* CDDL HEADER END
*/
/*
- * Copyright 2008 Sun Microsystems, Inc. All rights reserved.
- * Use is subject to license terms.
+ * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
*/
+/* Portions Copyright 2010 Robert Milkowski */
+
#include <sys/types.h>
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/cmn_err.h>
#include "fs/fs_subr.h"
#include <sys/zfs_znode.h>
+#include <sys/zfs_vnops.h>
#include <sys/zfs_dir.h>
#include <sys/zil.h>
#include <sys/fs/zfs.h>
#include <sys/dsl_deleg.h>
#include <sys/spa.h>
#include <sys/zap.h>
+#include <sys/sa.h>
#include <sys/varargs.h>
#include <sys/policy.h>
#include <sys/atomic.h>
#include <sys/modctl.h>
#include <sys/refstr.h>
#include <sys/zfs_ioctl.h>
-#include <sys/zfs_ctldir.h>
#include <sys/zfs_fuid.h>
#include <sys/bootconf.h>
#include <sys/sunddi.h>
#include <sys/dnlc.h>
#include <sys/dmu_objset.h>
#include <sys/spa_boot.h>
+#include <sys/sa.h>
+#include <sys/zpl.h>
+#include "zfs_comutil.h"
-int zfsfstype;
-vfsops_t *zfs_vfsops = NULL;
-static major_t zfs_major;
-static minor_t zfs_minor;
-static kmutex_t zfs_dev_mtx;
-
-static int zfs_mount(vfs_t *vfsp, vnode_t *mvp, struct mounta *uap, cred_t *cr);
-static int zfs_umount(vfs_t *vfsp, int fflag, cred_t *cr);
-static int zfs_mountroot(vfs_t *vfsp, enum whymountroot);
-static int zfs_root(vfs_t *vfsp, vnode_t **vpp);
-static int zfs_statvfs(vfs_t *vfsp, struct statvfs64 *statp);
-static int zfs_vget(vfs_t *vfsp, vnode_t **vpp, fid_t *fidp);
-static void zfs_freevfs(vfs_t *vfsp);
-
-static const fs_operation_def_t zfs_vfsops_template[] = {
- VFSNAME_MOUNT, { .vfs_mount = zfs_mount },
- VFSNAME_MOUNTROOT, { .vfs_mountroot = zfs_mountroot },
- VFSNAME_UNMOUNT, { .vfs_unmount = zfs_umount },
- VFSNAME_ROOT, { .vfs_root = zfs_root },
- VFSNAME_STATVFS, { .vfs_statvfs = zfs_statvfs },
- VFSNAME_SYNC, { .vfs_sync = zfs_sync },
- VFSNAME_VGET, { .vfs_vget = zfs_vget },
- VFSNAME_FREEVFS, { .vfs_freevfs = zfs_freevfs },
- NULL, NULL
-};
-
-static const fs_operation_def_t zfs_vfsops_eio_template[] = {
- VFSNAME_FREEVFS, { .vfs_freevfs = zfs_freevfs },
- NULL, NULL
-};
-
-/*
- * We need to keep a count of active fs's.
- * This is necessary to prevent our module
- * from being unloaded after a umount -f
- */
-static uint32_t zfs_active_fs_count = 0;
-
-static char *noatime_cancel[] = { MNTOPT_ATIME, NULL };
-static char *atime_cancel[] = { MNTOPT_NOATIME, NULL };
-static char *noxattr_cancel[] = { MNTOPT_XATTR, NULL };
-static char *xattr_cancel[] = { MNTOPT_NOXATTR, NULL };
-
-/*
- * MO_DEFAULT is not used since the default value is determined
- * by the equivalent property.
- */
-static mntopt_t mntopts[] = {
- { MNTOPT_NOXATTR, noxattr_cancel, NULL, 0, NULL },
- { MNTOPT_XATTR, xattr_cancel, NULL, 0, NULL },
- { MNTOPT_NOATIME, noatime_cancel, NULL, 0, NULL },
- { MNTOPT_ATIME, atime_cancel, NULL, 0, NULL }
-};
-
-static mntopts_t zfs_mntopts = {
- sizeof (mntopts) / sizeof (mntopt_t),
- mntopts
-};
/*ARGSUSED*/
int
-zfs_sync(vfs_t *vfsp, short flag, cred_t *cr)
+zfs_sync(struct super_block *sb, int wait, cred_t *cr)
{
+ zfs_sb_t *zsb = sb->s_fs_info;
+
/*
* 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)
+ if (unlikely(oops_in_progress))
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)
+ if (!wait)
return (0);
- if (vfsp != NULL) {
+ if (zsb != NULL) {
/*
* Sync a specific filesystem.
*/
- zfsvfs_t *zfsvfs = vfsp->vfs_data;
+ dsl_pool_t *dp;
- 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);
+ ZFS_ENTER(zsb);
+ dp = dmu_objset_pool(zsb->z_os);
+
+ /*
+ * If the system is shutting down, then skip any
+ * filesystems which may exist on a suspended pool.
+ */
+ if (spa_suspended(dp->dp_spa)) {
+ ZFS_EXIT(zsb);
+ return (0);
+ }
+
+ if (zsb->z_log != NULL)
+ zil_commit(zsb->z_log, 0);
+
+ ZFS_EXIT(zsb);
} else {
/*
* Sync all ZFS filesystems. This is what happens when you
return (0);
}
+EXPORT_SYMBOL(zfs_sync);
-static int
-zfs_create_unique_device(dev_t *dev)
+boolean_t
+zfs_is_readonly(zfs_sb_t *zsb)
{
- major_t new_major;
-
- do {
- ASSERT3U(zfs_minor, <=, MAXMIN32);
- minor_t start = zfs_minor;
- do {
- mutex_enter(&zfs_dev_mtx);
- if (zfs_minor >= MAXMIN32) {
- /*
- * If we're still using the real major
- * keep out of /dev/zfs and /dev/zvol minor
- * number space. If we're using a getudev()'ed
- * major number, we can use all of its minors.
- */
- if (zfs_major == ddi_name_to_major(ZFS_DRIVER))
- zfs_minor = ZFS_MIN_MINOR;
- else
- zfs_minor = 0;
- } else {
- zfs_minor++;
- }
- *dev = makedevice(zfs_major, zfs_minor);
- mutex_exit(&zfs_dev_mtx);
- } while (vfs_devismounted(*dev) && zfs_minor != start);
- if (zfs_minor == start) {
- /*
- * We are using all ~262,000 minor numbers for the
- * current major number. Create a new major number.
- */
- if ((new_major = getudev()) == (major_t)-1) {
- cmn_err(CE_WARN,
- "zfs_mount: Can't get unique major "
- "device number.");
- return (-1);
- }
- mutex_enter(&zfs_dev_mtx);
- zfs_major = new_major;
- zfs_minor = 0;
-
- mutex_exit(&zfs_dev_mtx);
- } else {
- break;
- }
- /* CONSTANTCONDITION */
- } while (1);
-
- return (0);
+ return (!!(zsb->z_sb->s_flags & MS_RDONLY));
}
+EXPORT_SYMBOL(zfs_is_readonly);
static void
atime_changed_cb(void *arg, uint64_t newval)
{
- zfsvfs_t *zfsvfs = arg;
-
- if (newval == TRUE) {
- zfsvfs->z_atime = TRUE;
- vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NOATIME);
- vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_ATIME, NULL, 0);
- } else {
- zfsvfs->z_atime = FALSE;
- vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_ATIME);
- vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NOATIME, NULL, 0);
- }
+ ((zfs_sb_t *)arg)->z_atime = newval;
}
static void
xattr_changed_cb(void *arg, uint64_t newval)
{
- zfsvfs_t *zfsvfs = arg;
+ zfs_sb_t *zsb = arg;
- if (newval == TRUE) {
- /* XXX locking on vfs_flag? */
- zfsvfs->z_vfs->vfs_flag |= VFS_XATTR;
- vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NOXATTR);
- vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_XATTR, NULL, 0);
- } else {
- /* XXX locking on vfs_flag? */
- zfsvfs->z_vfs->vfs_flag &= ~VFS_XATTR;
- vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_XATTR);
- vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NOXATTR, NULL, 0);
- }
+ if (newval == TRUE)
+ zsb->z_flags |= ZSB_XATTR;
+ else
+ zsb->z_flags &= ~ZSB_XATTR;
}
static void
blksz_changed_cb(void *arg, uint64_t newval)
{
- zfsvfs_t *zfsvfs = arg;
+ zfs_sb_t *zsb = arg;
if (newval < SPA_MINBLOCKSIZE ||
newval > SPA_MAXBLOCKSIZE || !ISP2(newval))
newval = SPA_MAXBLOCKSIZE;
- zfsvfs->z_max_blksz = newval;
- zfsvfs->z_vfs->vfs_bsize = newval;
+ zsb->z_max_blksz = newval;
}
static void
readonly_changed_cb(void *arg, uint64_t newval)
{
- zfsvfs_t *zfsvfs = arg;
+ zfs_sb_t *zsb = arg;
+ struct super_block *sb = zsb->z_sb;
- if (newval) {
- /* XXX locking on vfs_flag? */
- zfsvfs->z_vfs->vfs_flag |= VFS_RDONLY;
- vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_RW);
- vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_RO, NULL, 0);
- } else {
- /* XXX locking on vfs_flag? */
- zfsvfs->z_vfs->vfs_flag &= ~VFS_RDONLY;
- vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_RO);
- vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_RW, NULL, 0);
- }
+ if (sb == NULL)
+ return;
+
+ if (newval)
+ sb->s_flags |= MS_RDONLY;
+ else
+ sb->s_flags &= ~MS_RDONLY;
}
static void
devices_changed_cb(void *arg, uint64_t newval)
{
- zfsvfs_t *zfsvfs = arg;
-
- if (newval == FALSE) {
- zfsvfs->z_vfs->vfs_flag |= VFS_NODEVICES;
- vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_DEVICES);
- vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NODEVICES, NULL, 0);
- } else {
- zfsvfs->z_vfs->vfs_flag &= ~VFS_NODEVICES;
- vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NODEVICES);
- vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_DEVICES, NULL, 0);
- }
}
static void
setuid_changed_cb(void *arg, uint64_t newval)
{
- zfsvfs_t *zfsvfs = arg;
-
- if (newval == FALSE) {
- zfsvfs->z_vfs->vfs_flag |= VFS_NOSETUID;
- vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_SETUID);
- vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NOSETUID, NULL, 0);
- } else {
- zfsvfs->z_vfs->vfs_flag &= ~VFS_NOSETUID;
- vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NOSETUID);
- vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_SETUID, NULL, 0);
- }
}
static void
exec_changed_cb(void *arg, uint64_t newval)
{
- zfsvfs_t *zfsvfs = arg;
-
- if (newval == FALSE) {
- zfsvfs->z_vfs->vfs_flag |= VFS_NOEXEC;
- vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_EXEC);
- vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NOEXEC, NULL, 0);
- } else {
- zfsvfs->z_vfs->vfs_flag &= ~VFS_NOEXEC;
- vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NOEXEC);
- vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_EXEC, NULL, 0);
- }
}
-/*
- * The nbmand mount option can be changed at mount time.
- * We can't allow it to be toggled on live file systems or incorrect
- * behavior may be seen from cifs clients
- *
- * This property isn't registered via dsl_prop_register(), but this callback
- * will be called when a file system is first mounted
- */
static void
nbmand_changed_cb(void *arg, uint64_t newval)
{
- zfsvfs_t *zfsvfs = arg;
- if (newval == FALSE) {
- vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NBMAND);
- vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NONBMAND, NULL, 0);
- } else {
- vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NONBMAND);
- vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NBMAND, NULL, 0);
- }
-}
+ zfs_sb_t *zsb = arg;
+ struct super_block *sb = zsb->z_sb;
-static void
-snapdir_changed_cb(void *arg, uint64_t newval)
-{
- zfsvfs_t *zfsvfs = arg;
+ if (sb == NULL)
+ return;
- zfsvfs->z_show_ctldir = newval;
+ if (newval == TRUE)
+ sb->s_flags |= MS_MANDLOCK;
+ else
+ sb->s_flags &= ~MS_MANDLOCK;
}
static void
-vscan_changed_cb(void *arg, uint64_t newval)
+snapdir_changed_cb(void *arg, uint64_t newval)
{
- zfsvfs_t *zfsvfs = arg;
-
- zfsvfs->z_vscan = newval;
+ ((zfs_sb_t *)arg)->z_show_ctldir = newval;
}
static void
-acl_mode_changed_cb(void *arg, uint64_t newval)
+vscan_changed_cb(void *arg, uint64_t newval)
{
- zfsvfs_t *zfsvfs = arg;
-
- zfsvfs->z_acl_mode = newval;
+ ((zfs_sb_t *)arg)->z_vscan = newval;
}
static void
acl_inherit_changed_cb(void *arg, uint64_t newval)
{
- zfsvfs_t *zfsvfs = arg;
-
- zfsvfs->z_acl_inherit = newval;
+ ((zfs_sb_t *)arg)->z_acl_inherit = newval;
}
-static int
-zfs_register_callbacks(vfs_t *vfsp)
+int
+zfs_register_callbacks(zfs_sb_t *zsb)
{
struct dsl_dataset *ds = NULL;
- objset_t *os = NULL;
- zfsvfs_t *zfsvfs = NULL;
- uint64_t nbmand;
- int readonly, do_readonly = B_FALSE;
- int setuid, do_setuid = B_FALSE;
- int exec, do_exec = B_FALSE;
- int devices, do_devices = B_FALSE;
- int xattr, do_xattr = B_FALSE;
- int atime, do_atime = B_FALSE;
+ objset_t *os = zsb->z_os;
int error = 0;
- ASSERT(vfsp);
- zfsvfs = vfsp->vfs_data;
- ASSERT(zfsvfs);
- os = zfsvfs->z_os;
-
- /*
- * The act of registering our callbacks will destroy any mount
- * options we may have. In order to enable temporary overrides
- * of mount options, we stash away the current values and
- * restore them after we register the callbacks.
- */
- if (vfs_optionisset(vfsp, MNTOPT_RO, NULL)) {
- readonly = B_TRUE;
- do_readonly = B_TRUE;
- } else if (vfs_optionisset(vfsp, MNTOPT_RW, NULL)) {
- readonly = B_FALSE;
- do_readonly = B_TRUE;
- }
- if (vfs_optionisset(vfsp, MNTOPT_NOSUID, NULL)) {
- devices = B_FALSE;
- setuid = B_FALSE;
- do_devices = B_TRUE;
- do_setuid = B_TRUE;
- } else {
- if (vfs_optionisset(vfsp, MNTOPT_NODEVICES, NULL)) {
- devices = B_FALSE;
- do_devices = B_TRUE;
- } else if (vfs_optionisset(vfsp, MNTOPT_DEVICES, NULL)) {
- devices = B_TRUE;
- do_devices = B_TRUE;
- }
-
- if (vfs_optionisset(vfsp, MNTOPT_NOSETUID, NULL)) {
- setuid = B_FALSE;
- do_setuid = B_TRUE;
- } else if (vfs_optionisset(vfsp, MNTOPT_SETUID, NULL)) {
- setuid = B_TRUE;
- do_setuid = B_TRUE;
- }
- }
- if (vfs_optionisset(vfsp, MNTOPT_NOEXEC, NULL)) {
- exec = B_FALSE;
- do_exec = B_TRUE;
- } else if (vfs_optionisset(vfsp, MNTOPT_EXEC, NULL)) {
- exec = B_TRUE;
- do_exec = B_TRUE;
- }
- if (vfs_optionisset(vfsp, MNTOPT_NOXATTR, NULL)) {
- xattr = B_FALSE;
- do_xattr = B_TRUE;
- } else if (vfs_optionisset(vfsp, MNTOPT_XATTR, NULL)) {
- xattr = B_TRUE;
- do_xattr = B_TRUE;
- }
- if (vfs_optionisset(vfsp, MNTOPT_NOATIME, NULL)) {
- atime = B_FALSE;
- do_atime = B_TRUE;
- } else if (vfs_optionisset(vfsp, MNTOPT_ATIME, NULL)) {
- atime = B_TRUE;
- do_atime = B_TRUE;
- }
-
- /*
- * nbmand is a special property. It can only be changed at
- * mount time.
- *
- * This is weird, but it is documented to only be changeable
- * at mount time.
- */
- if (vfs_optionisset(vfsp, MNTOPT_NONBMAND, NULL)) {
- nbmand = B_FALSE;
- } else if (vfs_optionisset(vfsp, MNTOPT_NBMAND, NULL)) {
- nbmand = B_TRUE;
- } else {
- char osname[MAXNAMELEN];
-
- dmu_objset_name(os, osname);
- if (error = dsl_prop_get_integer(osname, "nbmand", &nbmand,
- NULL)) {
- return (error);
- }
- }
+ if (zfs_is_readonly(zsb) || !spa_writeable(dmu_objset_spa(os)))
+ readonly_changed_cb(zsb, B_TRUE);
/*
* Register property callbacks.
* overboard...
*/
ds = dmu_objset_ds(os);
- error = dsl_prop_register(ds, "atime", atime_changed_cb, zfsvfs);
+ error = dsl_prop_register(ds,
+ "atime", atime_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
- "xattr", xattr_changed_cb, zfsvfs);
+ "xattr", xattr_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
- "recordsize", blksz_changed_cb, zfsvfs);
+ "recordsize", blksz_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
- "readonly", readonly_changed_cb, zfsvfs);
+ "readonly", readonly_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
- "devices", devices_changed_cb, zfsvfs);
+ "devices", devices_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
- "setuid", setuid_changed_cb, zfsvfs);
+ "setuid", setuid_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
- "exec", exec_changed_cb, zfsvfs);
+ "exec", exec_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
- "snapdir", snapdir_changed_cb, zfsvfs);
+ "snapdir", snapdir_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
- "aclmode", acl_mode_changed_cb, zfsvfs);
+ "aclinherit", acl_inherit_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
- "aclinherit", acl_inherit_changed_cb, zfsvfs);
+ "vscan", vscan_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
- "vscan", vscan_changed_cb, zfsvfs);
+ "nbmand", nbmand_changed_cb, zsb);
if (error)
goto unregister;
- /*
- * Invoke our callbacks to restore temporary mount options.
- */
- if (do_readonly)
- readonly_changed_cb(zfsvfs, readonly);
- if (do_setuid)
- setuid_changed_cb(zfsvfs, setuid);
- if (do_exec)
- exec_changed_cb(zfsvfs, exec);
- if (do_devices)
- devices_changed_cb(zfsvfs, devices);
- if (do_xattr)
- xattr_changed_cb(zfsvfs, xattr);
- if (do_atime)
- atime_changed_cb(zfsvfs, atime);
-
- nbmand_changed_cb(zfsvfs, nbmand);
-
return (0);
unregister:
* registered, but this is OK; it will simply return ENOMSG,
* which we will ignore.
*/
- (void) dsl_prop_unregister(ds, "atime", atime_changed_cb, zfsvfs);
- (void) dsl_prop_unregister(ds, "xattr", xattr_changed_cb, zfsvfs);
- (void) dsl_prop_unregister(ds, "recordsize", blksz_changed_cb, zfsvfs);
- (void) dsl_prop_unregister(ds, "readonly", readonly_changed_cb, zfsvfs);
- (void) dsl_prop_unregister(ds, "devices", devices_changed_cb, zfsvfs);
- (void) dsl_prop_unregister(ds, "setuid", setuid_changed_cb, zfsvfs);
- (void) dsl_prop_unregister(ds, "exec", exec_changed_cb, zfsvfs);
- (void) dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb, zfsvfs);
- (void) dsl_prop_unregister(ds, "aclmode", acl_mode_changed_cb, zfsvfs);
+ (void) dsl_prop_unregister(ds, "atime", atime_changed_cb, zsb);
+ (void) dsl_prop_unregister(ds, "xattr", xattr_changed_cb, zsb);
+ (void) dsl_prop_unregister(ds, "recordsize", blksz_changed_cb, zsb);
+ (void) dsl_prop_unregister(ds, "readonly", readonly_changed_cb, zsb);
+ (void) dsl_prop_unregister(ds, "devices", devices_changed_cb, zsb);
+ (void) dsl_prop_unregister(ds, "setuid", setuid_changed_cb, zsb);
+ (void) dsl_prop_unregister(ds, "exec", exec_changed_cb, zsb);
+ (void) dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb, zsb);
(void) dsl_prop_unregister(ds, "aclinherit", acl_inherit_changed_cb,
- zfsvfs);
- (void) dsl_prop_unregister(ds, "vscan", vscan_changed_cb, zfsvfs);
- return (error);
+ zsb);
+ (void) dsl_prop_unregister(ds, "vscan", vscan_changed_cb, zsb);
+ (void) dsl_prop_unregister(ds, "nbmand", nbmand_changed_cb, zsb);
+ return (error);
}
+EXPORT_SYMBOL(zfs_register_callbacks);
static int
-zfsvfs_setup(zfsvfs_t *zfsvfs, boolean_t mounting)
+zfs_space_delta_cb(dmu_object_type_t bonustype, void *data,
+ uint64_t *userp, uint64_t *groupp)
{
- int error;
-
- error = zfs_register_callbacks(zfsvfs->z_vfs);
- if (error)
- return (error);
+ znode_phys_t *znp = data;
+ int error = 0;
/*
- * Set the objset user_ptr to track its zfsvfs.
+ * Is it a valid type of object to track?
*/
- mutex_enter(&zfsvfs->z_os->os->os_user_ptr_lock);
- dmu_objset_set_user(zfsvfs->z_os, zfsvfs);
- mutex_exit(&zfsvfs->z_os->os->os_user_ptr_lock);
+ if (bonustype != DMU_OT_ZNODE && bonustype != DMU_OT_SA)
+ return (ENOENT);
/*
- * If we are not mounting (ie: online recv), then we don't
- * have to worry about replaying the log as we blocked all
- * operations out since we closed the ZIL.
+ * If we have a NULL data pointer
+ * then assume the id's aren't changing and
+ * return EEXIST to the dmu to let it know to
+ * use the same ids
*/
- if (mounting) {
- boolean_t readonly;
+ if (data == NULL)
+ return (EEXIST);
- /*
- * During replay we remove the read only flag to
- * allow replays to succeed.
- */
- readonly = zfsvfs->z_vfs->vfs_flag & VFS_RDONLY;
- zfsvfs->z_vfs->vfs_flag &= ~VFS_RDONLY;
+ if (bonustype == DMU_OT_ZNODE) {
+ *userp = znp->zp_uid;
+ *groupp = znp->zp_gid;
+ } else {
+ int hdrsize;
- /*
- * Parse and replay the intent log.
- */
- zil_replay(zfsvfs->z_os, zfsvfs, &zfsvfs->z_assign,
- zfs_replay_vector, zfs_unlinked_drain);
+ ASSERT(bonustype == DMU_OT_SA);
+ hdrsize = sa_hdrsize(data);
- zfs_unlinked_drain(zfsvfs);
- zfsvfs->z_vfs->vfs_flag |= readonly; /* restore readonly bit */
+ if (hdrsize != 0) {
+ *userp = *((uint64_t *)((uintptr_t)data + hdrsize +
+ SA_UID_OFFSET));
+ *groupp = *((uint64_t *)((uintptr_t)data + hdrsize +
+ SA_GID_OFFSET));
+ } else {
+ /*
+ * This should only happen for newly created
+ * files that haven't had the znode data filled
+ * in yet.
+ */
+ *userp = 0;
+ *groupp = 0;
+ }
}
+ return (error);
+}
+
+static void
+fuidstr_to_sid(zfs_sb_t *zsb, const char *fuidstr,
+ char *domainbuf, int buflen, uid_t *ridp)
+{
+ uint64_t fuid;
+ const char *domain;
- if (!zil_disable)
- zfsvfs->z_log = zil_open(zfsvfs->z_os, zfs_get_data);
+ fuid = strtonum(fuidstr, NULL);
+ domain = zfs_fuid_find_by_idx(zsb, FUID_INDEX(fuid));
+ if (domain)
+ (void) strlcpy(domainbuf, domain, buflen);
+ else
+ domainbuf[0] = '\0';
+ *ridp = FUID_RID(fuid);
+}
+
+static uint64_t
+zfs_userquota_prop_to_obj(zfs_sb_t *zsb, zfs_userquota_prop_t type)
+{
+ switch (type) {
+ case ZFS_PROP_USERUSED:
+ return (DMU_USERUSED_OBJECT);
+ case ZFS_PROP_GROUPUSED:
+ return (DMU_GROUPUSED_OBJECT);
+ case ZFS_PROP_USERQUOTA:
+ return (zsb->z_userquota_obj);
+ case ZFS_PROP_GROUPQUOTA:
+ return (zsb->z_groupquota_obj);
+ default:
+ return (ENOTSUP);
+ }
return (0);
}
-static void
-zfs_freezfsvfs(zfsvfs_t *zfsvfs)
+int
+zfs_userspace_many(zfs_sb_t *zsb, zfs_userquota_prop_t type,
+ uint64_t *cookiep, void *vbuf, uint64_t *bufsizep)
{
- mutex_destroy(&zfsvfs->z_znodes_lock);
- mutex_destroy(&zfsvfs->z_online_recv_lock);
- list_destroy(&zfsvfs->z_all_znodes);
- rrw_destroy(&zfsvfs->z_teardown_lock);
- rw_destroy(&zfsvfs->z_teardown_inactive_lock);
- rw_destroy(&zfsvfs->z_fuid_lock);
- kmem_free(zfsvfs, sizeof (zfsvfs_t));
+ int error;
+ zap_cursor_t zc;
+ zap_attribute_t za;
+ zfs_useracct_t *buf = vbuf;
+ uint64_t obj;
+
+ if (!dmu_objset_userspace_present(zsb->z_os))
+ return (ENOTSUP);
+
+ obj = zfs_userquota_prop_to_obj(zsb, type);
+ if (obj == 0) {
+ *bufsizep = 0;
+ return (0);
+ }
+
+ for (zap_cursor_init_serialized(&zc, zsb->z_os, obj, *cookiep);
+ (error = zap_cursor_retrieve(&zc, &za)) == 0;
+ zap_cursor_advance(&zc)) {
+ if ((uintptr_t)buf - (uintptr_t)vbuf + sizeof (zfs_useracct_t) >
+ *bufsizep)
+ break;
+
+ fuidstr_to_sid(zsb, za.za_name,
+ buf->zu_domain, sizeof (buf->zu_domain), &buf->zu_rid);
+
+ buf->zu_space = za.za_first_integer;
+ buf++;
+ }
+ if (error == ENOENT)
+ error = 0;
+
+ ASSERT3U((uintptr_t)buf - (uintptr_t)vbuf, <=, *bufsizep);
+ *bufsizep = (uintptr_t)buf - (uintptr_t)vbuf;
+ *cookiep = zap_cursor_serialize(&zc);
+ zap_cursor_fini(&zc);
+ return (error);
}
+EXPORT_SYMBOL(zfs_userspace_many);
+/*
+ * buf must be big enough (eg, 32 bytes)
+ */
static int
-zfs_domount(vfs_t *vfsp, char *osname)
+id_to_fuidstr(zfs_sb_t *zsb, const char *domain, uid_t rid,
+ char *buf, boolean_t addok)
{
- dev_t mount_dev;
- uint64_t recordsize, readonly;
- int error = 0;
- int mode;
- zfsvfs_t *zfsvfs;
- znode_t *zp = NULL;
+ uint64_t fuid;
+ int domainid = 0;
- ASSERT(vfsp);
- ASSERT(osname);
+ if (domain && domain[0]) {
+ domainid = zfs_fuid_find_by_domain(zsb, domain, NULL, addok);
+ if (domainid == -1)
+ return (ENOENT);
+ }
+ fuid = FUID_ENCODE(domainid, rid);
+ (void) sprintf(buf, "%llx", (longlong_t)fuid);
+ return (0);
+}
- /*
- * Initialize the zfs-specific filesystem structure.
- * Should probably make this a kmem cache, shuffle fields,
- * and just bzero up to z_hold_mtx[].
- */
- zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP);
- zfsvfs->z_vfs = vfsp;
- zfsvfs->z_parent = zfsvfs;
- zfsvfs->z_assign = TXG_NOWAIT;
- zfsvfs->z_max_blksz = SPA_MAXBLOCKSIZE;
- zfsvfs->z_show_ctldir = ZFS_SNAPDIR_VISIBLE;
-
- mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
- mutex_init(&zfsvfs->z_online_recv_lock, NULL, MUTEX_DEFAULT, NULL);
- list_create(&zfsvfs->z_all_znodes, sizeof (znode_t),
- offsetof(znode_t, z_link_node));
- rrw_init(&zfsvfs->z_teardown_lock);
- rw_init(&zfsvfs->z_teardown_inactive_lock, NULL, RW_DEFAULT, NULL);
- rw_init(&zfsvfs->z_fuid_lock, NULL, RW_DEFAULT, NULL);
+int
+zfs_userspace_one(zfs_sb_t *zsb, zfs_userquota_prop_t type,
+ const char *domain, uint64_t rid, uint64_t *valp)
+{
+ char buf[32];
+ int err;
+ uint64_t obj;
- /* Initialize the generic filesystem structure. */
- vfsp->vfs_bcount = 0;
- vfsp->vfs_data = NULL;
+ *valp = 0;
- if (zfs_create_unique_device(&mount_dev) == -1) {
- error = ENODEV;
- goto out;
- }
- ASSERT(vfs_devismounted(mount_dev) == 0);
+ if (!dmu_objset_userspace_present(zsb->z_os))
+ return (ENOTSUP);
- if (error = dsl_prop_get_integer(osname, "recordsize", &recordsize,
- NULL))
- goto out;
+ obj = zfs_userquota_prop_to_obj(zsb, type);
+ if (obj == 0)
+ return (0);
- vfsp->vfs_dev = mount_dev;
- vfsp->vfs_fstype = zfsfstype;
- vfsp->vfs_bsize = recordsize;
- vfsp->vfs_flag |= VFS_NOTRUNC;
- vfsp->vfs_data = zfsvfs;
+ err = id_to_fuidstr(zsb, domain, rid, buf, B_FALSE);
+ if (err)
+ return (err);
- if (error = dsl_prop_get_integer(osname, "readonly", &readonly, NULL))
- goto out;
+ err = zap_lookup(zsb->z_os, obj, buf, 8, 1, valp);
+ if (err == ENOENT)
+ err = 0;
+ return (err);
+}
+EXPORT_SYMBOL(zfs_userspace_one);
- mode = DS_MODE_OWNER;
- if (readonly)
- mode |= DS_MODE_READONLY;
+int
+zfs_set_userquota(zfs_sb_t *zsb, zfs_userquota_prop_t type,
+ const char *domain, uint64_t rid, uint64_t quota)
+{
+ char buf[32];
+ int err;
+ dmu_tx_t *tx;
+ uint64_t *objp;
+ boolean_t fuid_dirtied;
- error = dmu_objset_open(osname, DMU_OST_ZFS, mode, &zfsvfs->z_os);
- if (error == EROFS) {
- mode = DS_MODE_OWNER | DS_MODE_READONLY;
- error = dmu_objset_open(osname, DMU_OST_ZFS, mode,
- &zfsvfs->z_os);
- }
+ if (type != ZFS_PROP_USERQUOTA && type != ZFS_PROP_GROUPQUOTA)
+ return (EINVAL);
- if (error)
- goto out;
+ if (zsb->z_version < ZPL_VERSION_USERSPACE)
+ return (ENOTSUP);
- if (error = zfs_init_fs(zfsvfs, &zp))
- goto out;
+ objp = (type == ZFS_PROP_USERQUOTA) ? &zsb->z_userquota_obj :
+ &zsb->z_groupquota_obj;
- /* The call to zfs_init_fs leaves the vnode held, release it here. */
- VN_RELE(ZTOV(zp));
+ err = id_to_fuidstr(zsb, domain, rid, buf, B_TRUE);
+ if (err)
+ return (err);
+ fuid_dirtied = zsb->z_fuid_dirty;
- /*
- * Set features for file system.
- */
- zfsvfs->z_use_fuids = USE_FUIDS(zfsvfs->z_version, zfsvfs->z_os);
- if (zfsvfs->z_use_fuids) {
- vfs_set_feature(vfsp, VFSFT_XVATTR);
- vfs_set_feature(vfsp, VFSFT_SYSATTR_VIEWS);
- vfs_set_feature(vfsp, VFSFT_ACEMASKONACCESS);
- vfs_set_feature(vfsp, VFSFT_ACLONCREATE);
+ tx = dmu_tx_create(zsb->z_os);
+ dmu_tx_hold_zap(tx, *objp ? *objp : DMU_NEW_OBJECT, B_TRUE, NULL);
+ if (*objp == 0) {
+ dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE,
+ zfs_userquota_prop_prefixes[type]);
}
- if (zfsvfs->z_case == ZFS_CASE_INSENSITIVE) {
- vfs_set_feature(vfsp, VFSFT_DIRENTFLAGS);
- vfs_set_feature(vfsp, VFSFT_CASEINSENSITIVE);
- vfs_set_feature(vfsp, VFSFT_NOCASESENSITIVE);
- } else if (zfsvfs->z_case == ZFS_CASE_MIXED) {
- vfs_set_feature(vfsp, VFSFT_DIRENTFLAGS);
- vfs_set_feature(vfsp, VFSFT_CASEINSENSITIVE);
+ if (fuid_dirtied)
+ zfs_fuid_txhold(zsb, tx);
+ err = dmu_tx_assign(tx, TXG_WAIT);
+ if (err) {
+ dmu_tx_abort(tx);
+ return (err);
}
- if (dmu_objset_is_snapshot(zfsvfs->z_os)) {
- uint64_t pval;
-
- ASSERT(mode & DS_MODE_READONLY);
- atime_changed_cb(zfsvfs, B_FALSE);
- readonly_changed_cb(zfsvfs, B_TRUE);
- if (error = dsl_prop_get_integer(osname, "xattr", &pval, NULL))
- goto out;
- xattr_changed_cb(zfsvfs, pval);
- zfsvfs->z_issnap = B_TRUE;
- } else {
- error = zfsvfs_setup(zfsvfs, B_TRUE);
+ mutex_enter(&zsb->z_lock);
+ if (*objp == 0) {
+ *objp = zap_create(zsb->z_os, DMU_OT_USERGROUP_QUOTA,
+ DMU_OT_NONE, 0, tx);
+ VERIFY(0 == zap_add(zsb->z_os, MASTER_NODE_OBJ,
+ zfs_userquota_prop_prefixes[type], 8, 1, objp, tx));
}
+ mutex_exit(&zsb->z_lock);
- if (!zfsvfs->z_issnap)
- zfsctl_create(zfsvfs);
-out:
- if (error) {
- if (zfsvfs->z_os)
- dmu_objset_close(zfsvfs->z_os);
- zfs_freezfsvfs(zfsvfs);
+ if (quota == 0) {
+ err = zap_remove(zsb->z_os, *objp, buf, tx);
+ if (err == ENOENT)
+ err = 0;
} else {
- atomic_add_32(&zfs_active_fs_count, 1);
+ err = zap_update(zsb->z_os, *objp, buf, 8, 1, "a, tx);
}
-
- return (error);
+ ASSERT(err == 0);
+ if (fuid_dirtied)
+ zfs_fuid_sync(zsb, tx);
+ dmu_tx_commit(tx);
+ return (err);
}
+EXPORT_SYMBOL(zfs_set_userquota);
-void
-zfs_unregister_callbacks(zfsvfs_t *zfsvfs)
+boolean_t
+zfs_fuid_overquota(zfs_sb_t *zsb, boolean_t isgroup, uint64_t fuid)
{
- objset_t *os = zfsvfs->z_os;
- struct dsl_dataset *ds;
+ char buf[32];
+ uint64_t used, quota, usedobj, quotaobj;
+ int err;
- /*
- * Unregister properties.
- */
- if (!dmu_objset_is_snapshot(os)) {
- ds = dmu_objset_ds(os);
- VERIFY(dsl_prop_unregister(ds, "atime", atime_changed_cb,
- zfsvfs) == 0);
+ usedobj = isgroup ? DMU_GROUPUSED_OBJECT : DMU_USERUSED_OBJECT;
+ quotaobj = isgroup ? zsb->z_groupquota_obj : zsb->z_userquota_obj;
- VERIFY(dsl_prop_unregister(ds, "xattr", xattr_changed_cb,
- zfsvfs) == 0);
+ if (quotaobj == 0 || zsb->z_replay)
+ return (B_FALSE);
- VERIFY(dsl_prop_unregister(ds, "recordsize", blksz_changed_cb,
- zfsvfs) == 0);
+ (void) sprintf(buf, "%llx", (longlong_t)fuid);
+ err = zap_lookup(zsb->z_os, quotaobj, buf, 8, 1, "a);
+ if (err != 0)
+ return (B_FALSE);
- VERIFY(dsl_prop_unregister(ds, "readonly", readonly_changed_cb,
- zfsvfs) == 0);
-
- VERIFY(dsl_prop_unregister(ds, "devices", devices_changed_cb,
- zfsvfs) == 0);
-
- VERIFY(dsl_prop_unregister(ds, "setuid", setuid_changed_cb,
- zfsvfs) == 0);
+ err = zap_lookup(zsb->z_os, usedobj, buf, 8, 1, &used);
+ if (err != 0)
+ return (B_FALSE);
+ return (used >= quota);
+}
+EXPORT_SYMBOL(zfs_fuid_overquota);
- VERIFY(dsl_prop_unregister(ds, "exec", exec_changed_cb,
- zfsvfs) == 0);
+boolean_t
+zfs_owner_overquota(zfs_sb_t *zsb, znode_t *zp, boolean_t isgroup)
+{
+ uint64_t fuid;
+ uint64_t quotaobj;
- VERIFY(dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb,
- zfsvfs) == 0);
+ quotaobj = isgroup ? zsb->z_groupquota_obj : zsb->z_userquota_obj;
- VERIFY(dsl_prop_unregister(ds, "aclmode", acl_mode_changed_cb,
- zfsvfs) == 0);
+ fuid = isgroup ? zp->z_gid : zp->z_uid;
- VERIFY(dsl_prop_unregister(ds, "aclinherit",
- acl_inherit_changed_cb, zfsvfs) == 0);
+ if (quotaobj == 0 || zsb->z_replay)
+ return (B_FALSE);
- VERIFY(dsl_prop_unregister(ds, "vscan",
- vscan_changed_cb, zfsvfs) == 0);
- }
+ return (zfs_fuid_overquota(zsb, isgroup, fuid));
}
+EXPORT_SYMBOL(zfs_owner_overquota);
-/*
- * Convert a decimal digit string to a uint64_t integer.
- */
-static int
-str_to_uint64(char *str, uint64_t *objnum)
+int
+zfs_sb_create(const char *osname, zfs_sb_t **zsbp)
{
- uint64_t num = 0;
+ objset_t *os;
+ zfs_sb_t *zsb;
+ uint64_t zval;
+ int i, error;
+ uint64_t sa_obj;
- while (*str) {
- if (*str < '0' || *str > '9')
- return (EINVAL);
+ zsb = kmem_zalloc(sizeof (zfs_sb_t), KM_SLEEP);
- num = num*10 + *str++ - '0';
+ /*
+ * We claim to always be readonly so we can open snapshots;
+ * other ZPL code will prevent us from writing to snapshots.
+ */
+ error = dmu_objset_own(osname, DMU_OST_ZFS, B_TRUE, zsb, &os);
+ if (error) {
+ kmem_free(zsb, sizeof (zfs_sb_t));
+ return (error);
}
- *objnum = num;
- return (0);
-}
+ /*
+ * Initialize the zfs-specific filesystem structure.
+ * Should probably make this a kmem cache, shuffle fields,
+ * and just bzero up to z_hold_mtx[].
+ */
+ zsb->z_sb = NULL;
+ zsb->z_parent = zsb;
+ zsb->z_max_blksz = SPA_MAXBLOCKSIZE;
+ zsb->z_show_ctldir = ZFS_SNAPDIR_VISIBLE;
+ zsb->z_os = os;
-/*
- * The boot path passed from the boot loader is in the form of
- * "rootpool-name/root-filesystem-object-number'. Convert this
- * string to a dataset name: "rootpool-name/root-filesystem-name".
- */
-static int
-zfs_parse_bootfs(char *bpath, char *outpath)
-{
- char *slashp;
- uint64_t objnum;
- int error;
+ error = -bdi_init(&zsb->z_bdi);
+ if (error) {
+ kmem_free(zsb, sizeof (zfs_sb_t));
+ return (error);
+ }
- if (*bpath == 0 || *bpath == '/')
- return (EINVAL);
+ error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zsb->z_version);
+ if (error) {
+ goto out;
+ } else if (zsb->z_version >
+ zfs_zpl_version_map(spa_version(dmu_objset_spa(os)))) {
+ (void) printk("Can't mount a version %lld file system "
+ "on a version %lld pool\n. Pool must be upgraded to mount "
+ "this file system.", (u_longlong_t)zsb->z_version,
+ (u_longlong_t)spa_version(dmu_objset_spa(os)));
+ error = ENOTSUP;
+ goto out;
+ }
+ if ((error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &zval)) != 0)
+ goto out;
+ zsb->z_norm = (int)zval;
- (void) strcpy(outpath, bpath);
+ if ((error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &zval)) != 0)
+ goto out;
+ zsb->z_utf8 = (zval != 0);
- slashp = strchr(bpath, '/');
+ if ((error = zfs_get_zplprop(os, ZFS_PROP_CASE, &zval)) != 0)
+ goto out;
+ zsb->z_case = (uint_t)zval;
- /* if no '/', just return the pool name */
- if (slashp == NULL) {
- return (0);
+ /*
+ * Fold case on file systems that are always or sometimes case
+ * insensitive.
+ */
+ if (zsb->z_case == ZFS_CASE_INSENSITIVE ||
+ zsb->z_case == ZFS_CASE_MIXED)
+ zsb->z_norm |= U8_TEXTPREP_TOUPPER;
+
+ zsb->z_use_fuids = USE_FUIDS(zsb->z_version, zsb->z_os);
+ zsb->z_use_sa = USE_SA(zsb->z_version, zsb->z_os);
+
+ if (zsb->z_use_sa) {
+ /* should either have both of these objects or none */
+ error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1,
+ &sa_obj);
+ if (error)
+ return (error);
+ } else {
+ /*
+ * Pre SA versions file systems should never touch
+ * either the attribute registration or layout objects.
+ */
+ sa_obj = 0;
}
- /* if not a number, just return the root dataset name */
- if (str_to_uint64(slashp+1, &objnum)) {
- return (0);
- }
+ error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END,
+ &zsb->z_attr_table);
+ if (error)
+ goto out;
+
+ if (zsb->z_version >= ZPL_VERSION_SA)
+ sa_register_update_callback(os, zfs_sa_upgrade);
- *slashp = '\0';
- error = dsl_dsobj_to_dsname(bpath, objnum, outpath);
- *slashp = '/';
+ error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1,
+ &zsb->z_root);
+ if (error)
+ goto out;
+ ASSERT(zsb->z_root != 0);
+
+ error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1,
+ &zsb->z_unlinkedobj);
+ if (error)
+ goto out;
+
+ error = zap_lookup(os, MASTER_NODE_OBJ,
+ zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA],
+ 8, 1, &zsb->z_userquota_obj);
+ if (error && error != ENOENT)
+ goto out;
+
+ error = zap_lookup(os, MASTER_NODE_OBJ,
+ zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA],
+ 8, 1, &zsb->z_groupquota_obj);
+ if (error && error != ENOENT)
+ goto out;
+ error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1,
+ &zsb->z_fuid_obj);
+ if (error && error != ENOENT)
+ goto out;
+
+ error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SHARES_DIR, 8, 1,
+ &zsb->z_shares_dir);
+ if (error && error != ENOENT)
+ goto out;
+
+ mutex_init(&zsb->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
+ mutex_init(&zsb->z_lock, NULL, MUTEX_DEFAULT, NULL);
+ list_create(&zsb->z_all_znodes, sizeof (znode_t),
+ offsetof(znode_t, z_link_node));
+ rrw_init(&zsb->z_teardown_lock);
+ rw_init(&zsb->z_teardown_inactive_lock, NULL, RW_DEFAULT, NULL);
+ rw_init(&zsb->z_fuid_lock, NULL, RW_DEFAULT, NULL);
+ for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
+ mutex_init(&zsb->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL);
+
+ *zsbp = zsb;
+ return (0);
+
+out:
+ dmu_objset_disown(os, zsb);
+ *zsbp = NULL;
+ kmem_free(zsb, sizeof (zfs_sb_t));
return (error);
}
+EXPORT_SYMBOL(zfs_sb_create);
-static int
-zfs_mountroot(vfs_t *vfsp, enum whymountroot why)
+int
+zfs_sb_setup(zfs_sb_t *zsb, boolean_t mounting)
{
- int error = 0;
- static int zfsrootdone = 0;
- zfsvfs_t *zfsvfs = NULL;
- znode_t *zp = NULL;
- vnode_t *vp = NULL;
- char *zfs_bootfs;
- char *zfs_devid;
+ int error;
- ASSERT(vfsp);
+ error = zfs_register_callbacks(zsb);
+ if (error)
+ return (error);
/*
- * The filesystem that we mount as root is defined in the
- * boot property "zfs-bootfs" with a format of
- * "poolname/root-dataset-objnum".
+ * Set the objset user_ptr to track its zsb.
*/
- if (why == ROOT_INIT) {
- if (zfsrootdone++)
- return (EBUSY);
- /*
- * the process of doing a spa_load will require the
- * clock to be set before we could (for example) do
- * something better by looking at the timestamp on
- * an uberblock, so just set it to -1.
- */
- clkset(-1);
+ mutex_enter(&zsb->z_os->os_user_ptr_lock);
+ dmu_objset_set_user(zsb->z_os, zsb);
+ mutex_exit(&zsb->z_os->os_user_ptr_lock);
- if ((zfs_bootfs = spa_get_bootprop("zfs-bootfs")) == NULL) {
- cmn_err(CE_NOTE, "spa_get_bootfs: can not get "
- "bootfs name");
- return (EINVAL);
- }
- zfs_devid = spa_get_bootprop("diskdevid");
- error = spa_import_rootpool(rootfs.bo_name, zfs_devid);
- if (zfs_devid)
- spa_free_bootprop(zfs_devid);
- if (error) {
- spa_free_bootprop(zfs_bootfs);
- cmn_err(CE_NOTE, "spa_import_rootpool: error %d",
- error);
- return (error);
- }
- if (error = zfs_parse_bootfs(zfs_bootfs, rootfs.bo_name)) {
- spa_free_bootprop(zfs_bootfs);
- cmn_err(CE_NOTE, "zfs_parse_bootfs: error %d",
- error);
- return (error);
- }
-
- spa_free_bootprop(zfs_bootfs);
-
- if (error = vfs_lock(vfsp))
- return (error);
-
- if (error = zfs_domount(vfsp, rootfs.bo_name)) {
- cmn_err(CE_NOTE, "zfs_domount: error %d", error);
- goto out;
- }
+ zsb->z_log = zil_open(zsb->z_os, zfs_get_data);
- zfsvfs = (zfsvfs_t *)vfsp->vfs_data;
- ASSERT(zfsvfs);
- if (error = zfs_zget(zfsvfs, zfsvfs->z_root, &zp)) {
- cmn_err(CE_NOTE, "zfs_zget: error %d", error);
- goto out;
- }
+ /*
+ * If we are not mounting (ie: online recv), then we don't
+ * have to worry about replaying the log as we blocked all
+ * operations out since we closed the ZIL.
+ */
+ if (mounting) {
+ boolean_t readonly;
- vp = ZTOV(zp);
- mutex_enter(&vp->v_lock);
- vp->v_flag |= VROOT;
- mutex_exit(&vp->v_lock);
- rootvp = vp;
+ /*
+ * During replay we remove the read only flag to
+ * allow replays to succeed.
+ */
+ readonly = zfs_is_readonly(zsb);
+ if (readonly != 0)
+ readonly_changed_cb(zsb, B_FALSE);
+ else
+ zfs_unlinked_drain(zsb);
/*
- * Leave rootvp held. The root file system is never unmounted.
+ * Parse and replay the intent log.
+ *
+ * Because of ziltest, this must be done after
+ * zfs_unlinked_drain(). (Further note: ziltest
+ * doesn't use readonly mounts, where
+ * zfs_unlinked_drain() isn't called.) This is because
+ * ziltest causes spa_sync() to think it's committed,
+ * but actually it is not, so the intent log contains
+ * many txg's worth of changes.
+ *
+ * In particular, if object N is in the unlinked set in
+ * the last txg to actually sync, then it could be
+ * actually freed in a later txg and then reallocated
+ * in a yet later txg. This would write a "create
+ * object N" record to the intent log. Normally, this
+ * would be fine because the spa_sync() would have
+ * written out the fact that object N is free, before
+ * we could write the "create object N" intent log
+ * record.
+ *
+ * But when we are in ziltest mode, we advance the "open
+ * txg" without actually spa_sync()-ing the changes to
+ * disk. So we would see that object N is still
+ * allocated and in the unlinked set, and there is an
+ * intent log record saying to allocate it.
*/
+ if (spa_writeable(dmu_objset_spa(zsb->z_os))) {
+ if (zil_replay_disable) {
+ zil_destroy(zsb->z_log, B_FALSE);
+ } else {
+ zsb->z_replay = B_TRUE;
+ zil_replay(zsb->z_os, zsb,
+ zfs_replay_vector);
+ zsb->z_replay = B_FALSE;
+ }
+ }
- vfs_add((struct vnode *)0, vfsp,
- (vfsp->vfs_flag & VFS_RDONLY) ? MS_RDONLY : 0);
-out:
- vfs_unlock(vfsp);
- return (error);
- } else if (why == ROOT_REMOUNT) {
- readonly_changed_cb(vfsp->vfs_data, B_FALSE);
- vfsp->vfs_flag |= VFS_REMOUNT;
+ /* restore readonly bit */
+ if (readonly != 0)
+ readonly_changed_cb(zsb, B_TRUE);
+ }
- /* refresh mount options */
- zfs_unregister_callbacks(vfsp->vfs_data);
- return (zfs_register_callbacks(vfsp));
+ return (0);
+}
+EXPORT_SYMBOL(zfs_sb_setup);
- } else if (why == ROOT_UNMOUNT) {
- zfs_unregister_callbacks((zfsvfs_t *)vfsp->vfs_data);
- (void) zfs_sync(vfsp, 0, 0);
- return (0);
- }
+void
+zfs_sb_free(zfs_sb_t *zsb)
+{
+ int i;
- /*
- * if "why" is equal to anything else other than ROOT_INIT,
- * ROOT_REMOUNT, or ROOT_UNMOUNT, we do not support it.
- */
- return (ENOTSUP);
+ zfs_fuid_destroy(zsb);
+
+ bdi_destroy(&zsb->z_bdi);
+ mutex_destroy(&zsb->z_znodes_lock);
+ mutex_destroy(&zsb->z_lock);
+ list_destroy(&zsb->z_all_znodes);
+ rrw_destroy(&zsb->z_teardown_lock);
+ rw_destroy(&zsb->z_teardown_inactive_lock);
+ rw_destroy(&zsb->z_fuid_lock);
+ for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
+ mutex_destroy(&zsb->z_hold_mtx[i]);
+ kmem_free(zsb, sizeof (zfs_sb_t));
}
+EXPORT_SYMBOL(zfs_sb_free);
-/*ARGSUSED*/
-static int
-zfs_mount(vfs_t *vfsp, vnode_t *mvp, struct mounta *uap, cred_t *cr)
+static void
+zfs_set_fuid_feature(zfs_sb_t *zsb)
{
- char *osname;
- pathname_t spn;
- int error = 0;
- uio_seg_t fromspace = (uap->flags & MS_SYSSPACE) ?
- UIO_SYSSPACE : UIO_USERSPACE;
- int canwrite;
-
- if (mvp->v_type != VDIR)
- return (ENOTDIR);
-
- mutex_enter(&mvp->v_lock);
- if ((uap->flags & MS_REMOUNT) == 0 &&
- (uap->flags & MS_OVERLAY) == 0 &&
- (mvp->v_count != 1 || (mvp->v_flag & VROOT))) {
- mutex_exit(&mvp->v_lock);
- return (EBUSY);
- }
- mutex_exit(&mvp->v_lock);
+ zsb->z_use_fuids = USE_FUIDS(zsb->z_version, zsb->z_os);
+ zsb->z_use_sa = USE_SA(zsb->z_version, zsb->z_os);
+}
- /*
- * ZFS does not support passing unparsed data in via MS_DATA.
- * Users should use the MS_OPTIONSTR interface; this means
- * that all option parsing is already done and the options struct
- * can be interrogated.
- */
- if ((uap->flags & MS_DATA) && uap->datalen > 0)
- return (EINVAL);
+void
+zfs_unregister_callbacks(zfs_sb_t *zsb)
+{
+ objset_t *os = zsb->z_os;
+ struct dsl_dataset *ds;
/*
- * Get the objset name (the "special" mount argument).
+ * Unregister properties.
*/
- if (error = pn_get(uap->spec, fromspace, &spn))
- return (error);
+ if (!dmu_objset_is_snapshot(os)) {
+ ds = dmu_objset_ds(os);
+ VERIFY(dsl_prop_unregister(ds, "atime", atime_changed_cb,
+ zsb) == 0);
- osname = spn.pn_path;
+ VERIFY(dsl_prop_unregister(ds, "xattr", xattr_changed_cb,
+ zsb) == 0);
- /*
- * Check for mount privilege?
- *
- * If we don't have privilege then see if
- * we have local permission to allow it
- */
- error = secpolicy_fs_mount(cr, mvp, vfsp);
- if (error) {
- error = dsl_deleg_access(osname, ZFS_DELEG_PERM_MOUNT, cr);
- if (error == 0) {
- vattr_t vattr;
+ VERIFY(dsl_prop_unregister(ds, "recordsize", blksz_changed_cb,
+ zsb) == 0);
- /*
- * Make sure user is the owner of the mount point
- * or has sufficient privileges.
- */
+ VERIFY(dsl_prop_unregister(ds, "readonly", readonly_changed_cb,
+ zsb) == 0);
- vattr.va_mask = AT_UID;
+ VERIFY(dsl_prop_unregister(ds, "devices", devices_changed_cb,
+ zsb) == 0);
- if (error = VOP_GETATTR(mvp, &vattr, 0, cr, NULL)) {
- goto out;
- }
+ VERIFY(dsl_prop_unregister(ds, "setuid", setuid_changed_cb,
+ zsb) == 0);
- if (secpolicy_vnode_owner(cr, vattr.va_uid) != 0 &&
- VOP_ACCESS(mvp, VWRITE, 0, cr, NULL) != 0) {
- error = EPERM;
- goto out;
- }
+ VERIFY(dsl_prop_unregister(ds, "exec", exec_changed_cb,
+ zsb) == 0);
- secpolicy_fs_mount_clearopts(cr, vfsp);
- } else {
- goto out;
- }
- }
+ VERIFY(dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb,
+ zsb) == 0);
- /*
- * Refuse to mount a filesystem if we are in a local zone and the
- * dataset is not visible.
- */
- if (!INGLOBALZONE(curproc) &&
- (!zone_dataset_visible(osname, &canwrite) || !canwrite)) {
- error = EPERM;
- goto out;
- }
+ VERIFY(dsl_prop_unregister(ds, "aclinherit",
+ acl_inherit_changed_cb, zsb) == 0);
- /*
- * When doing a remount, we simply refresh our temporary properties
- * according to those options set in the current VFS options.
- */
- if (uap->flags & MS_REMOUNT) {
- /* refresh mount options */
- zfs_unregister_callbacks(vfsp->vfs_data);
- error = zfs_register_callbacks(vfsp);
- goto out;
- }
+ VERIFY(dsl_prop_unregister(ds, "vscan",
+ vscan_changed_cb, zsb) == 0);
- error = zfs_domount(vfsp, osname);
+ VERIFY(dsl_prop_unregister(ds, "nbmand",
+ nbmand_changed_cb, zsb) == 0);
+ }
+}
+EXPORT_SYMBOL(zfs_unregister_callbacks);
-out:
- pn_free(&spn);
- return (error);
+#ifdef HAVE_MLSLABEL
+/*
+ * zfs_check_global_label:
+ * Check that the hex label string is appropriate for the dataset
+ * being mounted into the global_zone proper.
+ *
+ * Return an error if the hex label string is not default or
+ * admin_low/admin_high. For admin_low labels, the corresponding
+ * dataset must be readonly.
+ */
+int
+zfs_check_global_label(const char *dsname, const char *hexsl)
+{
+ if (strcasecmp(hexsl, ZFS_MLSLABEL_DEFAULT) == 0)
+ return (0);
+ if (strcasecmp(hexsl, ADMIN_HIGH) == 0)
+ return (0);
+ if (strcasecmp(hexsl, ADMIN_LOW) == 0) {
+ /* must be readonly */
+ uint64_t rdonly;
+
+ if (dsl_prop_get_integer(dsname,
+ zfs_prop_to_name(ZFS_PROP_READONLY), &rdonly, NULL))
+ return (EACCES);
+ return (rdonly ? 0 : EACCES);
+ }
+ return (EACCES);
}
+EXPORT_SYMBOL(zfs_check_global_label);
+#endif /* HAVE_MLSLABEL */
-static int
-zfs_statvfs(vfs_t *vfsp, struct statvfs64 *statp)
+int
+zfs_statvfs(struct dentry *dentry, struct kstatfs *statp)
{
- zfsvfs_t *zfsvfs = vfsp->vfs_data;
- dev32_t d32;
+ zfs_sb_t *zsb = dentry->d_sb->s_fs_info;
uint64_t refdbytes, availbytes, usedobjs, availobjs;
+ uint32_t bshift;
- ZFS_ENTER(zfsvfs);
+ ZFS_ENTER(zsb);
- dmu_objset_space(zfsvfs->z_os,
+ dmu_objset_space(zsb->z_os,
&refdbytes, &availbytes, &usedobjs, &availobjs);
/*
- * The underlying storage pool actually uses multiple block sizes.
- * We report the fragsize as the smallest block size we support,
- * and we report our blocksize as the filesystem's maximum blocksize.
+ * 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 = 1UL << SPA_MINBLOCKSHIFT;
- statp->f_bsize = zfsvfs->z_max_blksz;
+ 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_frsize - the
- * "fragment" size.
+ * 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) >> SPA_MINBLOCKSHIFT;
- statp->f_bfree = availbytes >> SPA_MINBLOCKSHIFT;
+ statp->f_blocks = (refdbytes + availbytes) >> bshift;
+ statp->f_bfree = availbytes >> bshift;
statp->f_bavail = statp->f_bfree; /* no root reservation */
/*
* 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, statp->f_bfree);
- statp->f_favail = statp->f_ffree; /* no "root reservation" */
+ statp->f_ffree = MIN(availobjs, availbytes >> DNODE_SHIFT);
statp->f_files = statp->f_ffree + usedobjs;
-
- (void) cmpldev(&d32, vfsp->vfs_dev);
- statp->f_fsid = d32;
+ statp->f_fsid.val[0] = dentry->d_sb->s_dev;
+ statp->f_fsid.val[1] = 0;
+ statp->f_type = ZFS_SUPER_MAGIC;
+ statp->f_namelen = ZFS_MAXNAMELEN;
/*
- * We're a zfs filesystem.
- */
- (void) strcpy(statp->f_basetype, vfssw[vfsp->vfs_fstype].vsw_name);
-
- statp->f_flag = vf_to_stf(vfsp->vfs_flag);
-
- statp->f_namemax = ZFS_MAXNAMELEN;
-
- /*
- * We have all of 32 characters to stuff a string here.
+ * We have all of 40 characters to stuff a string here.
* Is there anything useful we could/should provide?
*/
- bzero(statp->f_fstr, sizeof (statp->f_fstr));
+ bzero(statp->f_spare, sizeof (statp->f_spare));
- ZFS_EXIT(zfsvfs);
+ ZFS_EXIT(zsb);
return (0);
}
+EXPORT_SYMBOL(zfs_statvfs);
-static int
-zfs_root(vfs_t *vfsp, vnode_t **vpp)
+int
+zfs_root(zfs_sb_t *zsb, struct inode **ipp)
{
- zfsvfs_t *zfsvfs = vfsp->vfs_data;
znode_t *rootzp;
int error;
- ZFS_ENTER(zfsvfs);
+ ZFS_ENTER(zsb);
- error = zfs_zget(zfsvfs, zfsvfs->z_root, &rootzp);
+ error = zfs_zget(zsb, zsb->z_root, &rootzp);
if (error == 0)
- *vpp = ZTOV(rootzp);
+ *ipp = ZTOI(rootzp);
- ZFS_EXIT(zfsvfs);
+ ZFS_EXIT(zsb);
return (error);
}
+EXPORT_SYMBOL(zfs_root);
/*
- * Teardown the zfsvfs::z_os.
+ * Teardown the zfs_sb_t::z_os.
*
* Note, if 'unmounting' if FALSE, we return with the 'z_teardown_lock'
* and 'z_teardown_inactive_lock' held.
*/
-static int
-zfsvfs_teardown(zfsvfs_t *zfsvfs, boolean_t unmounting)
+int
+zfs_sb_teardown(zfs_sb_t *zsb, boolean_t unmounting)
{
znode_t *zp;
- rrw_enter(&zfsvfs->z_teardown_lock, RW_WRITER, FTAG);
+ rrw_enter(&zsb->z_teardown_lock, RW_WRITER, FTAG);
if (!unmounting) {
/*
- * 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.
+ * We purge the parent filesystem's super block as the
+ * parent filesystem and all of its snapshots have their
+ * inode's super block set to the parent's filesystem's
+ * super block. Note, 'z_parent' is self referential
+ * for non-snapshots.
*/
- (void) dnlc_purge_vfsp(zfsvfs->z_parent->z_vfs, 0);
+ shrink_dcache_sb(zsb->z_parent->z_sb);
+ (void) spl_invalidate_inodes(zsb->z_parent->z_sb, 0);
}
/*
+ * Drain the iput_taskq to ensure all active references to the
+ * zfs_sb_t have been handled only then can it be safely destroyed.
+ */
+ taskq_wait(dsl_pool_iput_taskq(dmu_objset_pool(zsb->z_os)));
+
+ /*
* 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;
+ if (zsb->z_log) {
+ zil_close(zsb->z_log);
+ zsb->z_log = NULL;
}
- rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_WRITER);
+ rw_enter(&zsb->z_teardown_inactive_lock, RW_WRITER);
/*
* If we are not unmounting (ie: online recv) and someone already
* unmounted this file system while we were doing the switcheroo,
* or a reopen of z_os failed then just bail out now.
*/
- if (!unmounting && (zfsvfs->z_unmounted || zfsvfs->z_os == NULL)) {
- rw_exit(&zfsvfs->z_teardown_inactive_lock);
- rrw_exit(&zfsvfs->z_teardown_lock, FTAG);
+ if (!unmounting && (zsb->z_unmounted || zsb->z_os == NULL)) {
+ rw_exit(&zsb->z_teardown_inactive_lock);
+ rrw_exit(&zsb->z_teardown_lock, FTAG);
return (EIO);
}
*
* Release all holds on dbufs.
*/
- mutex_enter(&zfsvfs->z_znodes_lock);
- for (zp = list_head(&zfsvfs->z_all_znodes); zp != NULL;
- zp = list_next(&zfsvfs->z_all_znodes, zp))
- if (zp->z_dbuf) {
- ASSERT(ZTOV(zp)->v_count > 0);
+ mutex_enter(&zsb->z_znodes_lock);
+ for (zp = list_head(&zsb->z_all_znodes); zp != NULL;
+ zp = list_next(&zsb->z_all_znodes, zp))
+ if (zp->z_sa_hdl) {
+ ASSERT(atomic_read(&ZTOI(zp)->i_count) > 0);
zfs_znode_dmu_fini(zp);
}
- mutex_exit(&zfsvfs->z_znodes_lock);
+ mutex_exit(&zsb->z_znodes_lock);
/*
* If we are unmounting, set the unmounted flag and let new vops
* other vops will fail with EIO.
*/
if (unmounting) {
- zfsvfs->z_unmounted = B_TRUE;
- rrw_exit(&zfsvfs->z_teardown_lock, FTAG);
- rw_exit(&zfsvfs->z_teardown_inactive_lock);
+ zsb->z_unmounted = B_TRUE;
+ rrw_exit(&zsb->z_teardown_lock, FTAG);
+ rw_exit(&zsb->z_teardown_inactive_lock);
}
/*
* z_os will be NULL if there was an error in attempting to reopen
- * zfsvfs, so just return as the properties had already been
+ * zsb, so just return as the properties had already been
+ *
* unregistered and cached data had been evicted before.
*/
- if (zfsvfs->z_os == NULL)
+ if (zsb->z_os == NULL)
return (0);
/*
* Unregister properties.
*/
- zfs_unregister_callbacks(zfsvfs);
+ zfs_unregister_callbacks(zsb);
/*
* Evict cached data
*/
- if (dmu_objset_evict_dbufs(zfsvfs->z_os)) {
- txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), 0);
- (void) dmu_objset_evict_dbufs(zfsvfs->z_os);
- }
+ if (dmu_objset_is_dirty_anywhere(zsb->z_os))
+ if (!zfs_is_readonly(zsb))
+ txg_wait_synced(dmu_objset_pool(zsb->z_os), 0);
+ (void) dmu_objset_evict_dbufs(zsb->z_os);
return (0);
}
+EXPORT_SYMBOL(zfs_sb_teardown);
-/*ARGSUSED*/
-static int
-zfs_umount(vfs_t *vfsp, int fflag, cred_t *cr)
+#ifdef HAVE_BDI
+static atomic_long_t bdi_seq = ATOMIC_LONG_INIT(0);
+#endif /* HAVE_BDI */
+
+int
+zfs_domount(struct super_block *sb, void *data, int silent)
{
- zfsvfs_t *zfsvfs = vfsp->vfs_data;
- objset_t *os;
- int ret;
-
- ret = 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);
- }
+ zpl_mount_data_t *zmd = data;
+ const char *osname = zmd->z_osname;
+ zfs_sb_t *zsb;
+ struct inode *root_inode;
+ uint64_t recordsize;
+ int error;
- /*
- * 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);
+ error = zfs_sb_create(osname, &zsb);
+ if (error)
+ return (error);
- /*
- * Unmount any snapshots mounted under .zfs before unmounting the
- * dataset itself.
- */
- if (zfsvfs->z_ctldir != NULL &&
- (ret = zfsctl_umount_snapshots(vfsp, fflag, cr)) != 0) {
- return (ret);
+ if ((error = dsl_prop_get_integer(osname, "recordsize",
+ &recordsize, NULL)))
+ goto out;
+
+ zsb->z_sb = sb;
+ sb->s_fs_info = zsb;
+ sb->s_magic = ZFS_SUPER_MAGIC;
+ sb->s_maxbytes = MAX_LFS_FILESIZE;
+ sb->s_time_gran = 1;
+ sb->s_blocksize = recordsize;
+ sb->s_blocksize_bits = ilog2(recordsize);
+ bdi_put_sb(sb, NULL);
+
+ /* Set callback operations for the file system. */
+ sb->s_op = &zpl_super_operations;
+ sb->s_xattr = zpl_xattr_handlers;
+ sb->s_export_op = &zpl_export_operations;
+
+ /* Set features for file system. */
+ zfs_set_fuid_feature(zsb);
+
+ if (dmu_objset_is_snapshot(zsb->z_os)) {
+ uint64_t pval;
+
+ atime_changed_cb(zsb, B_FALSE);
+ readonly_changed_cb(zsb, B_TRUE);
+ if ((error = dsl_prop_get_integer(osname,"xattr",&pval,NULL)))
+ goto out;
+ xattr_changed_cb(zsb, pval);
+ zsb->z_issnap = B_TRUE;
+ zsb->z_os->os_sync = ZFS_SYNC_DISABLED;
+
+ mutex_enter(&zsb->z_os->os_user_ptr_lock);
+ dmu_objset_set_user(zsb->z_os, zsb);
+ mutex_exit(&zsb->z_os->os_user_ptr_lock);
+ } else {
+ /* Disable Linux read-ahead handled by lower layers */
+ zsb->z_bdi.ra_pages = 0;
+
+ error = -bdi_register(&zsb->z_bdi, NULL, "zfs-%d",
+ atomic_long_inc_return(&bdi_seq));
+ if (error)
+ goto out;
+
+ bdi_put_sb(sb, &zsb->z_bdi);
+
+ error = zfs_sb_setup(zsb, B_TRUE);
+#ifdef HAVE_SNAPSHOT
+ (void) zfs_snap_create(zsb);
+#endif /* HAVE_SNAPSHOT */
}
- if (!(fflag & MS_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 (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);
- }
+ /* Allocate a root inode for the filesystem. */
+ error = zfs_root(zsb, &root_inode);
+ if (error) {
+ (void) zfs_umount(sb);
+ goto out;
}
- vfsp->vfs_flag |= VFS_UNMOUNTED;
+ /* Allocate a root dentry for the filesystem */
+ sb->s_root = d_alloc_root(root_inode);
+ if (sb->s_root == NULL) {
+ (void) zfs_umount(sb);
+ error = ENOMEM;
+ goto out;
+ }
+out:
+ if (error) {
+ dmu_objset_disown(zsb->z_os, zsb);
+ zfs_sb_free(zsb);
+ }
+
+ return (error);
+}
+EXPORT_SYMBOL(zfs_domount);
+
+/*ARGSUSED*/
+int
+zfs_umount(struct super_block *sb)
+{
+ zfs_sb_t *zsb = sb->s_fs_info;
+ objset_t *os;
+
+ VERIFY(zfs_sb_teardown(zsb, B_TRUE) == 0);
+ os = zsb->z_os;
- VERIFY(zfsvfs_teardown(zfsvfs, B_TRUE) == 0);
- os = zfsvfs->z_os;
+ if (bdi_get_sb(sb)) {
+ bdi_unregister(bdi_get_sb(sb));
+ bdi_put_sb(sb, NULL);
+ }
/*
* z_os will be NULL if there was an error in
- * attempting to reopen zfsvfs.
+ * attempting to reopen zsb.
*/
if (os != NULL) {
/*
* Unset the objset user_ptr.
*/
- mutex_enter(&os->os->os_user_ptr_lock);
+ mutex_enter(&os->os_user_ptr_lock);
dmu_objset_set_user(os, NULL);
- mutex_exit(&os->os->os_user_ptr_lock);
+ mutex_exit(&os->os_user_ptr_lock);
/*
* Finally release the objset
*/
- dmu_objset_close(os);
+ dmu_objset_disown(os, zsb);
}
+ zfs_sb_free(zsb);
+ return (0);
+}
+EXPORT_SYMBOL(zfs_umount);
+
+int
+zfs_remount(struct super_block *sb, int *flags, char *data)
+{
/*
- * We can now safely destroy the '.zfs' directory node.
+ * All namespace flags (MNT_*) and super block flags (MS_*) will
+ * be handled by the Linux VFS. Only handle custom options here.
*/
- if (zfsvfs->z_ctldir != NULL)
- zfsctl_destroy(zfsvfs);
-
return (0);
}
+EXPORT_SYMBOL(zfs_remount);
-static int
-zfs_vget(vfs_t *vfsp, vnode_t **vpp, fid_t *fidp)
+int
+zfs_vget(struct super_block *sb, struct inode **ipp, fid_t *fidp)
{
- zfsvfs_t *zfsvfs = vfsp->vfs_data;
+ 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;
+ int i, err;
- *vpp = NULL;
+ *ipp = NULL;
- ZFS_ENTER(zfsvfs);
+ ZFS_ENTER(zsb);
if (fidp->fid_len == LONG_FID_LEN) {
zfid_long_t *zlfid = (zfid_long_t *)fidp;
for (i = 0; i < sizeof (zlfid->zf_setgen); i++)
setgen |= ((uint64_t)zlfid->zf_setgen[i]) << (8 * i);
- ZFS_EXIT(zfsvfs);
+ ZFS_EXIT(zsb);
- err = zfsctl_lookup_objset(vfsp, objsetid, &zfsvfs);
+#ifdef HAVE_SNAPSHOT
+ err = zfsctl_lookup_objset(vfsp, objsetid, &zsb);
if (err)
return (EINVAL);
- ZFS_ENTER(zfsvfs);
+#endif /* HAVE_SNAPSHOT */
+ ZFS_ENTER(zsb);
}
if (fidp->fid_len == SHORT_FID_LEN || fidp->fid_len == LONG_FID_LEN) {
for (i = 0; i < sizeof (zfid->zf_gen); i++)
fid_gen |= ((uint64_t)zfid->zf_gen[i]) << (8 * i);
} else {
- ZFS_EXIT(zfsvfs);
+ ZFS_EXIT(zsb);
return (EINVAL);
}
+#ifdef HAVE_SNAPSHOT
/* A zero fid_gen means we are in the .zfs control directories */
if (fid_gen == 0 &&
(object == ZFSCTL_INO_ROOT || object == ZFSCTL_INO_SNAPDIR)) {
- *vpp = zfsvfs->z_ctldir;
- ASSERT(*vpp != NULL);
+ *ipp = zsb->z_ctldir;
+ ASSERT(*ipp != NULL);
if (object == ZFSCTL_INO_SNAPDIR) {
- VERIFY(zfsctl_root_lookup(*vpp, "snapshot", vpp, NULL,
+ VERIFY(zfsctl_root_lookup(*ipp, "snapshot", ipp, NULL,
0, NULL, NULL, NULL, NULL, NULL) == 0);
} else {
- VN_HOLD(*vpp);
+ igrab(*ipp);
}
- ZFS_EXIT(zfsvfs);
+ ZFS_EXIT(zsb);
return (0);
}
+#endif /* HAVE_SNAPSHOT */
gen_mask = -1ULL >> (64 - 8 * i);
dprintf("getting %llu [%u mask %llx]\n", object, fid_gen, gen_mask);
- if (err = zfs_zget(zfsvfs, object, &zp)) {
- ZFS_EXIT(zfsvfs);
+ if ((err = zfs_zget(zsb, object, &zp))) {
+ ZFS_EXIT(zsb);
return (err);
}
- zp_gen = zp->z_phys->zp_gen & gen_mask;
+ (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 (%u) != fid gen (%u)\n", zp_gen, fid_gen);
- VN_RELE(ZTOV(zp));
- ZFS_EXIT(zfsvfs);
+ iput(ZTOI(zp));
+ ZFS_EXIT(zsb);
return (EINVAL);
}
- *vpp = ZTOV(zp);
- ZFS_EXIT(zfsvfs);
+ *ipp = ZTOI(zp);
+ if (*ipp)
+ zfs_inode_update(ITOZ(*ipp));
+
+ ZFS_EXIT(zsb);
return (0);
}
+EXPORT_SYMBOL(zfs_vget);
/*
- * Block out VOPs and close zfsvfs_t::z_os
+ * Block out VOPs and close zfs_sb_t::z_os
*
* Note, if successful, then we return with the 'z_teardown_lock' and
* 'z_teardown_inactive_lock' write held.
*/
int
-zfs_suspend_fs(zfsvfs_t *zfsvfs, char *name, int *mode)
+zfs_suspend_fs(zfs_sb_t *zsb)
{
int error;
- if ((error = zfsvfs_teardown(zfsvfs, B_FALSE)) != 0)
+ if ((error = zfs_sb_teardown(zsb, B_FALSE)) != 0)
return (error);
-
- *mode = zfsvfs->z_os->os_mode;
- dmu_objset_name(zfsvfs->z_os, name);
- dmu_objset_close(zfsvfs->z_os);
+ dmu_objset_disown(zsb->z_os, zsb);
return (0);
}
+EXPORT_SYMBOL(zfs_suspend_fs);
/*
- * Reopen zfsvfs_t::z_os and release VOPs.
+ * Reopen zfs_sb_t::z_os and release VOPs.
*/
int
-zfs_resume_fs(zfsvfs_t *zfsvfs, const char *osname, int mode)
+zfs_resume_fs(zfs_sb_t *zsb, const char *osname)
{
- int err;
+ int err, err2;
- ASSERT(RRW_WRITE_HELD(&zfsvfs->z_teardown_lock));
- ASSERT(RW_WRITE_HELD(&zfsvfs->z_teardown_inactive_lock));
+ ASSERT(RRW_WRITE_HELD(&zsb->z_teardown_lock));
+ ASSERT(RW_WRITE_HELD(&zsb->z_teardown_inactive_lock));
- err = dmu_objset_open(osname, DMU_OST_ZFS, mode, &zfsvfs->z_os);
+ err = dmu_objset_own(osname, DMU_OST_ZFS, B_FALSE, zsb, &zsb->z_os);
if (err) {
- zfsvfs->z_os = NULL;
+ zsb->z_os = NULL;
} else {
znode_t *zp;
+ uint64_t sa_obj = 0;
+
+ err2 = zap_lookup(zsb->z_os, MASTER_NODE_OBJ,
+ ZFS_SA_ATTRS, 8, 1, &sa_obj);
+
+ if ((err || err2) && zsb->z_version >= ZPL_VERSION_SA)
+ goto bail;
- VERIFY(zfsvfs_setup(zfsvfs, B_FALSE) == 0);
+
+ if ((err = sa_setup(zsb->z_os, sa_obj,
+ zfs_attr_table, ZPL_END, &zsb->z_attr_table)) != 0)
+ goto bail;
+
+ VERIFY(zfs_sb_setup(zsb, B_FALSE) == 0);
/*
* Attempt to re-establish all the active znodes with
* any potential callers discover that via ZFS_ENTER_VERIFY_VP
* when they try to use their znode.
*/
- mutex_enter(&zfsvfs->z_znodes_lock);
- for (zp = list_head(&zfsvfs->z_all_znodes); zp;
- zp = list_next(&zfsvfs->z_all_znodes, zp)) {
+ mutex_enter(&zsb->z_znodes_lock);
+ for (zp = list_head(&zsb->z_all_znodes); zp;
+ zp = list_next(&zsb->z_all_znodes, zp)) {
(void) zfs_rezget(zp);
}
- mutex_exit(&zfsvfs->z_znodes_lock);
+ mutex_exit(&zsb->z_znodes_lock);
}
+bail:
/* release the VOPs */
- rw_exit(&zfsvfs->z_teardown_inactive_lock);
- rrw_exit(&zfsvfs->z_teardown_lock, FTAG);
+ rw_exit(&zsb->z_teardown_inactive_lock);
+ rrw_exit(&zsb->z_teardown_lock, FTAG);
if (err) {
/*
- * Since we couldn't reopen zfsvfs::z_os, force
+ * Since we couldn't reopen zfs_sb_t::z_os, force
* unmount this file system.
*/
- if (vn_vfswlock(zfsvfs->z_vfs->vfs_vnodecovered) == 0)
- (void) dounmount(zfsvfs->z_vfs, MS_FORCE, CRED());
+ (void) zfs_umount(zsb->z_sb);
}
return (err);
}
+EXPORT_SYMBOL(zfs_resume_fs);
-static void
-zfs_freevfs(vfs_t *vfsp)
+int
+zfs_set_version(zfs_sb_t *zsb, uint64_t newvers)
{
- zfsvfs_t *zfsvfs = vfsp->vfs_data;
- int i;
-
- for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
- mutex_destroy(&zfsvfs->z_hold_mtx[i]);
-
- zfs_fuid_destroy(zfsvfs);
- zfs_freezfsvfs(zfsvfs);
+ int error;
+ objset_t *os = zsb->z_os;
+ dmu_tx_t *tx;
- atomic_add_32(&zfs_active_fs_count, -1);
-}
+ if (newvers < ZPL_VERSION_INITIAL || newvers > ZPL_VERSION)
+ return (EINVAL);
-/*
- * VFS_INIT() initialization. Note that there is no VFS_FINI(),
- * so we can't safely do any non-idempotent initialization here.
- * Leave that to zfs_init() and zfs_fini(), which are called
- * from the module's _init() and _fini() entry points.
- */
-/*ARGSUSED*/
-static int
-zfs_vfsinit(int fstype, char *name)
-{
- int error;
+ if (newvers < zsb->z_version)
+ return (EINVAL);
- zfsfstype = fstype;
+ if (zfs_spa_version_map(newvers) >
+ spa_version(dmu_objset_spa(zsb->z_os)))
+ return (ENOTSUP);
- /*
- * Setup vfsops and vnodeops tables.
- */
- error = vfs_setfsops(fstype, zfs_vfsops_template, &zfs_vfsops);
- if (error != 0) {
- cmn_err(CE_WARN, "zfs: bad vfs ops template");
+ tx = dmu_tx_create(os);
+ dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_FALSE, ZPL_VERSION_STR);
+ if (newvers >= ZPL_VERSION_SA && !zsb->z_use_sa) {
+ dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE,
+ ZFS_SA_ATTRS);
+ dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
}
-
- error = zfs_create_op_tables();
+ error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
- zfs_remove_op_tables();
- cmn_err(CE_WARN, "zfs: bad vnode ops template");
- (void) vfs_freevfsops_by_type(zfsfstype);
+ dmu_tx_abort(tx);
return (error);
}
- mutex_init(&zfs_dev_mtx, NULL, MUTEX_DEFAULT, NULL);
-
- /*
- * Unique major number for all zfs mounts.
- * If we run out of 32-bit minors, we'll getudev() another major.
- */
- zfs_major = ddi_name_to_major(ZFS_DRIVER);
- zfs_minor = ZFS_MIN_MINOR;
-
- return (0);
-}
-
-void
-zfs_init(void)
-{
- /*
- * Initialize .zfs directory structures
- */
- zfsctl_init();
-
- /*
- * Initialize znode cache, vnode ops, etc...
- */
- zfs_znode_init();
-}
-
-void
-zfs_fini(void)
-{
- zfsctl_fini();
- zfs_znode_fini();
-}
-
-int
-zfs_busy(void)
-{
- return (zfs_active_fs_count != 0);
-}
+ error = zap_update(os, MASTER_NODE_OBJ, ZPL_VERSION_STR,
+ 8, 1, &newvers, tx);
-int
-zfs_set_version(const char *name, uint64_t newvers)
-{
- int error;
- objset_t *os;
- dmu_tx_t *tx;
- uint64_t curvers;
+ if (error) {
+ dmu_tx_commit(tx);
+ return (error);
+ }
- /*
- * XXX for now, require that the filesystem be unmounted. Would
- * be nice to find the zfsvfs_t and just update that if
- * possible.
- */
+ if (newvers >= ZPL_VERSION_SA && !zsb->z_use_sa) {
+ uint64_t sa_obj;
- if (newvers < ZPL_VERSION_INITIAL || newvers > ZPL_VERSION)
- return (EINVAL);
+ ASSERT3U(spa_version(dmu_objset_spa(zsb->z_os)), >=,
+ SPA_VERSION_SA);
+ sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE,
+ DMU_OT_NONE, 0, tx);
- error = dmu_objset_open(name, DMU_OST_ZFS, DS_MODE_OWNER, &os);
- if (error)
- return (error);
+ error = zap_add(os, MASTER_NODE_OBJ,
+ ZFS_SA_ATTRS, 8, 1, &sa_obj, tx);
+ ASSERT3U(error, ==, 0);
- error = zap_lookup(os, MASTER_NODE_OBJ, ZPL_VERSION_STR,
- 8, 1, &curvers);
- if (error)
- goto out;
- if (newvers < curvers) {
- error = EINVAL;
- goto out;
+ VERIFY(0 == sa_set_sa_object(os, sa_obj));
+ sa_register_update_callback(os, zfs_sa_upgrade);
}
- tx = dmu_tx_create(os);
- dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, 0, ZPL_VERSION_STR);
- error = dmu_tx_assign(tx, TXG_WAIT);
- if (error) {
- dmu_tx_abort(tx);
- goto out;
- }
- error = zap_update(os, MASTER_NODE_OBJ, ZPL_VERSION_STR, 8, 1,
- &newvers, tx);
+ spa_history_log_internal(LOG_DS_UPGRADE,
+ dmu_objset_spa(os), tx, "oldver=%llu newver=%llu dataset = %llu",
+ zsb->z_version, newvers, dmu_objset_id(os));
- spa_history_internal_log(LOG_DS_UPGRADE,
- dmu_objset_spa(os), tx, CRED(),
- "oldver=%llu newver=%llu dataset = %llu", curvers, newvers,
- dmu_objset_id(os));
dmu_tx_commit(tx);
-out:
- dmu_objset_close(os);
- return (error);
+ zsb->z_version = newvers;
+
+ if (zsb->z_version >= ZPL_VERSION_FUID)
+ zfs_set_fuid_feature(zsb);
+
+ return (0);
}
+EXPORT_SYMBOL(zfs_set_version);
/*
* Read a property stored within the master node.
}
return (error);
}
+EXPORT_SYMBOL(zfs_get_zplprop);
+
+void
+zfs_init(void)
+{
+ zfs_znode_init();
+ dmu_objset_register_type(DMU_OST_ZFS, zfs_space_delta_cb);
+ register_filesystem(&zpl_fs_type);
+}
-static vfsdef_t vfw = {
- VFSDEF_VERSION,
- MNTTYPE_ZFS,
- zfs_vfsinit,
- VSW_HASPROTO|VSW_CANRWRO|VSW_CANREMOUNT|VSW_VOLATILEDEV|VSW_STATS|
- VSW_XID,
- &zfs_mntopts
-};
-
-struct modlfs zfs_modlfs = {
- &mod_fsops, "ZFS filesystem version " SPA_VERSION_STRING, &vfw
-};
+void
+zfs_fini(void)
+{
+ unregister_filesystem(&zpl_fs_type);
+ zfs_znode_fini();
+}