free(outbuf, size);
return error;
}
static int
zfs_root(vfs_t *vfsp, int flags, vnode_t **vpp)
{
zfsvfs_t *zfsvfs = vfsp->vfs_data;
znode_t *rootzp;
int error;
ZFS_ENTER_NOERROR(zfsvfs);
error = zfs_zget(zfsvfs, zfsvfs->z_root, &rootzp);
if (error == 0) {
*vpp = ZTOV(rootzp);
error = vn_lock(*vpp, flags);
(*vpp)->v_vflag |= VV_ROOT;
}
ZFS_EXIT(zfsvfs);
return (error);
}
static int zfsfuse_write(fuse_req_t req, fuse_ino_t ino, const char *buf, size_t size, off_t off, struct fuse_file_info *fi)
{
file_info_t *info = (file_info_t *)(uintptr_t) fi->fh;
vnode_t *vp = info->vp;
ASSERT(vp != NULL);
ASSERT(VTOZ(vp) != NULL);
ASSERT(VTOZ(vp)->z_id == ino);
vfs_t *vfs = (vfs_t *) fuse_req_userdata(req);
zfsvfs_t *zfsvfs = vfs->vfs_data;
ZFS_ENTER(zfsvfs);
iovec_t iovec;
uio_t uio;
uio.uio_iov = &iovec;
uio.uio_iovcnt = 1;
uio.uio_segflg = UIO_SYSSPACE;
uio.uio_fmode = 0;
uio.uio_llimit = RLIM64_INFINITY;
iovec.iov_base = (void *) buf;
iovec.iov_len = size;
uio.uio_resid = iovec.iov_len;
uio.uio_loffset = off;
cred_t cred;
zfsfuse_getcred(req, &cred);
int error = VOP_WRITE(vp, &uio, info->flags, &cred, NULL);
ZFS_EXIT(zfsvfs);
if(!error) {
/* When not using direct_io, we must always write 'size' bytes */
VERIFY(uio.uio_resid == 0);
fuse_reply_write(req, size - uio.uio_resid);
}
return error;
}
/*
* The ARC has requested that the filesystem drop entries from the dentry
* and inode caches. This can occur when the ARC needs to free meta data
* blocks but can't because they are all pinned by entries in these caches.
*/
int
zfs_sb_prune(struct super_block *sb, unsigned long nr_to_scan, int *objects)
{
zfs_sb_t *zsb = sb->s_fs_info;
int error = 0;
#if defined(HAVE_SHRINK) || defined(HAVE_SPLIT_SHRINKER_CALLBACK)
struct shrinker *shrinker = &sb->s_shrink;
struct shrink_control sc = {
.nr_to_scan = nr_to_scan,
.gfp_mask = GFP_KERNEL,
};
#endif
ZFS_ENTER(zsb);
#if defined(HAVE_SPLIT_SHRINKER_CALLBACK) && \
defined(SHRINK_CONTROL_HAS_NID) && \
defined(SHRINKER_NUMA_AWARE)
if (sb->s_shrink.flags & SHRINKER_NUMA_AWARE) {
*objects = 0;
for_each_online_node(sc.nid)
*objects += (*shrinker->scan_objects)(shrinker, &sc);
} else {
*objects = (*shrinker->scan_objects)(shrinker, &sc);
}
#elif defined(HAVE_SPLIT_SHRINKER_CALLBACK)
*objects = (*shrinker->scan_objects)(shrinker, &sc);
#elif defined(HAVE_SHRINK)
*objects = (*shrinker->shrink)(shrinker, &sc);
#elif defined(HAVE_D_PRUNE_ALIASES)
*objects = zfs_sb_prune_aliases(zsb, nr_to_scan);
#else
#error "No available dentry and inode cache pruning mechanism."
#endif
ZFS_EXIT(zsb);
dprintf_ds(zsb->z_os->os_dsl_dataset,
"pruning, nr_to_scan=%lu objects=%d error=%d\n",
nr_to_scan, *objects, error);
return (error);
}
/*ARGSUSED*/
int
zfs_sync(vfs_t *vfsp, short flag, cred_t *cr)
{
/*
* Data integrity is job one. We don't want a compromised kernel
* writing to the storage pool, so we never sync during panic.
*/
if (panicstr)
return (0);
/*
* SYNC_ATTR is used by fsflush() to force old filesystems like UFS
* to sync metadata, which they would otherwise cache indefinitely.
* Semantically, the only requirement is that the sync be initiated.
* The DMU syncs out txgs frequently, so there's nothing to do.
*/
if (flag & SYNC_ATTR)
return (0);
if (vfsp != NULL) {
/*
* Sync a specific filesystem.
*/
zfsvfs_t *zfsvfs = vfsp->vfs_data;
ZFS_ENTER(zfsvfs);
if (zfsvfs->z_log != NULL)
zil_commit(zfsvfs->z_log, UINT64_MAX, 0);
else
txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), 0);
ZFS_EXIT(zfsvfs);
} else {
/*
* Sync all ZFS filesystems. This is what happens when you
* run sync(1M). Unlike other filesystems, ZFS honors the
* request by waiting for all pools to commit all dirty data.
*/
spa_sync_allpools();
}
return (0);
}
int
zfs_sb_prune(struct super_block *sb, unsigned long nr_to_scan, int *objects)
{
zfs_sb_t *zsb = sb->s_fs_info;
struct shrinker *shrinker = &sb->s_shrink;
struct shrink_control sc = {
.nr_to_scan = nr_to_scan,
.gfp_mask = GFP_KERNEL,
};
ZFS_ENTER(zsb);
#ifdef HAVE_SPLIT_SHRINKER_CALLBACK
*objects = (*shrinker->scan_objects)(shrinker, &sc);
#else
*objects = (*shrinker->shrink)(shrinker, &sc);
#endif
ZFS_EXIT(zsb);
return (0);
}
/*
* The ARC has requested that the filesystem drop entries from the dentry
* and inode caches. This can occur when the ARC needs to free meta data
* blocks but can't because they are all pinned by entries in these caches.
*/
int
zfs_sb_prune(struct super_block *sb, unsigned long nr_to_scan, int *objects)
{
zfs_sb_t *zsb = sb->s_fs_info;
int error = 0;
#if defined(HAVE_SHRINK) || defined(HAVE_SPLIT_SHRINKER_CALLBACK)
struct shrinker *shrinker = &sb->s_shrink;
struct shrink_control sc = {
.nr_to_scan = nr_to_scan,
.gfp_mask = GFP_KERNEL,
};
#endif
ZFS_ENTER(zsb);
#if defined(HAVE_SPLIT_SHRINKER_CALLBACK)
*objects = (*shrinker->scan_objects)(shrinker, &sc);
#elif defined(HAVE_SHRINK)
*objects = (*shrinker->shrink)(shrinker, &sc);
#else
/*
* Linux kernels older than 3.1 do not support a per-filesystem
* shrinker. Therefore, we must fall back to the only available
* interface which is to discard all unused dentries and inodes.
* This behavior clearly isn't ideal but it's required so the ARC
* may free memory. The performance impact is mitigated by the
* fact that the frequently accessed dentry and inode buffers will
* still be in the ARC making them relatively cheap to recreate.
*/
*objects = 0;
shrink_dcache_parent(sb->s_root);
#endif
ZFS_EXIT(zsb);
dprintf_ds(zsb->z_os->os_dsl_dataset,
"pruning, nr_to_scan=%lu objects=%d error=%d\n",
nr_to_scan, *objects, error);
return (error);
}
static int
zpl_snapdir_iterate(struct file *filp, struct dir_context *ctx)
{
zfs_sb_t *zsb = ITOZSB(filp->f_path.dentry->d_inode);
fstrans_cookie_t cookie;
char snapname[MAXNAMELEN];
boolean_t case_conflict;
uint64_t id, pos;
int error = 0;
ZFS_ENTER(zsb);
cookie = spl_fstrans_mark();
if (!dir_emit_dots(filp, ctx))
goto out;
pos = ctx->pos;
while (error == 0) {
dsl_pool_config_enter(dmu_objset_pool(zsb->z_os), FTAG);
error = -dmu_snapshot_list_next(zsb->z_os, MAXNAMELEN,
snapname, &id, &pos, &case_conflict);
dsl_pool_config_exit(dmu_objset_pool(zsb->z_os), FTAG);
if (error)
goto out;
if (!dir_emit(ctx, snapname, strlen(snapname),
ZFSCTL_INO_SHARES - id, DT_DIR))
goto out;
ctx->pos = pos;
}
out:
spl_fstrans_unmark(cookie);
ZFS_EXIT(zsb);
if (error == -ENOENT)
return (0);
return (error);
}
/* ARGSUSED */
int
zfsctl_snapdir_remove(struct inode *dip, char *name, cred_t *cr, int flags)
{
zfs_sb_t *zsb = ITOZSB(dip);
char *snapname, *real;
int error;
ZFS_ENTER(zsb);
snapname = kmem_alloc(MAXNAMELEN, KM_SLEEP);
real = kmem_alloc(MAXNAMELEN, KM_SLEEP);
if (zsb->z_case == ZFS_CASE_INSENSITIVE) {
error = dmu_snapshot_realname(zsb->z_os, name, real,
MAXNAMELEN, NULL);
if (error == 0) {
name = real;
} else if (error != ENOTSUP) {
goto out;
}
}
error = zfsctl_snapshot_zname(dip, name, MAXNAMELEN, snapname);
if (!error)
error = zfs_secpolicy_destroy_perms(snapname, cr);
if (error)
goto out;
error = zfsctl_unmount_snapshot(zsb, name, MNT_FORCE);
if ((error == 0) || (error == ENOENT))
error = dmu_objset_destroy(snapname, B_FALSE);
out:
kmem_free(snapname, MAXNAMELEN);
kmem_free(real, MAXNAMELEN);
ZFS_EXIT(zsb);
return (error);
}
/*ARGSUSED*/
static int
zfs_sync(vfs_t *vfsp, int waitfor)
{
/*
* Data integrity is job one. We don't want a compromised kernel
* writing to the storage pool, so we never sync during panic.
*/
if (panicstr)
return (0);
if (vfsp != NULL) {
/*
* Sync a specific filesystem.
*/
zfsvfs_t *zfsvfs = vfsp->vfs_data;
int error;
error = vfs_stdsync(vfsp, waitfor);
if (error != 0)
return (error);
ZFS_ENTER(zfsvfs);
if (zfsvfs->z_log != NULL)
zil_commit(zfsvfs->z_log, UINT64_MAX, 0);
else
txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), 0);
ZFS_EXIT(zfsvfs);
} else {
/*
* Sync all ZFS filesystems. This is what happens when you
* run sync(1M). Unlike other filesystems, ZFS honors the
* request by waiting for all pools to commit all dirty data.
*/
spa_sync_allpools();
}
return (0);
}
static int zfsfuse_fsync(fuse_req_t req, fuse_ino_t ino, int datasync, struct fuse_file_info *fi)
{
vfs_t *vfs = (vfs_t *) fuse_req_userdata(req);
zfsvfs_t *zfsvfs = vfs->vfs_data;
ZFS_ENTER(zfsvfs);
file_info_t *info = (file_info_t *)(uintptr_t) fi->fh;
ASSERT(info->vp != NULL);
ASSERT(VTOZ(info->vp) != NULL);
ASSERT(VTOZ(info->vp)->z_id == ino);
vnode_t *vp = info->vp;
cred_t cred;
zfsfuse_getcred(req, &cred);
int error = VOP_FSYNC(vp, datasync ? FDSYNC : FSYNC, &cred, NULL);
ZFS_EXIT(zfsvfs);
return error;
}
static int
zfs_vget(vfs_t *vfsp, ino_t ino, int flags, vnode_t **vpp)
{
zfsvfs_t *zfsvfs = vfsp->vfs_data;
znode_t *zp;
int err;
/*
* XXXPJD: zfs_zget() can't operate on virtual entires like .zfs/ or
* .zfs/snapshot/ directories, so for now just return EOPNOTSUPP.
* This will make NFS to fall back to using READDIR instead of
* READDIRPLUS.
* Also snapshots are stored in AVL tree, but based on their names,
* not inode numbers, so it will be very inefficient to iterate
* over all snapshots to find the right one.
* Note that OpenSolaris READDIRPLUS implementation does LOOKUP on
* d_name, and not VGET on d_fileno as we do.
*/
if (ino == ZFSCTL_INO_ROOT || ino == ZFSCTL_INO_SNAPDIR)
return (EOPNOTSUPP);
ZFS_ENTER(zfsvfs);
err = zfs_zget(zfsvfs, ino, &zp);
if (err == 0 && zp->z_unlinked) {
VN_RELE(ZTOV(zp));
err = EINVAL;
}
if (err != 0)
*vpp = NULL;
else {
*vpp = ZTOV(zp);
vn_lock(*vpp, flags);
}
ZFS_EXIT(zfsvfs);
return (err);
}
/*
* Vnode pointer to File handle
*
* XXX Do we want to check the DSL sharenfs property?
*/
static int
zfs_vfs_vptofh(struct vnode *vp, int *fhlenp, unsigned char *fhp, __unused vfs_context_t context)
{
zfsvfs_t *zfsvfs = vfs_fsprivate(vnode_mount(vp));
zfs_zfid_t *zfid = (zfs_zfid_t *)fhp;
znode_t *zp = VTOZ(vp);
uint64_t obj_num;
uint64_t zp_gen;
int i;
int error;
if (*fhlenp < sizeof (zfs_zfid_t)) {
return (EOVERFLOW);
}
ZFS_ENTER(zfsvfs);
obj_num = zp->z_id;
zp_gen = zp->z_phys->zp_gen;
if (zp_gen == 0)
zp_gen = 1;
/*
* Store the object and gen numbers in an endian neutral manner
*/
for (i = 0; i < sizeof (zfid->zf_object); i++)
zfid->zf_object[i] = (uint8_t)(obj_num >> (8 * i));
for (i = 0; i < sizeof (zfid->zf_gen); i++)
zfid->zf_gen[i] = (uint8_t)(zp_gen >> (8 * i));
*fhlenp = sizeof (zfs_zfid_t);
ZFS_EXIT(zfsvfs);
return (0);
}
int
zfsctl_mount_snapshot(struct path *path, int flags)
{
struct dentry *dentry = path->dentry;
struct inode *ip = dentry->d_inode;
zfs_sb_t *zsb = ITOZSB(ip);
char *full_name, *full_path;
zfs_snapentry_t *sep;
zfs_snapentry_t search;
char *argv[] = { "/bin/sh", "-c", NULL, NULL };
char *envp[] = { NULL };
int error;
ZFS_ENTER(zsb);
full_name = kmem_zalloc(MAXNAMELEN, KM_SLEEP);
full_path = kmem_zalloc(PATH_MAX, KM_SLEEP);
error = zfsctl_snapshot_zname(ip, dname(dentry), MAXNAMELEN, full_name);
if (error)
goto error;
error = zfsctl_snapshot_zpath(path, PATH_MAX, full_path);
if (error)
goto error;
/*
* Attempt to mount the snapshot from user space. Normally this
* would be done using the vfs_kern_mount() function, however that
* function is marked GPL-only and cannot be used. On error we
* careful to log the real error to the console and return EISDIR
* to safely abort the automount. This should be very rare.
*/
argv[2] = kmem_asprintf(SET_MOUNT_CMD, full_name, full_path);
error = call_usermodehelper(argv[0], argv, envp, UMH_WAIT_PROC);
strfree(argv[2]);
if (error) {
printk("ZFS: Unable to automount %s at %s: %d\n",
full_name, full_path, error);
error = EISDIR;
goto error;
}
mutex_enter(&zsb->z_ctldir_lock);
/*
* Ensure a previous entry does not exist, if it does safely remove
* it any cancel the outstanding expiration. This can occur when a
* snapshot is manually unmounted and then an automount is triggered.
*/
search.se_name = full_name;
sep = avl_find(&zsb->z_ctldir_snaps, &search, NULL);
if (sep) {
avl_remove(&zsb->z_ctldir_snaps, sep);
taskq_cancel_id(zfs_expire_taskq, sep->se_taskqid);
zfsctl_sep_free(sep);
}
sep = zfsctl_sep_alloc();
sep->se_name = full_name;
sep->se_path = full_path;
sep->se_inode = ip;
avl_add(&zsb->z_ctldir_snaps, sep);
sep->se_taskqid = taskq_dispatch_delay(zfs_expire_taskq,
zfsctl_expire_snapshot, sep, TQ_SLEEP,
ddi_get_lbolt() + zfs_expire_snapshot * HZ);
mutex_exit(&zsb->z_ctldir_lock);
error:
if (error) {
kmem_free(full_name, MAXNAMELEN);
kmem_free(full_path, PATH_MAX);
}
ZFS_EXIT(zsb);
return (error);
}
/*ARGSUSED*/
int
zfsctl_snapdir_rename(struct inode *sdip, char *sname,
struct inode *tdip, char *tname, cred_t *cr, int flags)
{
zfs_sb_t *zsb = ITOZSB(sdip);
zfs_snapentry_t search, *sep;
avl_index_t where;
char *to, *from, *real;
int error;
ZFS_ENTER(zsb);
to = kmem_alloc(MAXNAMELEN, KM_SLEEP);
from = kmem_alloc(MAXNAMELEN, KM_SLEEP);
real = kmem_alloc(MAXNAMELEN, KM_SLEEP);
if (zsb->z_case == ZFS_CASE_INSENSITIVE) {
error = dmu_snapshot_realname(zsb->z_os, sname, real,
MAXNAMELEN, NULL);
if (error == 0) {
sname = real;
} else if (error != ENOTSUP) {
goto out;
}
}
error = zfsctl_snapshot_zname(sdip, sname, MAXNAMELEN, from);
if (!error)
error = zfsctl_snapshot_zname(tdip, tname, MAXNAMELEN, to);
if (!error)
error = zfs_secpolicy_rename_perms(from, to, cr);
if (error)
goto out;
/*
* Cannot move snapshots out of the snapdir.
*/
if (sdip != tdip) {
error = EINVAL;
goto out;
}
/*
* No-op when names are identical.
*/
if (strcmp(sname, tname) == 0) {
error = 0;
goto out;
}
mutex_enter(&zsb->z_ctldir_lock);
error = dmu_objset_rename(from, to, B_FALSE);
if (error)
goto out_unlock;
search.se_name = (char *)sname;
sep = avl_find(&zsb->z_ctldir_snaps, &search, &where);
if (sep)
zfsctl_rename_snap(zsb, sep, tname);
out_unlock:
mutex_exit(&zsb->z_ctldir_lock);
out:
kmem_free(from, MAXNAMELEN);
kmem_free(to, MAXNAMELEN);
kmem_free(real, MAXNAMELEN);
ZFS_EXIT(zsb);
return (error);
}
int
zfs_vget(struct super_block *sb, struct inode **ipp, fid_t *fidp)
{
zfs_sb_t *zsb = sb->s_fs_info;
znode_t *zp;
uint64_t object = 0;
uint64_t fid_gen = 0;
uint64_t gen_mask;
uint64_t zp_gen;
int i, err;
*ipp = NULL;
ZFS_ENTER(zsb);
if (fidp->fid_len == LONG_FID_LEN) {
zfid_long_t *zlfid = (zfid_long_t *)fidp;
uint64_t objsetid = 0;
uint64_t setgen = 0;
for (i = 0; i < sizeof (zlfid->zf_setid); i++)
objsetid |= ((uint64_t)zlfid->zf_setid[i]) << (8 * i);
for (i = 0; i < sizeof (zlfid->zf_setgen); i++)
setgen |= ((uint64_t)zlfid->zf_setgen[i]) << (8 * i);
ZFS_EXIT(zsb);
err = zfsctl_lookup_objset(sb, objsetid, &zsb);
if (err)
return (SET_ERROR(EINVAL));
ZFS_ENTER(zsb);
}
if (fidp->fid_len == SHORT_FID_LEN || fidp->fid_len == LONG_FID_LEN) {
zfid_short_t *zfid = (zfid_short_t *)fidp;
for (i = 0; i < sizeof (zfid->zf_object); i++)
object |= ((uint64_t)zfid->zf_object[i]) << (8 * i);
for (i = 0; i < sizeof (zfid->zf_gen); i++)
fid_gen |= ((uint64_t)zfid->zf_gen[i]) << (8 * i);
} else {
ZFS_EXIT(zsb);
return (SET_ERROR(EINVAL));
}
/* A zero fid_gen means we are in the .zfs control directories */
if (fid_gen == 0 &&
(object == ZFSCTL_INO_ROOT || object == ZFSCTL_INO_SNAPDIR)) {
*ipp = zsb->z_ctldir;
ASSERT(*ipp != NULL);
if (object == ZFSCTL_INO_SNAPDIR) {
VERIFY(zfsctl_root_lookup(*ipp, "snapshot", ipp,
0, kcred, NULL, NULL) == 0);
} else {
igrab(*ipp);
}
ZFS_EXIT(zsb);
return (0);
}
gen_mask = -1ULL >> (64 - 8 * i);
dprintf("getting %llu [%llu mask %llx]\n", object, fid_gen, gen_mask);
if ((err = zfs_zget(zsb, object, &zp))) {
ZFS_EXIT(zsb);
return (err);
}
(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zsb), &zp_gen,
sizeof (uint64_t));
zp_gen = zp_gen & gen_mask;
if (zp_gen == 0)
zp_gen = 1;
if (zp->z_unlinked || zp_gen != fid_gen) {
dprintf("znode gen (%llu) != fid gen (%llu)\n", zp_gen,
fid_gen);
iput(ZTOI(zp));
ZFS_EXIT(zsb);
return (SET_ERROR(EINVAL));
}
*ipp = ZTOI(zp);
if (*ipp)
zfs_inode_update(ITOZ(*ipp));
ZFS_EXIT(zsb);
return (0);
}
int
zfs_statvfs(struct dentry *dentry, struct kstatfs *statp)
{
zfs_sb_t *zsb = dentry->d_sb->s_fs_info;
uint64_t refdbytes, availbytes, usedobjs, availobjs;
uint64_t fsid;
uint32_t bshift;
ZFS_ENTER(zsb);
dmu_objset_space(zsb->z_os,
&refdbytes, &availbytes, &usedobjs, &availobjs);
fsid = dmu_objset_fsid_guid(zsb->z_os);
/*
* The underlying storage pool actually uses multiple block
* size. Under Solaris frsize (fragment size) is reported as
* the smallest block size we support, and bsize (block size)
* as the filesystem's maximum block size. Unfortunately,
* under Linux the fragment size and block size are often used
* interchangeably. Thus we are forced to report both of them
* as the filesystem's maximum block size.
*/
statp->f_frsize = zsb->z_max_blksz;
statp->f_bsize = zsb->z_max_blksz;
bshift = fls(statp->f_bsize) - 1;
/*
* The following report "total" blocks of various kinds in
* the file system, but reported in terms of f_bsize - the
* "preferred" size.
*/
statp->f_blocks = (refdbytes + availbytes) >> bshift;
statp->f_bfree = availbytes >> bshift;
statp->f_bavail = statp->f_bfree; /* no root reservation */
/*
* statvfs() should really be called statufs(), because it assumes
* static metadata. ZFS doesn't preallocate files, so the best
* we can do is report the max that could possibly fit in f_files,
* and that minus the number actually used in f_ffree.
* For f_ffree, report the smaller of the number of object available
* and the number of blocks (each object will take at least a block).
*/
statp->f_ffree = MIN(availobjs, availbytes >> DNODE_SHIFT);
statp->f_files = statp->f_ffree + usedobjs;
statp->f_fsid.val[0] = (uint32_t)fsid;
statp->f_fsid.val[1] = (uint32_t)(fsid >> 32);
statp->f_type = ZFS_SUPER_MAGIC;
statp->f_namelen = ZFS_MAXNAMELEN;
/*
* We have all of 40 characters to stuff a string here.
* Is there anything useful we could/should provide?
*/
bzero(statp->f_spare, sizeof (statp->f_spare));
ZFS_EXIT(zsb);
return (0);
}
/*ARGSUSED*/
static int
zfs_vfs_unmount(struct mount *mp, int mntflags, vfs_context_t context)
{
zfsvfs_t *zfsvfs = vfs_fsprivate(mp);
objset_t *os = zfsvfs->z_os;
znode_t *zp, *nextzp;
int ret, i;
int flags;
/*XXX NOEL: delegation admin stuffs, add back if we use delg. admin */
#if 0
ret = 0; /* UNDEFINED: secpolicy_fs_unmount(cr, vfsp); */
if (ret) {
ret = dsl_deleg_access((char *)refstr_value(vfsp->vfs_resource),
ZFS_DELEG_PERM_MOUNT, cr);
if (ret)
return (ret);
}
/*
* We purge the parent filesystem's vfsp as the parent filesystem
* and all of its snapshots have their vnode's v_vfsp set to the
* parent's filesystem's vfsp. Note, 'z_parent' is self
* referential for non-snapshots.
*/
(void) dnlc_purge_vfsp(zfsvfs->z_parent->z_vfs, 0);
#endif
/*
* Unmount any snapshots mounted under .zfs before unmounting the
* dataset itself.
*/
#if 0
if (zfsvfs->z_ctldir != NULL &&
(ret = zfsctl_umount_snapshots(vfsp, fflag, cr)) != 0) {
return (ret);
#endif
flags = SKIPSYSTEM;
if (mntflags & MNT_FORCE)
flags |= FORCECLOSE;
ret = vflush(mp, NULLVP, flags);
/*
* Mac OS X needs a file system modify time
*
* We use the mtime of the "com.apple.system.mtime"
* extended attribute, which is associated with the
* file system root directory.
*
* Here we need to release the ref we took on z_mtime_vp during mount.
*/
if ((ret == 0) || (mntflags & MNT_FORCE)) {
if (zfsvfs->z_mtime_vp != NULL) {
struct vnode *mvp;
mvp = zfsvfs->z_mtime_vp;
zfsvfs->z_mtime_vp = NULL;
if (vnode_get(mvp) == 0) {
vnode_rele(mvp);
vnode_recycle(mvp);
vnode_put(mvp);
}
}
}
if (!(mntflags & MNT_FORCE)) {
/*
* Check the number of active vnodes in the file system.
* Our count is maintained in the vfs structure, but the
* number is off by 1 to indicate a hold on the vfs
* structure itself.
*
* The '.zfs' directory maintains a reference of its
* own, and any active references underneath are
* reflected in the vnode count.
*/
if (ret)
return (EBUSY);
#if 0
if (zfsvfs->z_ctldir == NULL) {
if (vfsp->vfs_count > 1)
return (EBUSY);
} else {
if (vfsp->vfs_count > 2 ||
zfsvfs->z_ctldir->v_count > 1) {
return (EBUSY);
}
}
#endif
}
rw_enter(&zfsvfs->z_unmount_lock, RW_WRITER);
rw_enter(&zfsvfs->z_unmount_inactive_lock, RW_WRITER);
/*
* At this point there are no vops active, and any new vops will
* fail with EIO since we have z_unmount_lock for writer (only
* relavent for forced unmount).
*
* Release all holds on dbufs.
* Note, the dmu can still callback via znode_pageout_func()
* which can zfs_znode_free() the znode. So we lock
* z_all_znodes; search the list for a held dbuf; drop the lock
* (we know zp can't disappear if we hold a dbuf lock) then
* regrab the lock and restart.
*/
mutex_enter(&zfsvfs->z_znodes_lock);
for (zp = list_head(&zfsvfs->z_all_znodes); zp; zp = nextzp) {
nextzp = list_next(&zfsvfs->z_all_znodes, zp);
if (zp->z_dbuf_held) {
/* dbufs should only be held when force unmounting */
zp->z_dbuf_held = 0;
mutex_exit(&zfsvfs->z_znodes_lock);
dmu_buf_rele(zp->z_dbuf, NULL);
/* Start again */
mutex_enter(&zfsvfs->z_znodes_lock);
nextzp = list_head(&zfsvfs->z_all_znodes);
}
}
mutex_exit(&zfsvfs->z_znodes_lock);
/*
* Set the unmounted flag and let new vops unblock.
* zfs_inactive will have the unmounted behavior, and all other
* vops will fail with EIO.
*/
zfsvfs->z_unmounted = B_TRUE;
rw_exit(&zfsvfs->z_unmount_lock);
rw_exit(&zfsvfs->z_unmount_inactive_lock);
/*
* Unregister properties.
*/
#ifndef __APPLE__
if (!dmu_objset_is_snapshot(os))
zfs_unregister_callbacks(zfsvfs);
#endif
/*
* Close the zil. NB: Can't close the zil while zfs_inactive
* threads are blocked as zil_close can call zfs_inactive.
*/
if (zfsvfs->z_log) {
zil_close(zfsvfs->z_log);
zfsvfs->z_log = NULL;
}
/*
* Evict all dbufs so that cached znodes will be freed
*/
if (dmu_objset_evict_dbufs(os, B_TRUE)) {
txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), 0);
(void) dmu_objset_evict_dbufs(os, B_FALSE);
}
/*
* Finally close the objset
*/
dmu_objset_close(os);
/*
* We can now safely destroy the '.zfs' directory node.
*/
#if 0
if (zfsvfs->z_ctldir != NULL)
zfsctl_destroy(zfsvfs);
#endif
/*
* Note that this work is normally done in zfs_freevfs, but since
* there is no VOP_FREEVFS in OSX, we free VFS items here
*/
OSDecrementAtomic((SInt32 *)&zfs_active_fs_count);
for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
mutex_destroy(&zfsvfs->z_hold_mtx[i]);
mutex_destroy(&zfsvfs->z_znodes_lock);
list_destroy(&zfsvfs->z_all_znodes);
rw_destroy(&zfsvfs->z_unmount_lock);
rw_destroy(&zfsvfs->z_unmount_inactive_lock);
return (0);
}
struct vnode* vnode_getparent(struct vnode *vp); /* sys/vnode_internal.h */
static int
zfs_vget_internal(zfsvfs_t *zfsvfs, ino64_t ino, struct vnode **vpp)
{
struct vnode *vp;
struct vnode *dvp = NULL;
znode_t *zp;
int error;
*vpp = NULL;
/*
* On Mac OS X we always export the root directory id as 2
* and its parent as 1
*/
if (ino == 2 || ino == 1)
ino = zfsvfs->z_root;
if ((error = zfs_zget(zfsvfs, ino, &zp)))
goto out;
/* Don't expose EA objects! */
if (zp->z_phys->zp_flags & ZFS_XATTR) {
vnode_put(ZTOV(zp));
error = ENOENT;
goto out;
}
*vpp = vp = ZTOV(zp);
if (vnode_isvroot(vp))
goto out;
/*
* If this znode didn't just come from the cache then
* it won't have a valid identity (parent and name).
*
* Manually fix its identity here (normally done by namei lookup).
*/
if ((dvp = vnode_getparent(vp)) == NULL) {
if (zp->z_phys->zp_parent != 0 &&
zfs_vget_internal(zfsvfs, zp->z_phys->zp_parent, &dvp)) {
goto out;
}
if ( vnode_isdir(dvp) ) {
char objname[ZAP_MAXNAMELEN]; /* 256 bytes */
int flags = VNODE_UPDATE_PARENT;
/* Look for znode's name in its parent's zap */
if ( zap_value_search(zfsvfs->z_os,
zp->z_phys->zp_parent,
zp->z_id,
ZFS_DIRENT_OBJ(-1ULL),
objname) == 0 ) {
flags |= VNODE_UPDATE_NAME;
}
/* Update the znode's parent and name */
vnode_update_identity(vp, dvp, objname, 0, 0, flags);
}
}
/* All done with znode's parent */
vnode_put(dvp);
out:
return (error);
}
/*
* Get a vnode from a file id (ignoring the generation)
*
* Use by NFS Server (readdirplus) and VFS (build_path)
*/
static int
zfs_vfs_vget(struct mount *mp, ino64_t ino, struct vnode **vpp, __unused vfs_context_t context)
{
zfsvfs_t *zfsvfs = vfs_fsprivate(mp);
int error;
ZFS_ENTER(zfsvfs);
/*
* On Mac OS X we always export the root directory id as 2.
* So we don't expect to see the real root directory id
* from zfs_vfs_vget KPI (unless of course the real id was
* already 2).
*/
if ((ino == zfsvfs->z_root) && (zfsvfs->z_root != 2)) {
ZFS_EXIT(zfsvfs);
return (ENOENT);
}
error = zfs_vget_internal(zfsvfs, ino, vpp);
ZFS_EXIT(zfsvfs);
return (error);
}
