Page MenuHomeFreeBSD
Files F106105377

D7421.id25481.diff
No OneTemporary
Actions

D7421.id25481.diff
View Options

Index: head/sys/cddl/compat/opensolaris/kern/opensolaris_vfs.c
===================================================================
--- head/sys/cddl/compat/opensolaris/kern/opensolaris_vfs.c
+++ head/sys/cddl/compat/opensolaris/kern/opensolaris_vfs.c
@@ -196,10 +196,17 @@
td->td_ucred = cr;
if (error != 0) {
+ /*
+ * Clear VI_MOUNT and decrement the use count "atomically",
+ * under the vnode lock. This is not strictly required,
+ * but makes it easier to reason about the life-cycle and
+ * ownership of the covered vnode.
+ */
+ vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
VI_LOCK(vp);
vp->v_iflag &= ~VI_MOUNT;
VI_UNLOCK(vp);
- vrele(vp);
+ vput(vp);
vfs_unbusy(mp);
vfs_freeopts(mp->mnt_optnew);
vfs_mount_destroy(mp);
Index: head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/zfs_ctldir.h
===================================================================
--- head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/zfs_ctldir.h
+++ head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/zfs_ctldir.h
@@ -44,7 +44,7 @@
void zfsctl_create(zfsvfs_t *);
void zfsctl_destroy(zfsvfs_t *);
-vnode_t *zfsctl_root(znode_t *);
+int zfsctl_root(zfsvfs_t *, int, vnode_t **);
void zfsctl_init(void);
void zfsctl_fini(void);
boolean_t zfsctl_is_node(vnode_t *);
@@ -53,10 +53,6 @@
int zfsctl_destroy_snapshot(const char *snapname, int force);
int zfsctl_umount_snapshots(vfs_t *, int, cred_t *);
-int zfsctl_root_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, pathname_t *pnp,
- int flags, vnode_t *rdir, cred_t *cr, caller_context_t *ct,
- int *direntflags, pathname_t *realpnp);
-
int zfsctl_lookup_objset(vfs_t *vfsp, uint64_t objsetid, zfsvfs_t **zfsvfsp);
#define ZFSCTL_INO_ROOT 0x1
Index: head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/zfs_vfsops.h
===================================================================
--- head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/zfs_vfsops.h
+++ head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/zfs_vfsops.h
@@ -68,7 +68,7 @@
krwlock_t z_teardown_inactive_lock;
list_t z_all_znodes; /* all vnodes in the fs */
kmutex_t z_znodes_lock; /* lock for z_all_znodes */
- vnode_t *z_ctldir; /* .zfs directory pointer */
+ struct zfsctl_root *z_ctldir; /* .zfs directory pointer */
boolean_t z_show_ctldir; /* expose .zfs in the root dir */
boolean_t z_issnap; /* true if this is a snapshot */
boolean_t z_vscan; /* virus scan on/off */
Index: head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_ctldir.c
===================================================================
--- head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_ctldir.c
+++ head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_ctldir.c
@@ -70,136 +70,249 @@
#include <sys/zfs_ioctl.h>
#include <sys/zfs_vfsops.h>
#include <sys/namei.h>
-#include <sys/gfs.h>
#include <sys/stat.h>
#include <sys/dmu.h>
+#include <sys/dsl_dataset.h>
#include <sys/dsl_destroy.h>
#include <sys/dsl_deleg.h>
#include <sys/mount.h>
-#include <sys/sunddi.h>
+#include <sys/zap.h>
#include "zfs_namecheck.h"
-typedef struct zfsctl_node {
- gfs_dir_t zc_gfs_private;
- uint64_t zc_id;
- timestruc_t zc_cmtime; /* ctime and mtime, always the same */
-} zfsctl_node_t;
-
-typedef struct zfsctl_snapdir {
- zfsctl_node_t sd_node;
- kmutex_t sd_lock;
- avl_tree_t sd_snaps;
-} zfsctl_snapdir_t;
+/*
+ * "Synthetic" filesystem implementation.
+ */
-typedef struct {
- char *se_name;
- vnode_t *se_root;
- avl_node_t se_node;
-} zfs_snapentry_t;
-
-static int
-snapentry_compare(const void *a, const void *b)
-{
- const zfs_snapentry_t *sa = a;
- const zfs_snapentry_t *sb = b;
- int ret = strcmp(sa->se_name, sb->se_name);
-
- if (ret < 0)
- return (-1);
- else if (ret > 0)
- return (1);
- else
- return (0);
-}
-
-#ifdef illumos
-vnodeops_t *zfsctl_ops_root;
-vnodeops_t *zfsctl_ops_snapdir;
-vnodeops_t *zfsctl_ops_snapshot;
-vnodeops_t *zfsctl_ops_shares;
-vnodeops_t *zfsctl_ops_shares_dir;
-
-static const fs_operation_def_t zfsctl_tops_root[];
-static const fs_operation_def_t zfsctl_tops_snapdir[];
-static const fs_operation_def_t zfsctl_tops_snapshot[];
-static const fs_operation_def_t zfsctl_tops_shares[];
-#else
-static struct vop_vector zfsctl_ops_root;
-static struct vop_vector zfsctl_ops_snapdir;
-static struct vop_vector zfsctl_ops_snapshot;
-static struct vop_vector zfsctl_ops_shares;
-static struct vop_vector zfsctl_ops_shares_dir;
-#endif
+/*
+ * Assert that A implies B.
+ */
+#define KASSERT_IMPLY(A, B, msg) KASSERT(!(A) || (B), (msg));
-static vnode_t *zfsctl_mknode_snapdir(vnode_t *);
-static vnode_t *zfsctl_mknode_shares(vnode_t *);
-static vnode_t *zfsctl_snapshot_mknode(vnode_t *, uint64_t objset);
-static int zfsctl_unmount_snap(zfs_snapentry_t *, int, cred_t *);
-
-#ifdef illumos
-static gfs_opsvec_t zfsctl_opsvec[] = {
- { ".zfs", zfsctl_tops_root, &zfsctl_ops_root },
- { ".zfs/snapshot", zfsctl_tops_snapdir, &zfsctl_ops_snapdir },
- { ".zfs/snapshot/vnode", zfsctl_tops_snapshot, &zfsctl_ops_snapshot },
- { ".zfs/shares", zfsctl_tops_shares, &zfsctl_ops_shares_dir },
- { ".zfs/shares/vnode", zfsctl_tops_shares, &zfsctl_ops_shares },
- { NULL }
-};
-#endif
+static MALLOC_DEFINE(M_SFSNODES, "sfs_nodes", "synthetic-fs nodes");
+
+typedef struct sfs_node {
+ char sn_name[ZFS_MAX_DATASET_NAME_LEN];
+ uint64_t sn_parent_id;
+ uint64_t sn_id;
+} sfs_node_t;
/*
- * Root directory elements. We only have two entries
- * snapshot and shares.
+ * Check the parent's ID as well as the node's to account for a chance
+ * that IDs originating from different domains (snapshot IDs, artifical
+ * IDs, znode IDs) may clash.
*/
-static gfs_dirent_t zfsctl_root_entries[] = {
- { "snapshot", zfsctl_mknode_snapdir, GFS_CACHE_VNODE },
- { "shares", zfsctl_mknode_shares, GFS_CACHE_VNODE },
- { NULL }
-};
+static int
+sfs_compare_ids(struct vnode *vp, void *arg)
+{
+ sfs_node_t *n1 = vp->v_data;
+ sfs_node_t *n2 = arg;
+ bool equal;
+
+ equal = n1->sn_id == n2->sn_id &&
+ n1->sn_parent_id == n2->sn_parent_id;
+
+ /* Zero means equality. */
+ return (!equal);
+}
+
+static int
+sfs_vnode_get(const struct mount *mp, int flags, uint64_t parent_id,
+ uint64_t id, struct vnode **vpp)
+{
+ sfs_node_t search;
+ int err;
+
+ search.sn_id = id;
+ search.sn_parent_id = parent_id;
+ err = vfs_hash_get(mp, (u_int)id, flags, curthread, vpp,
+ sfs_compare_ids, &search);
+ return (err);
+}
+
+static int
+sfs_vnode_insert(struct vnode *vp, int flags, uint64_t parent_id,
+ uint64_t id, struct vnode **vpp)
+{
+ int err;
+
+ KASSERT(vp->v_data != NULL, ("sfs_vnode_insert with NULL v_data"));
+ err = vfs_hash_insert(vp, (u_int)id, flags, curthread, vpp,
+ sfs_compare_ids, vp->v_data);
+ return (err);
+}
+
+static void
+sfs_vnode_remove(struct vnode *vp)
+{
+ vfs_hash_remove(vp);
+}
+
+typedef void sfs_vnode_setup_fn(vnode_t *vp, void *arg);
+
+static int
+sfs_vgetx(struct mount *mp, int flags, uint64_t parent_id, uint64_t id,
+ const char *tag, struct vop_vector *vops,
+ sfs_vnode_setup_fn setup, void *arg,
+ struct vnode **vpp)
+{
+ struct vnode *vp;
+ int error;
+
+ error = sfs_vnode_get(mp, flags, parent_id, id, vpp);
+ if (error != 0 || *vpp != NULL) {
+ KASSERT_IMPLY(error == 0, (*vpp)->v_data != NULL,
+ "sfs vnode with no data");
+ return (error);
+ }
+
+ /* Allocate a new vnode/inode. */
+ error = getnewvnode(tag, mp, vops, &vp);
+ if (error != 0) {
+ *vpp = NULL;
+ return (error);
+ }
+
+ /*
+ * Exclusively lock the vnode vnode while it's being constructed.
+ */
+ lockmgr(vp->v_vnlock, LK_EXCLUSIVE, NULL);
+ error = insmntque(vp, mp);
+ if (error != 0) {
+ *vpp = NULL;
+ return (error);
+ }
+
+ setup(vp, arg);
+
+ error = sfs_vnode_insert(vp, flags, parent_id, id, vpp);
+ if (error != 0 || *vpp != NULL) {
+ KASSERT_IMPLY(error == 0, (*vpp)->v_data != NULL,
+ "sfs vnode with no data");
+ return (error);
+ }
+
+ *vpp = vp;
+ return (0);
+}
+
+static void
+sfs_print_node(sfs_node_t *node)
+{
+ printf("\tname = %s\n", node->sn_name);
+ printf("\tparent_id = %ju\n", (uintmax_t)node->sn_parent_id);
+ printf("\tid = %ju\n", (uintmax_t)node->sn_id);
+}
+
+static sfs_node_t *
+sfs_alloc_node(size_t size, const char *name, uint64_t parent_id, uint64_t id)
+{
+ struct sfs_node *node;
+
+ KASSERT(strlen(name) < sizeof(node->sn_name),
+ ("sfs node name is too long"));
+ KASSERT(size >= sizeof(*node), ("sfs node size is too small"));
+ node = malloc(size, M_SFSNODES, M_WAITOK | M_ZERO);
+ strlcpy(node->sn_name, name, sizeof(node->sn_name));
+ node->sn_parent_id = parent_id;
+ node->sn_id = id;
+
+ return (node);
+}
+
+static void
+sfs_destroy_node(sfs_node_t *node)
+{
+ free(node, M_SFSNODES);
+}
+
+static void *
+sfs_reclaim_vnode(vnode_t *vp)
+{
+ sfs_node_t *node;
+ void *data;
+
+ sfs_vnode_remove(vp);
+ data = vp->v_data;
+ vp->v_data = NULL;
+ return (data);
+}
+
+static int
+sfs_readdir_common(uint64_t parent_id, uint64_t id, struct vop_readdir_args *ap,
+ uio_t *uio, off_t *offp)
+{
+ struct dirent entry;
+ int error;
+
+ /* Reset ncookies for subsequent use of vfs_read_dirent. */
+ if (ap->a_ncookies != NULL)
+ *ap->a_ncookies = 0;
+
+ if (uio->uio_resid < sizeof(entry))
+ return (SET_ERROR(EINVAL));
+
+ if (uio->uio_offset < 0)
+ return (SET_ERROR(EINVAL));
+ if (uio->uio_offset == 0) {
+ entry.d_fileno = id;
+ entry.d_type = DT_DIR;
+ entry.d_name[0] = '.';
+ entry.d_name[1] = '\0';
+ entry.d_namlen = 1;
+ entry.d_reclen = sizeof(entry);
+ error = vfs_read_dirent(ap, &entry, uio->uio_offset);
+ if (error != 0)
+ return (SET_ERROR(error));
+ }
+
+ if (uio->uio_offset < sizeof(entry))
+ return (SET_ERROR(EINVAL));
+ if (uio->uio_offset == sizeof(entry)) {
+ entry.d_fileno = parent_id;
+ entry.d_type = DT_DIR;
+ entry.d_name[0] = '.';
+ entry.d_name[1] = '.';
+ entry.d_name[2] = '\0';
+ entry.d_namlen = 2;
+ entry.d_reclen = sizeof(entry);
+ error = vfs_read_dirent(ap, &entry, uio->uio_offset);
+ if (error != 0)
+ return (SET_ERROR(error));
+ }
-/* include . and .. in the calculation */
-#define NROOT_ENTRIES ((sizeof (zfsctl_root_entries) / \
- sizeof (gfs_dirent_t)) + 1)
+ if (offp != NULL)
+ *offp = 2 * sizeof(entry);
+ return (0);
+}
/*
- * Initialize the various GFS pieces we'll need to create and manipulate .zfs
- * directories. This is called from the ZFS init routine, and initializes the
- * vnode ops vectors that we'll be using.
+ * .zfs inode namespace
+ *
+ * We need to generate unique inode numbers for all files and directories
+ * within the .zfs pseudo-filesystem. We use the following scheme:
+ *
+ * ENTRY ZFSCTL_INODE
+ * .zfs 1
+ * .zfs/snapshot 2
+ * .zfs/snapshot/<snap> objectid(snap)
*/
+#define ZFSCTL_INO_SNAP(id) (id)
+
+static struct vop_vector zfsctl_ops_root;
+static struct vop_vector zfsctl_ops_snapdir;
+static struct vop_vector zfsctl_ops_snapshot;
+static struct vop_vector zfsctl_ops_shares_dir;
+
void
zfsctl_init(void)
{
-#ifdef illumos
- VERIFY(gfs_make_opsvec(zfsctl_opsvec) == 0);
-#endif
}
void
zfsctl_fini(void)
{
-#ifdef illumos
- /*
- * Remove vfsctl vnode ops
- */
- if (zfsctl_ops_root)
- vn_freevnodeops(zfsctl_ops_root);
- if (zfsctl_ops_snapdir)
- vn_freevnodeops(zfsctl_ops_snapdir);
- if (zfsctl_ops_snapshot)
- vn_freevnodeops(zfsctl_ops_snapshot);
- if (zfsctl_ops_shares)
- vn_freevnodeops(zfsctl_ops_shares);
- if (zfsctl_ops_shares_dir)
- vn_freevnodeops(zfsctl_ops_shares_dir);
-
- zfsctl_ops_root = NULL;
- zfsctl_ops_snapdir = NULL;
- zfsctl_ops_snapshot = NULL;
- zfsctl_ops_shares = NULL;
- zfsctl_ops_shares_dir = NULL;
-#endif /* illumos */
}
boolean_t
@@ -208,106 +321,114 @@
return (vn_matchops(vp, zfsctl_ops_root) ||
vn_matchops(vp, zfsctl_ops_snapdir) ||
vn_matchops(vp, zfsctl_ops_snapshot) ||
- vn_matchops(vp, zfsctl_ops_shares) ||
vn_matchops(vp, zfsctl_ops_shares_dir));
}
-/*
- * Return the inode number associated with the 'snapshot' or
- * 'shares' directory.
- */
-/* ARGSUSED */
-static ino64_t
-zfsctl_root_inode_cb(vnode_t *vp, int index)
-{
- zfsvfs_t *zfsvfs = vp->v_vfsp->vfs_data;
-
- ASSERT(index < 2);
-
- if (index == 0)
- return (ZFSCTL_INO_SNAPDIR);
+typedef struct zfsctl_root {
+ sfs_node_t node;
+ sfs_node_t *snapdir;
+ timestruc_t cmtime;
+} zfsctl_root_t;
- return (zfsvfs->z_shares_dir);
-}
/*
- * Create the '.zfs' directory. This directory is cached as part of the VFS
- * structure. This results in a hold on the vfs_t. The code in zfs_umount()
- * therefore checks against a vfs_count of 2 instead of 1. This reference
- * is removed when the ctldir is destroyed in the unmount.
+ * Create the '.zfs' directory.
*/
void
zfsctl_create(zfsvfs_t *zfsvfs)
{
- vnode_t *vp, *rvp;
- zfsctl_node_t *zcp;
+ zfsctl_root_t *dot_zfs;
+ sfs_node_t *snapdir;
+ vnode_t *rvp;
uint64_t crtime[2];
ASSERT(zfsvfs->z_ctldir == NULL);
- vp = gfs_root_create(sizeof (zfsctl_node_t), zfsvfs->z_vfs,
- &zfsctl_ops_root, ZFSCTL_INO_ROOT, zfsctl_root_entries,
- zfsctl_root_inode_cb, MAXNAMELEN, NULL, NULL);
- zcp = vp->v_data;
- zcp->zc_id = ZFSCTL_INO_ROOT;
+ snapdir = sfs_alloc_node(sizeof(*snapdir), "snapshot", ZFSCTL_INO_ROOT,
+ ZFSCTL_INO_SNAPDIR);
+ dot_zfs = (zfsctl_root_t *)sfs_alloc_node(sizeof(*dot_zfs), ".zfs", 0,
+ ZFSCTL_INO_ROOT);
+ dot_zfs->snapdir = snapdir;
VERIFY(VFS_ROOT(zfsvfs->z_vfs, LK_EXCLUSIVE, &rvp) == 0);
VERIFY(0 == sa_lookup(VTOZ(rvp)->z_sa_hdl, SA_ZPL_CRTIME(zfsvfs),
- &crtime, sizeof (crtime)));
- ZFS_TIME_DECODE(&zcp->zc_cmtime, crtime);
- VN_URELE(rvp);
-
- /*
- * We're only faking the fact that we have a root of a filesystem for
- * the sake of the GFS interfaces. Undo the flag manipulation it did
- * for us.
- */
- vp->v_vflag &= ~VV_ROOT;
+ &crtime, sizeof(crtime)));
+ ZFS_TIME_DECODE(&dot_zfs->cmtime, crtime);
+ vput(rvp);
- zfsvfs->z_ctldir = vp;
-
- VOP_UNLOCK(vp, 0);
+ zfsvfs->z_ctldir = dot_zfs;
}
/*
* Destroy the '.zfs' directory. Only called when the filesystem is unmounted.
- * There might still be more references if we were force unmounted, but only
- * new zfs_inactive() calls can occur and they don't reference .zfs
+ * The nodes must not have any associated vnodes by now as they should be
+ * vflush-ed.
*/
void
zfsctl_destroy(zfsvfs_t *zfsvfs)
{
- VN_RELE(zfsvfs->z_ctldir);
+ sfs_destroy_node(zfsvfs->z_ctldir->snapdir);
+ sfs_destroy_node((sfs_node_t *)zfsvfs->z_ctldir);
zfsvfs->z_ctldir = NULL;
}
-/*
- * Given a root znode, retrieve the associated .zfs directory.
- * Add a hold to the vnode and return it.
- */
-vnode_t *
-zfsctl_root(znode_t *zp)
+static int
+zfsctl_fs_root_vnode(struct mount *mp, void *arg __unused, int flags,
+ struct vnode **vpp)
+{
+ return (VFS_ROOT(mp, flags, vpp));
+}
+
+static void
+zfsctl_common_vnode_setup(vnode_t *vp, void *arg)
{
- ASSERT(zfs_has_ctldir(zp));
- VN_HOLD(zp->z_zfsvfs->z_ctldir);
- return (zp->z_zfsvfs->z_ctldir);
+ ASSERT_VOP_ELOCKED(vp, __func__);
+
+ /* We support shared locking. */
+ VN_LOCK_ASHARE(vp);
+ vp->v_type = VDIR;
+ vp->v_data = arg;
}
static int
-zfsctl_common_print(ap)
- struct vop_print_args /* {
- struct vnode *a_vp;
- } */ *ap;
+zfsctl_root_vnode(struct mount *mp, void *arg __unused, int flags,
+ struct vnode **vpp)
{
- vnode_t *vp = ap->a_vp;
- gfs_file_t *fp = vp->v_data;
+ void *node;
+ int err;
- printf(" parent = %p\n", fp->gfs_parent);
- printf(" type = %d\n", fp->gfs_type);
- printf(" index = %d\n", fp->gfs_index);
- printf(" ino = %ju\n", (uintmax_t)fp->gfs_ino);
- return (0);
+ node = ((zfsvfs_t*)mp->mnt_data)->z_ctldir;
+ err = sfs_vgetx(mp, flags, 0, ZFSCTL_INO_ROOT, "zfs", &zfsctl_ops_root,
+ zfsctl_common_vnode_setup, node, vpp);
+ return (err);
+}
+
+static int
+zfsctl_snapdir_vnode(struct mount *mp, void *arg __unused, int flags,
+ struct vnode **vpp)
+{
+ void *node;
+ int err;
+
+ node = ((zfsvfs_t*)mp->mnt_data)->z_ctldir->snapdir;
+ err = sfs_vgetx(mp, flags, ZFSCTL_INO_ROOT, ZFSCTL_INO_SNAPDIR, "zfs",
+ &zfsctl_ops_snapdir, zfsctl_common_vnode_setup, node, vpp);
+ return (err);
+}
+
+/*
+ * Given a root znode, retrieve the associated .zfs directory.
+ * Add a hold to the vnode and return it.
+ */
+int
+zfsctl_root(zfsvfs_t *zfsvfs, int flags, vnode_t **vpp)
+{
+ vnode_t *vp;
+ int error;
+
+ error = zfsctl_root_vnode(zfsvfs->z_vfs, NULL, flags, vpp);
+ return (error);
}
/*
@@ -350,18 +471,8 @@
{
accmode_t accmode = ap->a_accmode;
-#ifdef TODO
- if (flags & V_ACE_MASK) {
- if (accmode & ACE_ALL_WRITE_PERMS)
- return (SET_ERROR(EACCES));
- } else {
-#endif
- if (accmode & VWRITE)
- return (SET_ERROR(EACCES));
-#ifdef TODO
- }
-#endif
-
+ if (accmode & VWRITE)
+ return (SET_ERROR(EACCES));
return (0);
}
@@ -372,6 +483,9 @@
zfsctl_common_getattr(vnode_t *vp, vattr_t *vap)
{
timestruc_t now;
+ sfs_node_t *node;
+
+ node = vp->v_data;
vap->va_uid = 0;
vap->va_gid = 0;
@@ -394,6 +508,11 @@
vap->va_atime = now;
/* FreeBSD: Reset chflags(2) flags. */
vap->va_flags = 0;
+
+ vap->va_nodeid = node->sn_id;
+
+ /* At least '.' and '..'. */
+ vap->va_nlink = 2;
}
/*ARGSUSED*/
@@ -406,81 +525,46 @@
{
vnode_t *vp = ap->a_vp;
fid_t *fidp = (void *)ap->a_fid;
- zfsvfs_t *zfsvfs = vp->v_vfsp->vfs_data;
- zfsctl_node_t *zcp = vp->v_data;
- uint64_t object = zcp->zc_id;
+ sfs_node_t *node = vp->v_data;
+ uint64_t object = node->sn_id;
zfid_short_t *zfid;
int i;
- ZFS_ENTER(zfsvfs);
-
-#ifdef illumos
- if (fidp->fid_len < SHORT_FID_LEN) {
- fidp->fid_len = SHORT_FID_LEN;
- ZFS_EXIT(zfsvfs);
- return (SET_ERROR(ENOSPC));
- }
-#endif
-
zfid = (zfid_short_t *)fidp;
-
zfid->zf_len = SHORT_FID_LEN;
- for (i = 0; i < sizeof (zfid->zf_object); i++)
+ for (i = 0; i < sizeof(zfid->zf_object); i++)
zfid->zf_object[i] = (uint8_t)(object >> (8 * i));
- /* .zfs znodes always have a generation number of 0 */
- for (i = 0; i < sizeof (zfid->zf_gen); i++)
+ /* .zfs nodes always have a generation number of 0 */
+ for (i = 0; i < sizeof(zfid->zf_gen); i++)
zfid->zf_gen[i] = 0;
- ZFS_EXIT(zfsvfs);
return (0);
}
-
-/*ARGSUSED*/
static int
-zfsctl_shares_fid(ap)
- struct vop_fid_args /* {
+zfsctl_common_reclaim(ap)
+ struct vop_reclaim_args /* {
struct vnode *a_vp;
- struct fid *a_fid;
+ struct thread *a_td;
} */ *ap;
{
- vnode_t *vp = ap->a_vp;
- fid_t *fidp = (void *)ap->a_fid;
- 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 (SET_ERROR(ENOTSUP));
- }
-
- if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &dzp)) == 0) {
- error = VOP_FID(ZTOV(dzp), fidp);
- VN_RELE(ZTOV(dzp));
- }
+ vnode_t *vp = ap->a_vp;
- ZFS_EXIT(zfsvfs);
- return (error);
+ (void) sfs_reclaim_vnode(vp);
+ return (0);
}
-/*
- * .zfs inode namespace
- *
- * We need to generate unique inode numbers for all files and directories
- * within the .zfs pseudo-filesystem. We use the following scheme:
- *
- * ENTRY ZFSCTL_INODE
- * .zfs 1
- * .zfs/snapshot 2
- * .zfs/snapshot/<snap> objectid(snap)
- */
-
-#define ZFSCTL_INO_SNAP(id) (id)
+static int
+zfsctl_common_print(ap)
+ struct vop_print_args /* {
+ struct vnode *a_vp;
+ } */ *ap;
+{
+ sfs_print_node(ap->a_vp->v_data);
+ return (0);
+}
/*
* Get root directory attributes.
@@ -496,156 +580,132 @@
{
struct vnode *vp = ap->a_vp;
struct vattr *vap = ap->a_vap;
- zfsvfs_t *zfsvfs = vp->v_vfsp->vfs_data;
- zfsctl_node_t *zcp = vp->v_data;
-
- ZFS_ENTER(zfsvfs);
- vap->va_nodeid = ZFSCTL_INO_ROOT;
- vap->va_nlink = vap->va_size = NROOT_ENTRIES;
- vap->va_mtime = vap->va_ctime = zcp->zc_cmtime;
- vap->va_birthtime = vap->va_ctime;
+ zfsctl_root_t *node = vp->v_data;
zfsctl_common_getattr(vp, vap);
- ZFS_EXIT(zfsvfs);
-
+ vap->va_ctime = node->cmtime;
+ vap->va_mtime = vap->va_ctime;
+ vap->va_birthtime = vap->va_ctime;
+ vap->va_nlink += 1; /* snapdir */
+ vap->va_size = vap->va_nlink;
return (0);
}
/*
- * Special case the handling of "..".
+ * When we lookup "." we still can be asked to lock it
+ * differently, can't we?
*/
-/* ARGSUSED */
int
-zfsctl_root_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, pathname_t *pnp,
- int flags, vnode_t *rdir, cred_t *cr, caller_context_t *ct,
- int *direntflags, pathname_t *realpnp)
+zfsctl_relock_dot(vnode_t *dvp, int ltype)
{
- zfsvfs_t *zfsvfs = dvp->v_vfsp->vfs_data;
- int err;
-
- /*
- * No extended attributes allowed under .zfs
- */
- if (flags & LOOKUP_XATTR)
- return (SET_ERROR(EINVAL));
-
- ZFS_ENTER(zfsvfs);
-
- if (strcmp(nm, "..") == 0) {
-#ifdef illumos
- err = VFS_ROOT(dvp->v_vfsp, LK_EXCLUSIVE, vpp);
-#else
- /*
- * NB: can not use VFS_ROOT here as it would acquire
- * the vnode lock of the parent (root) vnode while
- * holding the child's (.zfs) lock.
- */
- znode_t *rootzp;
-
- err = zfs_zget(zfsvfs, zfsvfs->z_root, &rootzp);
- if (err == 0)
- *vpp = ZTOV(rootzp);
-#endif
- } else {
- err = gfs_vop_lookup(dvp, nm, vpp, pnp, flags, rdir,
- cr, ct, direntflags, realpnp);
+ vref(dvp);
+ if (ltype != VOP_ISLOCKED(dvp)) {
+ if (ltype == LK_EXCLUSIVE)
+ vn_lock(dvp, LK_UPGRADE | LK_RETRY);
+ else /* if (ltype == LK_SHARED) */
+ vn_lock(dvp, LK_DOWNGRADE | LK_RETRY);
+
+ /* Relock for the "." case may left us with reclaimed vnode. */
+ if ((dvp->v_iflag & VI_DOOMED) != 0) {
+ vrele(dvp);
+ return (SET_ERROR(ENOENT));
+ }
}
-
- ZFS_EXIT(zfsvfs);
-
- return (err);
+ return (0);
}
-static int
-zfsctl_freebsd_root_lookup(ap)
+/*
+ * Special case the handling of "..".
+ */
+int
+zfsctl_root_lookup(ap)
struct vop_lookup_args /* {
struct vnode *a_dvp;
struct vnode **a_vpp;
struct componentname *a_cnp;
} */ *ap;
{
+ struct componentname *cnp = ap->a_cnp;
vnode_t *dvp = ap->a_dvp;
vnode_t **vpp = ap->a_vpp;
cred_t *cr = ap->a_cnp->cn_cred;
int flags = ap->a_cnp->cn_flags;
int lkflags = ap->a_cnp->cn_lkflags;
int nameiop = ap->a_cnp->cn_nameiop;
- char nm[NAME_MAX + 1];
int err;
+ int ltype;
- if ((flags & ISLASTCN) && (nameiop == RENAME || nameiop == CREATE))
- return (EOPNOTSUPP);
+ ASSERT(dvp->v_type == VDIR);
- ASSERT(ap->a_cnp->cn_namelen < sizeof(nm));
- strlcpy(nm, ap->a_cnp->cn_nameptr, ap->a_cnp->cn_namelen + 1);
-relookup:
- err = zfsctl_root_lookup(dvp, nm, vpp, NULL, 0, NULL, cr, NULL, NULL, NULL);
- if (err == 0 && (nm[0] != '.' || nm[1] != '\0')) {
- if (flags & ISDOTDOT) {
- VOP_UNLOCK(dvp, 0);
- err = vn_lock(*vpp, lkflags);
- if (err != 0) {
- vrele(*vpp);
- *vpp = NULL;
- }
- vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY);
- } else {
- err = vn_lock(*vpp, LK_EXCLUSIVE);
- if (err != 0) {
- VERIFY3S(err, ==, ENOENT);
- goto relookup;
- }
- }
- }
- return (err);
-}
+ if ((flags & ISLASTCN) != 0 && nameiop != LOOKUP)
+ return (SET_ERROR(ENOTSUP));
-static int
-zfsctl_root_print(ap)
- struct vop_print_args /* {
+ if (cnp->cn_namelen == 1 && *cnp->cn_nameptr == '.') {
+ err = zfsctl_relock_dot(dvp, lkflags & LK_TYPE_MASK);
+ if (err == 0)
+ *vpp = dvp;
+ } else if ((flags & ISDOTDOT) != 0) {
+ err = vn_vget_ino_gen(dvp, zfsctl_fs_root_vnode, NULL,
+ lkflags, vpp);
+ } else if (strncmp(cnp->cn_nameptr, "snapshot", cnp->cn_namelen) == 0) {
+ err = zfsctl_snapdir_vnode(dvp->v_mount, NULL, lkflags, vpp);
+ } else {
+ err = SET_ERROR(ENOENT);
+ }
+ if (err != 0)
+ *vpp = NULL;
+ return (err);
+}
+
+static int
+zfsctl_root_readdir(ap)
+ struct vop_readdir_args /* {
struct vnode *a_vp;
+ struct uio *a_uio;
+ struct ucred *a_cred;
+ int *a_eofflag;
+ int *ncookies;
+ u_long **a_cookies;
} */ *ap;
{
- printf(" .zfs node\n");
- zfsctl_common_print(ap);
+ struct dirent entry;
+ vnode_t *vp = ap->a_vp;
+ zfsvfs_t *zfsvfs = vp->v_vfsp->vfs_data;
+ zfsctl_root_t *node = vp->v_data;
+ uio_t *uio = ap->a_uio;
+ int *eofp = ap->a_eofflag;
+ off_t dots_offset;
+ int error;
+
+ ASSERT(vp->v_type == VDIR);
+
+ error = sfs_readdir_common(zfsvfs->z_root, ZFSCTL_INO_ROOT, ap, uio,
+ &dots_offset);
+ if (error != 0) {
+ if (error == ENAMETOOLONG) /* ran out of destination space */
+ error = 0;
+ return (error);
+ }
+ if (uio->uio_offset != dots_offset)
+ return (SET_ERROR(EINVAL));
+
+ CTASSERT(sizeof(node->snapdir->sn_name) <= sizeof(entry.d_name));
+ entry.d_fileno = node->snapdir->sn_id;
+ entry.d_type = DT_DIR;
+ strcpy(entry.d_name, node->snapdir->sn_name);
+ entry.d_namlen = strlen(entry.d_name);
+ entry.d_reclen = sizeof(entry);
+ error = vfs_read_dirent(ap, &entry, uio->uio_offset);
+ if (error != 0) {
+ if (error == ENAMETOOLONG)
+ error = 0;
+ return (SET_ERROR(error));
+ }
+ if (eofp != NULL)
+ *eofp = 1;
return (0);
}
-#ifdef illumos
-static int
-zfsctl_pathconf(vnode_t *vp, int cmd, ulong_t *valp, cred_t *cr,
- caller_context_t *ct)
-{
- /*
- * We only care about ACL_ENABLED so that libsec can
- * display ACL correctly and not default to POSIX draft.
- */
- if (cmd == _PC_ACL_ENABLED) {
- *valp = _ACL_ACE_ENABLED;
- return (0);
- }
-
- return (fs_pathconf(vp, cmd, valp, cr, ct));
-}
-#endif /* illumos */
-
-#ifdef illumos
-static const fs_operation_def_t zfsctl_tops_root[] = {
- { VOPNAME_OPEN, { .vop_open = zfsctl_common_open } },
- { VOPNAME_CLOSE, { .vop_close = zfsctl_common_close } },
- { VOPNAME_IOCTL, { .error = fs_inval } },
- { VOPNAME_GETATTR, { .vop_getattr = zfsctl_root_getattr } },
- { VOPNAME_ACCESS, { .vop_access = zfsctl_common_access } },
- { VOPNAME_READDIR, { .vop_readdir = gfs_vop_readdir } },
- { VOPNAME_LOOKUP, { .vop_lookup = zfsctl_root_lookup } },
- { VOPNAME_SEEK, { .vop_seek = fs_seek } },
- { VOPNAME_INACTIVE, { .vop_inactive = gfs_vop_inactive } },
- { VOPNAME_PATHCONF, { .vop_pathconf = zfsctl_pathconf } },
- { VOPNAME_FID, { .vop_fid = zfsctl_common_fid } },
- { NULL }
-};
-#endif /* illumos */
-
static struct vop_vector zfsctl_ops_root = {
.vop_default = &default_vnodeops,
.vop_open = zfsctl_common_open,
@@ -653,29 +713,19 @@
.vop_ioctl = VOP_EINVAL,
.vop_getattr = zfsctl_root_getattr,
.vop_access = zfsctl_common_access,
- .vop_readdir = gfs_vop_readdir,
- .vop_lookup = zfsctl_freebsd_root_lookup,
+ .vop_readdir = zfsctl_root_readdir,
+ .vop_lookup = zfsctl_root_lookup,
.vop_inactive = VOP_NULL,
- .vop_reclaim = gfs_vop_reclaim,
-#ifdef TODO
- .vop_pathconf = zfsctl_pathconf,
-#endif
+ .vop_reclaim = zfsctl_common_reclaim,
.vop_fid = zfsctl_common_fid,
- .vop_print = zfsctl_root_print,
+ .vop_print = zfsctl_common_print,
};
-/*
- * Gets the full dataset name that corresponds to the given snapshot name
- * Example:
- * zfsctl_snapshot_zname("snap1") -> "mypool/myfs@snap1"
- */
static int
zfsctl_snapshot_zname(vnode_t *vp, const char *name, int len, char *zname)
{
objset_t *os = ((zfsvfs_t *)((vp)->v_vfsp->vfs_data))->z_os;
- if (zfs_component_namecheck(name, NULL, NULL) != 0)
- return (SET_ERROR(EILSEQ));
dmu_objset_name(os, zname);
if (strlen(zname) + 1 + strlen(name) >= len)
return (SET_ERROR(ENAMETOOLONG));
@@ -685,272 +735,62 @@
}
static int
-zfsctl_unmount_snap(zfs_snapentry_t *sep, int fflags, cred_t *cr)
+zfsctl_snapshot_lookup(vnode_t *vp, const char *name, uint64_t *id)
{
- vnode_t *svp = sep->se_root;
- int error;
-
- ASSERT(vn_ismntpt(svp));
-
- /* this will be dropped by dounmount() */
- if ((error = vn_vfswlock(svp)) != 0)
- return (error);
-
-#ifdef illumos
- VN_HOLD(svp);
- error = dounmount(vn_mountedvfs(svp), fflags, cr);
- if (error) {
- VN_RELE(svp);
- return (error);
- }
-
- /*
- * We can't use VN_RELE(), as that will try to invoke
- * zfsctl_snapdir_inactive(), which would cause us to destroy
- * the sd_lock mutex held by our caller.
- */
- ASSERT(svp->v_count == 1);
- gfs_vop_reclaim(svp, cr, NULL);
-
- kmem_free(sep->se_name, strlen(sep->se_name) + 1);
- kmem_free(sep, sizeof (zfs_snapentry_t));
-
- return (0);
-#else
- vfs_ref(vn_mountedvfs(svp));
- return (dounmount(vn_mountedvfs(svp), fflags, curthread));
-#endif
-}
-
-#ifdef illumos
-static void
-zfsctl_rename_snap(zfsctl_snapdir_t *sdp, zfs_snapentry_t *sep, const char *nm)
-{
- avl_index_t where;
- vfs_t *vfsp;
- refstr_t *pathref;
- char newpath[MAXNAMELEN];
- char *tail;
-
- ASSERT(MUTEX_HELD(&sdp->sd_lock));
- ASSERT(sep != NULL);
-
- vfsp = vn_mountedvfs(sep->se_root);
- ASSERT(vfsp != NULL);
-
- vfs_lock_wait(vfsp);
-
- /*
- * Change the name in the AVL tree.
- */
- avl_remove(&sdp->sd_snaps, sep);
- kmem_free(sep->se_name, strlen(sep->se_name) + 1);
- sep->se_name = kmem_alloc(strlen(nm) + 1, KM_SLEEP);
- (void) strcpy(sep->se_name, nm);
- VERIFY(avl_find(&sdp->sd_snaps, sep, &where) == NULL);
- avl_insert(&sdp->sd_snaps, sep, where);
-
- /*
- * Change the current mountpoint info:
- * - update the tail of the mntpoint path
- * - update the tail of the resource path
- */
- pathref = vfs_getmntpoint(vfsp);
- (void) strncpy(newpath, refstr_value(pathref), sizeof (newpath));
- VERIFY((tail = strrchr(newpath, '/')) != NULL);
- *(tail+1) = '\0';
- ASSERT3U(strlen(newpath) + strlen(nm), <, sizeof (newpath));
- (void) strcat(newpath, nm);
- refstr_rele(pathref);
- vfs_setmntpoint(vfsp, newpath, 0);
-
- pathref = vfs_getresource(vfsp);
- (void) strncpy(newpath, refstr_value(pathref), sizeof (newpath));
- VERIFY((tail = strrchr(newpath, '@')) != NULL);
- *(tail+1) = '\0';
- ASSERT3U(strlen(newpath) + strlen(nm), <, sizeof (newpath));
- (void) strcat(newpath, nm);
- refstr_rele(pathref);
- vfs_setresource(vfsp, newpath, 0);
-
- vfs_unlock(vfsp);
-}
-#endif /* illumos */
-
-#ifdef illumos
-/*ARGSUSED*/
-static int
-zfsctl_snapdir_rename(vnode_t *sdvp, char *snm, vnode_t *tdvp, char *tnm,
- cred_t *cr, caller_context_t *ct, int flags)
-{
- zfsctl_snapdir_t *sdp = sdvp->v_data;
- zfs_snapentry_t search, *sep;
- zfsvfs_t *zfsvfs;
- avl_index_t where;
- char from[ZFS_MAX_DATASET_NAME_LEN], to[ZFS_MAX_DATASET_NAME_LEN];
- char real[ZFS_MAX_DATASET_NAME_LEN], fsname[ZFS_MAX_DATASET_NAME_LEN];
+ objset_t *os = ((zfsvfs_t *)((vp)->v_vfsp->vfs_data))->z_os;
int err;
- zfsvfs = sdvp->v_vfsp->vfs_data;
- ZFS_ENTER(zfsvfs);
-
- if ((flags & FIGNORECASE) || zfsvfs->z_case == ZFS_CASE_INSENSITIVE) {
- err = dmu_snapshot_realname(zfsvfs->z_os, snm, real,
- sizeof (real), NULL);
- if (err == 0) {
- snm = real;
- } else if (err != ENOTSUP) {
- ZFS_EXIT(zfsvfs);
- return (err);
- }
- }
-
- ZFS_EXIT(zfsvfs);
-
- dmu_objset_name(zfsvfs->z_os, fsname);
-
- err = zfsctl_snapshot_zname(sdvp, snm, sizeof (from), from);
- if (err == 0)
- err = zfsctl_snapshot_zname(tdvp, tnm, sizeof (to), to);
- if (err == 0)
- err = zfs_secpolicy_rename_perms(from, to, cr);
- if (err != 0)
- return (err);
-
- /*
- * Cannot move snapshots out of the snapdir.
- */
- if (sdvp != tdvp)
- return (SET_ERROR(EINVAL));
-
- if (strcmp(snm, tnm) == 0)
- return (0);
-
- mutex_enter(&sdp->sd_lock);
-
- search.se_name = (char *)snm;
- if ((sep = avl_find(&sdp->sd_snaps, &search, &where)) == NULL) {
- mutex_exit(&sdp->sd_lock);
- return (SET_ERROR(ENOENT));
- }
-
- err = dsl_dataset_rename_snapshot(fsname, snm, tnm, 0);
- if (err == 0)
- zfsctl_rename_snap(sdp, sep, tnm);
-
- mutex_exit(&sdp->sd_lock);
-
+ err = dsl_dataset_snap_lookup(dmu_objset_ds(os), name, id);
return (err);
}
-#endif /* illumos */
-
-#ifdef illumos
-/* ARGSUSED */
-static int
-zfsctl_snapdir_remove(vnode_t *dvp, char *name, vnode_t *cwd, cred_t *cr,
- caller_context_t *ct, int flags)
-{
- zfsctl_snapdir_t *sdp = dvp->v_data;
- zfs_snapentry_t *sep;
- zfs_snapentry_t search;
- zfsvfs_t *zfsvfs;
- char snapname[ZFS_MAX_DATASET_NAME_LEN];
- char real[ZFS_MAX_DATASET_NAME_LEN];
- int err;
-
- zfsvfs = dvp->v_vfsp->vfs_data;
- ZFS_ENTER(zfsvfs);
-
- if ((flags & FIGNORECASE) || zfsvfs->z_case == ZFS_CASE_INSENSITIVE) {
-
- err = dmu_snapshot_realname(zfsvfs->z_os, name, real,
- sizeof (real), NULL);
- if (err == 0) {
- name = real;
- } else if (err != ENOTSUP) {
- ZFS_EXIT(zfsvfs);
- return (err);
- }
- }
-
- ZFS_EXIT(zfsvfs);
-
- err = zfsctl_snapshot_zname(dvp, name, sizeof (snapname), snapname);
- if (err == 0)
- err = zfs_secpolicy_destroy_perms(snapname, cr);
- if (err != 0)
- return (err);
-
- mutex_enter(&sdp->sd_lock);
-
- search.se_name = name;
- sep = avl_find(&sdp->sd_snaps, &search, NULL);
- if (sep) {
- avl_remove(&sdp->sd_snaps, sep);
- err = zfsctl_unmount_snap(sep, MS_FORCE, cr);
- if (err != 0)
- avl_add(&sdp->sd_snaps, sep);
- else
- err = dsl_destroy_snapshot(snapname, B_FALSE);
- } else {
- err = SET_ERROR(ENOENT);
- }
-
- mutex_exit(&sdp->sd_lock);
-
- return (err);
-}
-#endif /* illumos */
/*
- * This creates a snapshot under '.zfs/snapshot'.
+ * Given a vnode get a root vnode of a filesystem mounted on top of
+ * the vnode, if any. The root vnode is referenced and locked.
+ * If no filesystem is mounted then the orinal vnode remains referenced
+ * and locked. If any error happens the orinal vnode is unlocked and
+ * released.
*/
-/* ARGSUSED */
static int
-zfsctl_snapdir_mkdir(vnode_t *dvp, char *dirname, vattr_t *vap, vnode_t **vpp,
- cred_t *cr, caller_context_t *cc, int flags, vsecattr_t *vsecp)
+zfsctl_mounted_here(vnode_t **vpp, int flags)
{
- zfsvfs_t *zfsvfs = dvp->v_vfsp->vfs_data;
- char name[ZFS_MAX_DATASET_NAME_LEN];
+ struct mount *mp;
int err;
- static enum symfollow follow = NO_FOLLOW;
- static enum uio_seg seg = UIO_SYSSPACE;
-
- if (zfs_component_namecheck(dirname, NULL, NULL) != 0)
- return (SET_ERROR(EILSEQ));
- dmu_objset_name(zfsvfs->z_os, name);
+ ASSERT_VOP_LOCKED(*vpp, __func__);
+ ASSERT3S((*vpp)->v_type, ==, VDIR);
- *vpp = NULL;
-
- err = zfs_secpolicy_snapshot_perms(name, cr);
- if (err != 0)
+ if ((mp = (*vpp)->v_mountedhere) != NULL) {
+ err = vfs_busy(mp, 0);
+ KASSERT(err == 0, ("vfs_busy(mp, 0) failed with %d", err));
+ KASSERT(vrefcnt(*vpp) > 1, ("unreferenced mountpoint"));
+ vput(*vpp);
+ err = VFS_ROOT(mp, flags, vpp);
+ vfs_unbusy(mp);
return (err);
-
- if (err == 0) {
- err = dmu_objset_snapshot_one(name, dirname);
- if (err != 0)
- return (err);
- err = lookupnameat(dirname, seg, follow, NULL, vpp, dvp);
}
-
- return (err);
+ return (EJUSTRETURN);
}
-static int
-zfsctl_freebsd_snapdir_mkdir(ap)
- struct vop_mkdir_args /* {
- struct vnode *a_dvp;
- struct vnode **a_vpp;
- struct componentname *a_cnp;
- struct vattr *a_vap;
- } */ *ap;
+typedef struct {
+ const char *snap_name;
+ uint64_t snap_id;
+} snapshot_setup_arg_t;
+
+static void
+zfsctl_snapshot_vnode_setup(vnode_t *vp, void *arg)
{
+ snapshot_setup_arg_t *ssa = arg;
+ sfs_node_t *node;
+
+ ASSERT_VOP_ELOCKED(vp, __func__);
- ASSERT(ap->a_cnp->cn_flags & SAVENAME);
+ node = sfs_alloc_node(sizeof(sfs_node_t),
+ ssa->snap_name, ZFSCTL_INO_SNAPDIR, ssa->snap_id);
+ zfsctl_common_vnode_setup(vp, node);
- return (zfsctl_snapdir_mkdir(ap->a_dvp, ap->a_cnp->cn_nameptr, NULL,
- ap->a_vpp, ap->a_cnp->cn_cred, NULL, 0, NULL));
+ /* We have to support recursive locking. */
+ VN_LOCK_AREC(vp);
}
/*
@@ -970,150 +810,100 @@
vnode_t *dvp = ap->a_dvp;
vnode_t **vpp = ap->a_vpp;
struct componentname *cnp = ap->a_cnp;
- char nm[NAME_MAX + 1];
- zfsctl_snapdir_t *sdp = dvp->v_data;
- objset_t *snap;
- char snapname[ZFS_MAX_DATASET_NAME_LEN];
- char real[ZFS_MAX_DATASET_NAME_LEN];
+ char name[NAME_MAX + 1];
+ char fullname[ZFS_MAX_DATASET_NAME_LEN];
char *mountpoint;
- zfs_snapentry_t *sep, search;
size_t mountpoint_len;
- avl_index_t where;
zfsvfs_t *zfsvfs = dvp->v_vfsp->vfs_data;
+ uint64_t snap_id;
+ int nameiop = cnp->cn_nameiop;
+ int lkflags = cnp->cn_lkflags;
+ int flags = cnp->cn_flags;
int err;
- int ltype, flags = 0;
-
- /*
- * No extended attributes allowed under .zfs
- */
- if (flags & LOOKUP_XATTR)
- return (SET_ERROR(EINVAL));
- ASSERT(ap->a_cnp->cn_namelen < sizeof(nm));
- strlcpy(nm, ap->a_cnp->cn_nameptr, ap->a_cnp->cn_namelen + 1);
ASSERT(dvp->v_type == VDIR);
- *vpp = NULL;
+ if ((flags & ISLASTCN) != 0 && nameiop != LOOKUP)
+ return (SET_ERROR(ENOTSUP));
- /*
- * If we get a recursive call, that means we got called
- * from the domount() code while it was trying to look up the
- * spec (which looks like a local path for zfs). We need to
- * add some flag to domount() to tell it not to do this lookup.
- */
- if (MUTEX_HELD(&sdp->sd_lock))
+ if (cnp->cn_namelen == 1 && *cnp->cn_nameptr == '.') {
+ err = zfsctl_relock_dot(dvp, lkflags & LK_TYPE_MASK);
+ if (err == 0)
+ *vpp = dvp;
+ return (err);
+ }
+ if (flags & ISDOTDOT) {
+ err = vn_vget_ino_gen(dvp, zfsctl_root_vnode, NULL, lkflags,
+ vpp);
+ return (err);
+ }
+
+ if (cnp->cn_namelen >= sizeof(name))
+ return (SET_ERROR(ENAMETOOLONG));
+
+ strlcpy(name, ap->a_cnp->cn_nameptr, ap->a_cnp->cn_namelen + 1);
+ err = zfsctl_snapshot_lookup(dvp, name, &snap_id);
+ if (err != 0)
return (SET_ERROR(ENOENT));
- ZFS_ENTER(zfsvfs);
- if (gfs_lookup_dot(vpp, dvp, zfsvfs->z_ctldir, nm) == 0) {
- if (nm[0] == '.' && nm[1] == '.' && nm[2] =='\0') {
- VOP_UNLOCK(dvp, 0);
- VERIFY0(vn_lock(*vpp, LK_EXCLUSIVE));
- VERIFY0(vn_lock(dvp, LK_EXCLUSIVE));
- }
- ZFS_EXIT(zfsvfs);
- return (0);
- }
+ for (;;) {
+ snapshot_setup_arg_t ssa;
+
+ ssa.snap_name = name;
+ ssa.snap_id = snap_id;
+ err = sfs_vgetx(dvp->v_mount, LK_SHARED, ZFSCTL_INO_SNAPDIR,
+ snap_id, "zfs", &zfsctl_ops_snapshot,
+ zfsctl_snapshot_vnode_setup, &ssa, vpp);
+ if (err != 0)
+ return (err);
- if (flags & FIGNORECASE) {
- boolean_t conflict = B_FALSE;
+ /* Check if a new vnode has just been created. */
+ if (VOP_ISLOCKED(*vpp) == LK_EXCLUSIVE)
+ break;
- err = dmu_snapshot_realname(zfsvfs->z_os, nm, real,
- sizeof (real), &conflict);
- if (err == 0) {
- strlcpy(nm, real, sizeof(nm));
- } else if (err != ENOTSUP) {
- ZFS_EXIT(zfsvfs);
+ /*
+ * The vnode must be referenced at least by this thread and
+ * the mounted snapshot or the thread doing the mounting.
+ * There can be more references from concurrent lookups.
+ */
+ KASSERT(vrefcnt(*vpp) > 1, ("found unreferenced mountpoint"));
+
+ /*
+ * Check if a snapshot is already mounted on top of the vnode.
+ */
+ err = zfsctl_mounted_here(vpp, lkflags);
+ if (err != EJUSTRETURN)
return (err);
- }
-#if 0
- if (realpnp)
- (void) strlcpy(realpnp->pn_buf, nm,
- realpnp->pn_bufsize);
- if (conflict && direntflags)
- *direntflags = ED_CASE_CONFLICT;
-#endif
- }
-relookup:
- mutex_enter(&sdp->sd_lock);
- search.se_name = (char *)nm;
- if ((sep = avl_find(&sdp->sd_snaps, &search, &where)) != NULL) {
- *vpp = sep->se_root;
- VN_HOLD(*vpp);
- err = traverse(vpp, LK_EXCLUSIVE | LK_RETRY);
- if (err != 0) {
- *vpp = NULL;
- } else if (*vpp == sep->se_root) {
- /*
- * The snapshot was unmounted behind our backs,
- * try to remount it.
- */
- VERIFY(zfsctl_snapshot_zname(dvp, nm, MAXNAMELEN, snapname) == 0);
- goto domount;
- }
- mutex_exit(&sdp->sd_lock);
- ZFS_EXIT(zfsvfs);
- return (err);
- }
+#ifdef INVARIANTS
+ /*
+ * If the vnode not covered yet, then the mount operation
+ * must be in progress.
+ */
+ VI_LOCK(*vpp);
+ KASSERT(((*vpp)->v_iflag & VI_MOUNT) != 0,
+ ("snapshot vnode not covered"));
+ VI_UNLOCK(*vpp);
+#endif
+ vput(*vpp);
- /*
- * The requested snapshot is not currently mounted, look it up.
- */
- err = zfsctl_snapshot_zname(dvp, nm, sizeof (snapname), snapname);
- if (err != 0) {
- mutex_exit(&sdp->sd_lock);
- ZFS_EXIT(zfsvfs);
/*
- * handle "ls *" or "?" in a graceful manner,
- * forcing EILSEQ to ENOENT.
- * Since shell ultimately passes "*" or "?" as name to lookup
+ * In this situation we can loop on uncontested locks and starve
+ * the thread doing the lengthy, non-trivial mount operation.
*/
- return (err == EILSEQ ? ENOENT : err);
- }
- if (dmu_objset_hold(snapname, FTAG, &snap) != 0) {
- mutex_exit(&sdp->sd_lock);
-#ifdef illumos
- ZFS_EXIT(zfsvfs);
- return (SET_ERROR(ENOENT));
-#else /* !illumos */
- /* Translate errors and add SAVENAME when needed. */
- if ((cnp->cn_flags & ISLASTCN) && cnp->cn_nameiop == CREATE) {
- err = EJUSTRETURN;
- cnp->cn_flags |= SAVENAME;
- } else {
- err = SET_ERROR(ENOENT);
- }
- ZFS_EXIT(zfsvfs);
- return (err);
-#endif /* illumos */
+ kern_yield(PRI_USER);
}
- sep = kmem_alloc(sizeof (zfs_snapentry_t), KM_SLEEP);
- sep->se_name = kmem_alloc(strlen(nm) + 1, KM_SLEEP);
- (void) strcpy(sep->se_name, nm);
- *vpp = sep->se_root = zfsctl_snapshot_mknode(dvp, dmu_objset_id(snap));
- avl_insert(&sdp->sd_snaps, sep, where);
+ VERIFY0(zfsctl_snapshot_zname(dvp, name, sizeof(fullname), fullname));
- dmu_objset_rele(snap, FTAG);
-domount:
mountpoint_len = strlen(dvp->v_vfsp->mnt_stat.f_mntonname) +
- strlen("/" ZFS_CTLDIR_NAME "/snapshot/") + strlen(nm) + 1;
+ strlen("/" ZFS_CTLDIR_NAME "/snapshot/") + strlen(name) + 1;
mountpoint = kmem_alloc(mountpoint_len, KM_SLEEP);
(void) snprintf(mountpoint, mountpoint_len,
"%s/" ZFS_CTLDIR_NAME "/snapshot/%s",
- dvp->v_vfsp->mnt_stat.f_mntonname, nm);
- mutex_exit(&sdp->sd_lock);
+ dvp->v_vfsp->mnt_stat.f_mntonname, name);
- /*
- * The vnode may get reclaimed between dropping sd_lock and
- * getting the vnode lock.
- * */
- err = vn_lock(*vpp, LK_EXCLUSIVE);
- if (err == ENOENT)
- goto relookup;
- VERIFY0(err);
- err = mount_snapshot(curthread, vpp, "zfs", mountpoint, snapname, 0);
+ err = mount_snapshot(curthread, vpp, "zfs", mountpoint, fullname, 0);
kmem_free(mountpoint, mountpoint_len);
if (err == 0) {
/*
@@ -1125,231 +915,79 @@
*/
ASSERT(VTOZ(*vpp)->z_zfsvfs != zfsvfs);
VTOZ(*vpp)->z_zfsvfs->z_parent = zfsvfs;
- (*vpp)->v_flag &= ~VROOT;
- }
- ZFS_EXIT(zfsvfs);
-#ifdef illumos
- /*
- * If we had an error, drop our hold on the vnode and
- * zfsctl_snapshot_inactive() will clean up.
- */
- if (err != 0) {
- VN_RELE(*vpp);
- *vpp = NULL;
+ /* Clear the root flag (set via VFS_ROOT) as well. */
+ (*vpp)->v_vflag &= ~VV_ROOT;
}
-#else
+
if (err != 0)
*vpp = NULL;
-#endif
return (err);
}
-/* ARGSUSED */
-int
-zfsctl_shares_lookup(ap)
- struct vop_lookup_args /* {
- struct vnode *a_dvp;
- struct vnode **a_vpp;
- struct componentname *a_cnp;
- } */ *ap;
-{
- vnode_t *dvp = ap->a_dvp;
- vnode_t **vpp = ap->a_vpp;
- struct componentname *cnp = ap->a_cnp;
- zfsvfs_t *zfsvfs = dvp->v_vfsp->vfs_data;
- char nm[NAME_MAX + 1];
- znode_t *dzp;
- int error;
-
- ZFS_ENTER(zfsvfs);
-
- ASSERT(cnp->cn_namelen < sizeof(nm));
- strlcpy(nm, cnp->cn_nameptr, cnp->cn_namelen + 1);
-
- if (gfs_lookup_dot(vpp, dvp, zfsvfs->z_ctldir, nm) == 0) {
- if (nm[0] == '.' && nm[1] == '.' && nm[2] =='\0') {
- VOP_UNLOCK(dvp, 0);
- VERIFY0(vn_lock(*vpp, LK_EXCLUSIVE));
- VERIFY0(vn_lock(dvp, LK_EXCLUSIVE));
- }
- ZFS_EXIT(zfsvfs);
- return (0);
- }
-
- if (zfsvfs->z_shares_dir == 0) {
- ZFS_EXIT(zfsvfs);
- return (SET_ERROR(ENOTSUP));
- }
- if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &dzp)) == 0) {
- error = VOP_LOOKUP(ZTOV(dzp), vpp, cnp);
- VN_RELE(ZTOV(dzp));
- }
-
- ZFS_EXIT(zfsvfs);
-
- return (error);
-}
-
-/* ARGSUSED */
static int
-zfsctl_snapdir_readdir_cb(vnode_t *vp, void *dp, int *eofp,
- offset_t *offp, offset_t *nextp, void *data, int flags)
-{
- zfsvfs_t *zfsvfs = vp->v_vfsp->vfs_data;
- char snapname[ZFS_MAX_DATASET_NAME_LEN];
- uint64_t id, cookie;
- boolean_t case_conflict;
- int error;
-
- ZFS_ENTER(zfsvfs);
-
- cookie = *offp;
- dsl_pool_config_enter(dmu_objset_pool(zfsvfs->z_os), FTAG);
- error = dmu_snapshot_list_next(zfsvfs->z_os,
- sizeof (snapname), snapname, &id, &cookie, &case_conflict);
- dsl_pool_config_exit(dmu_objset_pool(zfsvfs->z_os), FTAG);
- if (error) {
- ZFS_EXIT(zfsvfs);
- if (error == ENOENT) {
- *eofp = 1;
- return (0);
- }
- return (error);
- }
-
- if (flags & V_RDDIR_ENTFLAGS) {
- edirent_t *eodp = dp;
-
- (void) strcpy(eodp->ed_name, snapname);
- eodp->ed_ino = ZFSCTL_INO_SNAP(id);
- eodp->ed_eflags = case_conflict ? ED_CASE_CONFLICT : 0;
- } else {
- struct dirent64 *odp = dp;
-
- (void) strcpy(odp->d_name, snapname);
- odp->d_ino = ZFSCTL_INO_SNAP(id);
- }
- *nextp = cookie;
-
- ZFS_EXIT(zfsvfs);
-
- return (0);
-}
-
-/* ARGSUSED */
-static int
-zfsctl_shares_readdir(ap)
+zfsctl_snapdir_readdir(ap)
struct vop_readdir_args /* {
struct vnode *a_vp;
struct uio *a_uio;
struct ucred *a_cred;
int *a_eofflag;
- int *a_ncookies;
+ int *ncookies;
u_long **a_cookies;
} */ *ap;
{
+ char snapname[ZFS_MAX_DATASET_NAME_LEN];
+ struct dirent entry;
vnode_t *vp = ap->a_vp;
- uio_t *uiop = ap->a_uio;
- cred_t *cr = ap->a_cred;
- int *eofp = ap->a_eofflag;
zfsvfs_t *zfsvfs = vp->v_vfsp->vfs_data;
- znode_t *dzp;
+ uio_t *uio = ap->a_uio;
+ int *eofp = ap->a_eofflag;
+ off_t dots_offset;
int error;
- ZFS_ENTER(zfsvfs);
+ ASSERT(vp->v_type == VDIR);
- if (zfsvfs->z_shares_dir == 0) {
- ZFS_EXIT(zfsvfs);
- return (SET_ERROR(ENOTSUP));
- }
- if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &dzp)) == 0) {
- vn_lock(ZTOV(dzp), LK_SHARED | LK_RETRY);
- error = VOP_READDIR(ZTOV(dzp), uiop, cr, eofp, ap->a_ncookies, ap->a_cookies);
- VN_URELE(ZTOV(dzp));
- } else {
- *eofp = 1;
- error = SET_ERROR(ENOENT);
+ error = sfs_readdir_common(ZFSCTL_INO_ROOT, ZFSCTL_INO_SNAPDIR, ap, uio,
+ &dots_offset);
+ if (error != 0) {
+ if (error == ENAMETOOLONG) /* ran out of destination space */
+ error = 0;
+ return (error);
}
- ZFS_EXIT(zfsvfs);
- return (error);
-}
-
-/*
- * pvp is the '.zfs' directory (zfsctl_node_t).
- *
- * Creates vp, which is '.zfs/snapshot' (zfsctl_snapdir_t).
- *
- * This function is the callback to create a GFS vnode for '.zfs/snapshot'
- * when a lookup is performed on .zfs for "snapshot".
- */
-vnode_t *
-zfsctl_mknode_snapdir(vnode_t *pvp)
-{
- vnode_t *vp;
- zfsctl_snapdir_t *sdp;
-
- vp = gfs_dir_create(sizeof (zfsctl_snapdir_t), pvp, pvp->v_vfsp,
- &zfsctl_ops_snapdir, NULL, NULL, MAXNAMELEN,
- zfsctl_snapdir_readdir_cb, NULL);
- sdp = vp->v_data;
- sdp->sd_node.zc_id = ZFSCTL_INO_SNAPDIR;
- sdp->sd_node.zc_cmtime = ((zfsctl_node_t *)pvp->v_data)->zc_cmtime;
- mutex_init(&sdp->sd_lock, NULL, MUTEX_DEFAULT, NULL);
- avl_create(&sdp->sd_snaps, snapentry_compare,
- sizeof (zfs_snapentry_t), offsetof(zfs_snapentry_t, se_node));
- VOP_UNLOCK(vp, 0);
- return (vp);
-}
-
-vnode_t *
-zfsctl_mknode_shares(vnode_t *pvp)
-{
- vnode_t *vp;
- zfsctl_node_t *sdp;
-
- vp = gfs_dir_create(sizeof (zfsctl_node_t), pvp, pvp->v_vfsp,
- &zfsctl_ops_shares, NULL, NULL, MAXNAMELEN,
- NULL, NULL);
- sdp = vp->v_data;
- sdp->zc_cmtime = ((zfsctl_node_t *)pvp->v_data)->zc_cmtime;
- VOP_UNLOCK(vp, 0);
- return (vp);
-
-}
-
-/* ARGSUSED */
-static int
-zfsctl_shares_getattr(ap)
- struct vop_getattr_args /* {
- struct vnode *a_vp;
- struct vattr *a_vap;
- struct ucred *a_cred;
- struct thread *a_td;
- } */ *ap;
-{
- vnode_t *vp = ap->a_vp;
- vattr_t *vap = ap->a_vap;
- cred_t *cr = ap->a_cred;
- zfsvfs_t *zfsvfs = vp->v_vfsp->vfs_data;
- znode_t *dzp;
- int error;
+ for (;;) {
+ uint64_t cookie;
+ uint64_t id;
+
+ cookie = uio->uio_offset - dots_offset;
+
+ dsl_pool_config_enter(dmu_objset_pool(zfsvfs->z_os), FTAG);
+ error = dmu_snapshot_list_next(zfsvfs->z_os, sizeof(snapname),
+ snapname, &id, &cookie, NULL);
+ dsl_pool_config_exit(dmu_objset_pool(zfsvfs->z_os), FTAG);
+ if (error != 0) {
+ if (error == ENOENT) {
+ if (eofp != NULL)
+ *eofp = 1;
+ error = 0;
+ }
+ return (error);
+ }
- ZFS_ENTER(zfsvfs);
- if (zfsvfs->z_shares_dir == 0) {
- ZFS_EXIT(zfsvfs);
- return (SET_ERROR(ENOTSUP));
- }
- if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &dzp)) == 0) {
- vn_lock(ZTOV(dzp), LK_SHARED | LK_RETRY);
- error = VOP_GETATTR(ZTOV(dzp), vap, cr);
- VN_URELE(ZTOV(dzp));
+ entry.d_fileno = id;
+ entry.d_type = DT_DIR;
+ strcpy(entry.d_name, snapname);
+ entry.d_namlen = strlen(entry.d_name);
+ entry.d_reclen = sizeof(entry);
+ error = vfs_read_dirent(ap, &entry, uio->uio_offset);
+ if (error != 0) {
+ if (error == ENAMETOOLONG)
+ error = 0;
+ return (SET_ERROR(error));
+ }
+ uio->uio_offset = cookie + dots_offset;
}
- ZFS_EXIT(zfsvfs);
- return (error);
-
-
+ /* NOTREACHED */
}
/* ARGSUSED */
@@ -1364,151 +1002,40 @@
vnode_t *vp = ap->a_vp;
vattr_t *vap = ap->a_vap;
zfsvfs_t *zfsvfs = vp->v_vfsp->vfs_data;
- zfsctl_snapdir_t *sdp = vp->v_data;
+ dsl_dataset_t *ds = dmu_objset_ds(zfsvfs->z_os);
+ sfs_node_t *node = vp->v_data;
+ uint64_t snap_count;
+ int err;
- 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;
- vap->va_ctime = vap->va_mtime = dmu_objset_snap_cmtime(zfsvfs->z_os);
+ vap->va_ctime = dmu_objset_snap_cmtime(zfsvfs->z_os);
+ vap->va_mtime = vap->va_ctime;
vap->va_birthtime = vap->va_ctime;
- ZFS_EXIT(zfsvfs);
-
- return (0);
-}
-
-/* ARGSUSED */
-static int
-zfsctl_snapdir_reclaim(ap)
- struct vop_reclaim_args /* {
- struct vnode *a_vp;
- struct thread *a_td;
- } */ *ap;
-{
- vnode_t *vp = ap->a_vp;
- zfsctl_snapdir_t *sdp = vp->v_data;
- zfs_snapentry_t *sep;
-
- ASSERT(avl_numnodes(&sdp->sd_snaps) == 0);
- mutex_destroy(&sdp->sd_lock);
- avl_destroy(&sdp->sd_snaps);
- gfs_vop_reclaim(ap);
-
- return (0);
-}
-
-static int
-zfsctl_shares_print(ap)
- struct vop_print_args /* {
- struct vnode *a_vp;
- } */ *ap;
-{
- printf(" .zfs/shares node\n");
- zfsctl_common_print(ap);
- return (0);
-}
-
-static int
-zfsctl_snapdir_print(ap)
- struct vop_print_args /* {
- struct vnode *a_vp;
- } */ *ap;
-{
- vnode_t *vp = ap->a_vp;
- zfsctl_snapdir_t *sdp = vp->v_data;
+ if (dsl_dataset_phys(ds)->ds_snapnames_zapobj != 0) {
+ err = zap_count(dmu_objset_pool(ds->ds_objset)->dp_meta_objset,
+ dsl_dataset_phys(ds)->ds_snapnames_zapobj, &snap_count);
+ if (err != 0)
+ return (err);
+ vap->va_nlink += snap_count;
+ }
+ vap->va_size = vap->va_nlink;
- printf(" .zfs/snapshot node\n");
- printf(" number of children = %lu\n", avl_numnodes(&sdp->sd_snaps));
- zfsctl_common_print(ap);
return (0);
}
-#ifdef illumos
-static const fs_operation_def_t zfsctl_tops_snapdir[] = {
- { VOPNAME_OPEN, { .vop_open = zfsctl_common_open } },
- { VOPNAME_CLOSE, { .vop_close = zfsctl_common_close } },
- { VOPNAME_IOCTL, { .error = fs_inval } },
- { VOPNAME_GETATTR, { .vop_getattr = zfsctl_snapdir_getattr } },
- { VOPNAME_ACCESS, { .vop_access = zfsctl_common_access } },
- { VOPNAME_RENAME, { .vop_rename = zfsctl_snapdir_rename } },
- { VOPNAME_RMDIR, { .vop_rmdir = zfsctl_snapdir_remove } },
- { VOPNAME_MKDIR, { .vop_mkdir = zfsctl_snapdir_mkdir } },
- { VOPNAME_READDIR, { .vop_readdir = gfs_vop_readdir } },
- { VOPNAME_LOOKUP, { .vop_lookup = zfsctl_snapdir_lookup } },
- { VOPNAME_SEEK, { .vop_seek = fs_seek } },
- { VOPNAME_INACTIVE, { .vop_inactive = zfsctl_snapdir_inactive } },
- { VOPNAME_FID, { .vop_fid = zfsctl_common_fid } },
- { NULL }
-};
-
-static const fs_operation_def_t zfsctl_tops_shares[] = {
- { VOPNAME_OPEN, { .vop_open = zfsctl_common_open } },
- { VOPNAME_CLOSE, { .vop_close = zfsctl_common_close } },
- { VOPNAME_IOCTL, { .error = fs_inval } },
- { VOPNAME_GETATTR, { .vop_getattr = zfsctl_shares_getattr } },
- { VOPNAME_ACCESS, { .vop_access = zfsctl_common_access } },
- { VOPNAME_READDIR, { .vop_readdir = zfsctl_shares_readdir } },
- { VOPNAME_LOOKUP, { .vop_lookup = zfsctl_shares_lookup } },
- { VOPNAME_SEEK, { .vop_seek = fs_seek } },
- { VOPNAME_INACTIVE, { .vop_inactive = gfs_vop_inactive } },
- { VOPNAME_FID, { .vop_fid = zfsctl_shares_fid } },
- { NULL }
-};
-#else /* !illumos */
static struct vop_vector zfsctl_ops_snapdir = {
.vop_default = &default_vnodeops,
.vop_open = zfsctl_common_open,
.vop_close = zfsctl_common_close,
- .vop_ioctl = VOP_EINVAL,
.vop_getattr = zfsctl_snapdir_getattr,
.vop_access = zfsctl_common_access,
- .vop_mkdir = zfsctl_freebsd_snapdir_mkdir,
- .vop_readdir = gfs_vop_readdir,
+ .vop_readdir = zfsctl_snapdir_readdir,
.vop_lookup = zfsctl_snapdir_lookup,
- .vop_inactive = VOP_NULL,
- .vop_reclaim = zfsctl_snapdir_reclaim,
+ .vop_reclaim = zfsctl_common_reclaim,
.vop_fid = zfsctl_common_fid,
- .vop_print = zfsctl_snapdir_print,
+ .vop_print = zfsctl_common_print,
};
-static struct vop_vector zfsctl_ops_shares = {
- .vop_default = &default_vnodeops,
- .vop_open = zfsctl_common_open,
- .vop_close = zfsctl_common_close,
- .vop_ioctl = VOP_EINVAL,
- .vop_getattr = zfsctl_shares_getattr,
- .vop_access = zfsctl_common_access,
- .vop_readdir = zfsctl_shares_readdir,
- .vop_lookup = zfsctl_shares_lookup,
- .vop_inactive = VOP_NULL,
- .vop_reclaim = gfs_vop_reclaim,
- .vop_fid = zfsctl_shares_fid,
- .vop_print = zfsctl_shares_print,
-};
-#endif /* illumos */
-
-/*
- * pvp is the GFS vnode '.zfs/snapshot'.
- *
- * This creates a GFS node under '.zfs/snapshot' representing each
- * snapshot. This newly created GFS node is what we mount snapshot
- * vfs_t's ontop of.
- */
-static vnode_t *
-zfsctl_snapshot_mknode(vnode_t *pvp, uint64_t objset)
-{
- vnode_t *vp;
- zfsctl_node_t *zcp;
-
- vp = gfs_dir_create(sizeof (zfsctl_node_t), pvp, pvp->v_vfsp,
- &zfsctl_ops_snapshot, NULL, NULL, MAXNAMELEN, NULL, NULL);
- zcp = vp->v_data;
- zcp->zc_id = objset;
- VOP_UNLOCK(vp, 0);
-
- return (vp);
-}
-
static int
zfsctl_snapshot_inactive(ap)
struct vop_inactive_args /* {
@@ -1518,7 +1045,7 @@
{
vnode_t *vp = ap->a_vp;
- vrecycle(vp);
+ VERIFY(vrecycle(vp) == 1);
return (0);
}
@@ -1530,183 +1057,104 @@
} */ *ap;
{
vnode_t *vp = ap->a_vp;
- cred_t *cr = ap->a_td->td_ucred;
- zfsctl_snapdir_t *sdp;
- zfs_snapentry_t *sep, *next;
- int locked;
- vnode_t *dvp;
+ void *data = vp->v_data;
- VERIFY(gfs_dir_lookup(vp, "..", &dvp, cr, 0, NULL, NULL) == 0);
- sdp = dvp->v_data;
- /* this may already have been unmounted */
- if (sdp == NULL) {
- VN_RELE(dvp);
- return (0);
- }
- if (!(locked = MUTEX_HELD(&sdp->sd_lock)))
- mutex_enter(&sdp->sd_lock);
-
- ASSERT(!vn_ismntpt(vp));
-
- sep = avl_first(&sdp->sd_snaps);
- while (sep != NULL) {
- next = AVL_NEXT(&sdp->sd_snaps, sep);
-
- if (sep->se_root == vp) {
- avl_remove(&sdp->sd_snaps, sep);
- kmem_free(sep->se_name, strlen(sep->se_name) + 1);
- kmem_free(sep, sizeof (zfs_snapentry_t));
- break;
- }
- sep = next;
- }
- ASSERT(sep != NULL);
-
- if (!locked)
- mutex_exit(&sdp->sd_lock);
- VN_RELE(dvp);
-
- /*
- * Dispose of the vnode for the snapshot mount point.
- * This is safe to do because once this entry has been removed
- * from the AVL tree, it can't be found again, so cannot become
- * "active". If we lookup the same name again we will end up
- * creating a new vnode.
- */
- gfs_vop_reclaim(ap);
+ sfs_reclaim_vnode(vp);
+ sfs_destroy_node(data);
return (0);
-
}
static int
zfsctl_snapshot_vptocnp(struct vop_vptocnp_args *ap)
{
- zfsvfs_t *zfsvfs = ap->a_vp->v_vfsp->vfs_data;
- vnode_t *dvp, *vp;
- zfsctl_snapdir_t *sdp;
- zfs_snapentry_t *sep;
+ struct mount *mp;
+ vnode_t *dvp;
+ vnode_t *vp;
+ sfs_node_t *node;
+ size_t len;
+ int locked;
int error;
- ASSERT(zfsvfs->z_ctldir != NULL);
- error = zfsctl_root_lookup(zfsvfs->z_ctldir, "snapshot", &dvp,
- NULL, 0, NULL, kcred, NULL, NULL, NULL);
- if (error != 0)
- return (error);
- sdp = dvp->v_data;
+ vp = ap->a_vp;
+ node = vp->v_data;
+ len = strlen(node->sn_name);
+ if (*ap->a_buflen < len)
+ return (SET_ERROR(ENOMEM));
- mutex_enter(&sdp->sd_lock);
- sep = avl_first(&sdp->sd_snaps);
- while (sep != NULL) {
- vp = sep->se_root;
- if (vp == ap->a_vp)
- break;
- sep = AVL_NEXT(&sdp->sd_snaps, sep);
- }
- if (sep == NULL) {
- mutex_exit(&sdp->sd_lock);
- error = SET_ERROR(ENOENT);
- } else {
- size_t len;
+ /*
+ * Prevent unmounting of the snapshot while the vnode lock
+ * is not held. That is not strictly required, but allows
+ * us to assert that an uncovered snapshot vnode is never
+ * "leaked".
+ */
+ mp = vp->v_mountedhere;
+ if (mp == NULL)
+ return (SET_ERROR(ENOENT));
+ error = vfs_busy(mp, 0);
+ KASSERT(error == 0, ("vfs_busy(mp, 0) failed with %d", error));
- len = strlen(sep->se_name);
- if (*ap->a_buflen < len) {
- mutex_exit(&sdp->sd_lock);
- error = SET_ERROR(ENOMEM);
- } else {
- *ap->a_buflen -= len;
- bcopy(sep->se_name, ap->a_buf + *ap->a_buflen, len);
- mutex_exit(&sdp->sd_lock);
- vref(dvp);
- *ap->a_vpp = dvp;
- }
+ /*
+ * We can vput the vnode as we can now depend on the reference owned
+ * by the busied mp. But we also need to hold the vnode, because
+ * the reference may go after vfs_unbusy() which has to be called
+ * before we can lock the vnode again.
+ */
+ locked = VOP_ISLOCKED(vp);
+ vhold(vp);
+ vput(vp);
+
+ /* Look up .zfs/snapshot, our parent. */
+ error = zfsctl_snapdir_vnode(vp->v_mount, NULL, LK_SHARED, &dvp);
+ if (error == 0) {
+ VOP_UNLOCK(dvp, 0);
+ *ap->a_vpp = dvp;
+ *ap->a_buflen -= len;
+ bcopy(node->sn_name, ap->a_buf + *ap->a_buflen, len);
}
- VN_RELE(dvp);
-
+ vfs_unbusy(mp);
+ vget(vp, locked | LK_VNHELD | LK_RETRY, curthread);
return (error);
}
-static int
-zfsctl_snaphot_print(ap)
- struct vop_print_args /* {
- struct vnode *a_vp;
- } */ *ap;
-{
- vnode_t *vp = ap->a_vp;
- zfsctl_node_t *zcp = vp->v_data;
-
- printf(" .zfs/snapshot/<snap> node\n");
- printf(" id = %ju\n", (uintmax_t)zcp->zc_id);
- zfsctl_common_print(ap);
- return (0);
-}
-
/*
* These VP's should never see the light of day. They should always
* be covered.
*/
static struct vop_vector zfsctl_ops_snapshot = {
- .vop_default = &default_vnodeops,
- .vop_inactive = zfsctl_snapshot_inactive,
- .vop_reclaim = zfsctl_snapshot_reclaim,
- .vop_vptocnp = zfsctl_snapshot_vptocnp,
- .vop_print = zfsctl_snaphot_print,
+ .vop_default = NULL, /* ensure very restricted access */
+ .vop_inactive = zfsctl_snapshot_inactive,
+ .vop_reclaim = zfsctl_snapshot_reclaim,
+ .vop_vptocnp = zfsctl_snapshot_vptocnp,
+ .vop_lock1 = vop_stdlock,
+ .vop_unlock = vop_stdunlock,
+ .vop_islocked = vop_stdislocked,
+ .vop_advlockpurge = vop_stdadvlockpurge, /* called by vgone */
+ .vop_print = zfsctl_common_print,
};
int
zfsctl_lookup_objset(vfs_t *vfsp, uint64_t objsetid, zfsvfs_t **zfsvfsp)
{
+ struct mount *mp;
zfsvfs_t *zfsvfs = vfsp->vfs_data;
- vnode_t *dvp, *vp;
- zfsctl_snapdir_t *sdp;
- zfsctl_node_t *zcp;
- zfs_snapentry_t *sep;
+ vnode_t *vp;
int error;
ASSERT(zfsvfs->z_ctldir != NULL);
- error = zfsctl_root_lookup(zfsvfs->z_ctldir, "snapshot", &dvp,
- NULL, 0, NULL, kcred, NULL, NULL, NULL);
- if (error != 0)
- return (error);
- sdp = dvp->v_data;
-
- mutex_enter(&sdp->sd_lock);
- sep = avl_first(&sdp->sd_snaps);
- while (sep != NULL) {
- vp = sep->se_root;
- zcp = vp->v_data;
- if (zcp->zc_id == objsetid)
- break;
-
- sep = AVL_NEXT(&sdp->sd_snaps, sep);
- }
-
- if (sep != NULL) {
- VN_HOLD(vp);
+ *zfsvfsp = NULL;
+ error = sfs_vnode_get(vfsp, LK_EXCLUSIVE,
+ ZFSCTL_INO_SNAPDIR, objsetid, &vp);
+ if (error == 0 && vp != NULL) {
/*
- * Return the mounted root rather than the covered mount point.
- * Takes the GFS vnode at .zfs/snapshot/<snapshot objsetid>
- * and returns the ZFS vnode mounted on top of the GFS node.
- * This ZFS vnode is the root of the vfs for objset 'objsetid'.
+ * XXX Probably need to at least reference, if not busy, the mp.
*/
- error = traverse(&vp, LK_SHARED | LK_RETRY);
- if (error == 0) {
- if (vp == sep->se_root) {
- VN_RELE(vp); /* release covered vp */
- error = SET_ERROR(EINVAL);
- } else {
- *zfsvfsp = VTOZ(vp)->z_zfsvfs;
- VN_URELE(vp); /* put snapshot's root vp */
- }
- }
- mutex_exit(&sdp->sd_lock);
- } else {
- error = SET_ERROR(EINVAL);
- mutex_exit(&sdp->sd_lock);
+ if (vp->v_mountedhere != NULL)
+ *zfsvfsp = vp->v_mountedhere->mnt_data;
+ vput(vp);
}
-
- VN_RELE(dvp);
-
- return (error);
+ if (*zfsvfsp == NULL)
+ return (SET_ERROR(EINVAL));
+ return (0);
}
/*
@@ -1717,52 +1165,70 @@
int
zfsctl_umount_snapshots(vfs_t *vfsp, int fflags, cred_t *cr)
{
+ char snapname[ZFS_MAX_DATASET_NAME_LEN];
zfsvfs_t *zfsvfs = vfsp->vfs_data;
+ struct mount *mp;
vnode_t *dvp;
- zfsctl_snapdir_t *sdp;
- zfs_snapentry_t *sep, *next;
+ vnode_t *vp;
+ sfs_node_t *node;
+ sfs_node_t *snap;
+ uint64_t cookie;
int error;
ASSERT(zfsvfs->z_ctldir != NULL);
- error = zfsctl_root_lookup(zfsvfs->z_ctldir, "snapshot", &dvp,
- NULL, 0, NULL, cr, NULL, NULL, NULL);
- if (error != 0)
- return (error);
- sdp = dvp->v_data;
- mutex_enter(&sdp->sd_lock);
+ cookie = 0;
+ for (;;) {
+ uint64_t id;
+
+ dsl_pool_config_enter(dmu_objset_pool(zfsvfs->z_os), FTAG);
+ error = dmu_snapshot_list_next(zfsvfs->z_os, sizeof(snapname),
+ snapname, &id, &cookie, NULL);
+ dsl_pool_config_exit(dmu_objset_pool(zfsvfs->z_os), FTAG);
+ if (error != 0) {
+ if (error == ENOENT)
+ error = 0;
+ break;
+ }
- sep = avl_first(&sdp->sd_snaps);
- while (sep != NULL) {
- next = AVL_NEXT(&sdp->sd_snaps, sep);
+ for (;;) {
+ error = sfs_vnode_get(vfsp, LK_EXCLUSIVE,
+ ZFSCTL_INO_SNAPDIR, id, &vp);
+ if (error != 0 || vp == NULL)
+ break;
- /*
- * If this snapshot is not mounted, then it must
- * have just been unmounted by somebody else, and
- * will be cleaned up by zfsctl_snapdir_inactive().
- */
- if (vn_ismntpt(sep->se_root)) {
- error = zfsctl_unmount_snap(sep, fflags, cr);
- if (error) {
- avl_index_t where;
-
- /*
- * Before reinserting snapshot to the tree,
- * check if it was actually removed. For example
- * when snapshot mount point is busy, we will
- * have an error here, but there will be no need
- * to reinsert snapshot.
- */
- if (avl_find(&sdp->sd_snaps, sep, &where) == NULL)
- avl_insert(&sdp->sd_snaps, sep, where);
+ mp = vp->v_mountedhere;
+
+ /*
+ * v_mountedhere being NULL means that the
+ * (uncovered) vnode is in a transient state
+ * (mounting or unmounting), so loop until it
+ * settles down.
+ */
+ if (mp != NULL)
break;
- }
+ vput(vp);
}
- sep = next;
- }
-
- mutex_exit(&sdp->sd_lock);
- VN_RELE(dvp);
+ if (error != 0)
+ break;
+ if (vp == NULL)
+ continue; /* no mountpoint, nothing to do */
+ /*
+ * The mount-point vnode is kept locked to avoid spurious EBUSY
+ * from a concurrent umount.
+ * The vnode lock must have recursive locking enabled.
+ */
+ vfs_ref(mp);
+ error = dounmount(mp, fflags, curthread);
+ KASSERT_IMPLY(error == 0, vrefcnt(vp) == 1,
+ ("extra references after unmount"));
+ vput(vp);
+ if (error != 0)
+ break;
+ }
+ KASSERT_IMPLY((fflags & MS_FORCE) != 0, error == 0,
+ ("force unmounting failed"));
return (error);
}
+
Index: head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_vfsops.c
===================================================================
--- head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_vfsops.c
+++ head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_vfsops.c
@@ -1952,15 +1952,6 @@
if (zfsvfs->z_ctldir != NULL) {
if ((ret = zfsctl_umount_snapshots(vfsp, fflag, cr)) != 0)
return (ret);
- ret = vflush(vfsp, 0, 0, td);
- ASSERT(ret == EBUSY);
- if (!(fflag & MS_FORCE)) {
- if (zfsvfs->z_ctldir->v_count > 1)
- return (EBUSY);
- ASSERT(zfsvfs->z_ctldir->v_count == 1);
- }
- zfsctl_destroy(zfsvfs);
- ASSERT(zfsvfs->z_ctldir == NULL);
}
if (fflag & MS_FORCE) {
@@ -1978,13 +1969,8 @@
* Flush all the files.
*/
ret = vflush(vfsp, 0, (fflag & MS_FORCE) ? FORCECLOSE : 0, td);
- if (ret != 0) {
- if (!zfsvfs->z_issnap) {
- zfsctl_create(zfsvfs);
- ASSERT(zfsvfs->z_ctldir != NULL);
- }
+ if (ret != 0)
return (ret);
- }
#ifdef illumos
if (!(fflag & MS_FORCE)) {
@@ -2095,8 +2081,10 @@
static int
zfs_fhtovp(vfs_t *vfsp, fid_t *fidp, int flags, vnode_t **vpp)
{
+ struct componentname cn;
zfsvfs_t *zfsvfs = vfsp->vfs_data;
znode_t *zp;
+ vnode_t *dvp;
uint64_t object = 0;
uint64_t fid_gen = 0;
uint64_t gen_mask;
@@ -2151,21 +2139,32 @@
if ((fid_gen == 0 &&
(object == ZFSCTL_INO_ROOT || object == ZFSCTL_INO_SNAPDIR)) ||
(zfsvfs->z_shares_dir != 0 && object == zfsvfs->z_shares_dir)) {
- *vpp = zfsvfs->z_ctldir;
- ASSERT(*vpp != NULL);
+ ZFS_EXIT(zfsvfs);
+ VERIFY0(zfsctl_root(zfsvfs, LK_SHARED, &dvp));
if (object == ZFSCTL_INO_SNAPDIR) {
- VERIFY(zfsctl_root_lookup(*vpp, "snapshot", vpp, NULL,
- 0, NULL, NULL, NULL, NULL, NULL) == 0);
+ cn.cn_nameptr = "snapshot";
+ cn.cn_namelen = strlen(cn.cn_nameptr);
+ cn.cn_nameiop = LOOKUP;
+ cn.cn_flags = ISLASTCN | LOCKLEAF;
+ cn.cn_lkflags = flags;
+ VERIFY0(VOP_LOOKUP(dvp, vpp, &cn));
+ vput(dvp);
} else if (object == zfsvfs->z_shares_dir) {
- VERIFY(zfsctl_root_lookup(*vpp, "shares", vpp, NULL,
- 0, NULL, NULL, NULL, NULL, NULL) == 0);
+ /*
+ * XXX This branch must not be taken,
+ * if it is, then the lookup below will
+ * explode.
+ */
+ cn.cn_nameptr = "shares";
+ cn.cn_namelen = strlen(cn.cn_nameptr);
+ cn.cn_nameiop = LOOKUP;
+ cn.cn_flags = ISLASTCN;
+ cn.cn_lkflags = flags;
+ VERIFY0(VOP_LOOKUP(dvp, vpp, &cn));
+ vput(dvp);
} else {
- vref(*vpp);
+ *vpp = dvp;
}
- ZFS_EXIT(zfsvfs);
- err = vn_lock(*vpp, flags);
- if (err != 0)
- *vpp = NULL;
return (err);
}
@@ -2190,7 +2189,7 @@
*vpp = ZTOV(zp);
ZFS_EXIT(zfsvfs);
- err = vn_lock(*vpp, flags | LK_RETRY);
+ err = vn_lock(*vpp, flags);
if (err == 0)
vnode_create_vobject(*vpp, zp->z_size, curthread);
else
Index: head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_vnops.c
===================================================================
--- head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_vnops.c
+++ head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_vnops.c
@@ -1612,27 +1612,34 @@
* the vp for the snapshot directory.
*/
if (zdp->z_id == zfsvfs->z_root && zfsvfs->z_parent != zfsvfs) {
- error = zfsctl_root_lookup(zfsvfs->z_parent->z_ctldir,
- "snapshot", vpp, NULL, 0, NULL, kcred,
- NULL, NULL, NULL);
+ struct componentname cn;
+ vnode_t *zfsctl_vp;
+ int ltype;
+
ZFS_EXIT(zfsvfs);
+ ltype = VOP_ISLOCKED(dvp);
+ VOP_UNLOCK(dvp, 0);
+ error = zfsctl_root(zfsvfs->z_parent, LK_SHARED,
+ &zfsctl_vp);
if (error == 0) {
- error = zfs_lookup_lock(dvp, *vpp, nm,
- cnp->cn_lkflags);
+ cn.cn_nameptr = "snapshot";
+ cn.cn_namelen = strlen(cn.cn_nameptr);
+ cn.cn_nameiop = cnp->cn_nameiop;
+ cn.cn_flags = cnp->cn_flags;
+ cn.cn_lkflags = cnp->cn_lkflags;
+ error = VOP_LOOKUP(zfsctl_vp, vpp, &cn);
+ vput(zfsctl_vp);
}
- goto out;
+ vn_lock(dvp, ltype | LK_RETRY);
+ return (error);
}
}
if (zfs_has_ctldir(zdp) && strcmp(nm, ZFS_CTLDIR_NAME) == 0) {
- error = 0;
- if ((cnp->cn_flags & ISLASTCN) != 0 && nameiop != LOOKUP)
- error = SET_ERROR(ENOTSUP);
- else
- *vpp = zfsctl_root(zdp);
ZFS_EXIT(zfsvfs);
- if (error == 0)
- error = zfs_lookup_lock(dvp, *vpp, nm, cnp->cn_lkflags);
- goto out;
+ if ((cnp->cn_flags & ISLASTCN) != 0 && nameiop != LOOKUP)
+ return (SET_ERROR(ENOTSUP));
+ error = zfsctl_root(zfsvfs, cnp->cn_lkflags, vpp);
+ return (error);
}
/*
@@ -5956,7 +5963,7 @@
vhold(covered_vp);
ltype = VOP_ISLOCKED(vp);
VOP_UNLOCK(vp, 0);
- error = vget(covered_vp, LK_EXCLUSIVE | LK_VNHELD, curthread);
+ error = vget(covered_vp, LK_SHARED | LK_VNHELD, curthread);
if (error == 0) {
error = VOP_VPTOCNP(covered_vp, ap->a_vpp, ap->a_cred,
ap->a_buf, ap->a_buflen);

File Metadata

Mime Type
text/plain
Expires
Thu, Dec 26, 11:58 AM (11 h, 37 m)
Storage Engine
blob
Storage Format
Raw Data
Storage Handle
15604483
Default Alt Text
D7421.id25481.diff (66 KB)

Event Timeline