X-Git-Url: https://git.camperquake.de/gitweb.cgi?a=blobdiff_plain;f=module%2Fzfs%2Fzfs_vfsops.c;h=f68dde85f82eddebda6e63a5e367b2ae08740c4c;hb=428870ff734fdaccc342b33fc53cf94724409a46;hp=06b4dee4620bbb4f24c62be01c0d48063d180c4b;hpb=172bb4bd5e4afef721dd4d2972d8680d983f144b;p=zfs.git diff --git a/module/zfs/zfs_vfsops.c b/module/zfs/zfs_vfsops.c index 06b4dee..f68dde8 100644 --- a/module/zfs/zfs_vfsops.c +++ b/module/zfs/zfs_vfsops.c @@ -19,10 +19,11 @@ * 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 #include #include @@ -46,6 +47,7 @@ #include #include #include +#include #include #include #include @@ -60,6 +62,8 @@ #include #include #include +#include +#include "zfs_comutil.h" int zfsfstype; vfsops_t *zfs_vfsops = NULL; @@ -67,6 +71,8 @@ static major_t zfs_major; static minor_t zfs_minor; static kmutex_t zfs_dev_mtx; +extern int sys_shutdown; + 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); @@ -145,12 +151,23 @@ zfs_sync(vfs_t *vfsp, short flag, cred_t *cr) * Sync a specific filesystem. */ zfsvfs_t *zfsvfs = vfsp->vfs_data; + dsl_pool_t *dp; ZFS_ENTER(zfsvfs); + dp = dmu_objset_pool(zfsvfs->z_os); + + /* + * If the system is shutting down, then skip any + * filesystems which may exist on a suspended pool. + */ + if (sys_shutdown && spa_suspended(dp->dp_spa)) { + ZFS_EXIT(zfsvfs); + return (0); + } + if (zfsvfs->z_log != NULL) zil_commit(zfsvfs->z_log, UINT64_MAX, 0); - else - txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), 0); + ZFS_EXIT(zfsvfs); } else { /* @@ -367,14 +384,6 @@ vscan_changed_cb(void *arg, uint64_t newval) } static void -acl_mode_changed_cb(void *arg, uint64_t newval) -{ - zfsvfs_t *zfsvfs = arg; - - zfsvfs->z_acl_mode = newval; -} - -static void acl_inherit_changed_cb(void *arg, uint64_t newval) { zfsvfs_t *zfsvfs = arg; @@ -504,8 +513,6 @@ zfs_register_callbacks(vfs_t *vfsp) error = error ? error : dsl_prop_register(ds, "snapdir", snapdir_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, - "aclmode", acl_mode_changed_cb, zfsvfs); - error = error ? error : dsl_prop_register(ds, "aclinherit", acl_inherit_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, "vscan", vscan_changed_cb, zfsvfs); @@ -546,7 +553,6 @@ unregister: (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, "aclinherit", acl_inherit_changed_cb, zfsvfs); (void) dsl_prop_unregister(ds, "vscan", vscan_changed_cb, zfsvfs); @@ -555,6 +561,426 @@ unregister: } static int +zfs_space_delta_cb(dmu_object_type_t bonustype, void *data, + uint64_t *userp, uint64_t *groupp) +{ + znode_phys_t *znp = data; + int error = 0; + + /* + * Is it a valid type of object to track? + */ + if (bonustype != DMU_OT_ZNODE && bonustype != DMU_OT_SA) + return (ENOENT); + + /* + * 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 (data == NULL) + return (EEXIST); + + if (bonustype == DMU_OT_ZNODE) { + *userp = znp->zp_uid; + *groupp = znp->zp_gid; + } else { + int hdrsize; + + ASSERT(bonustype == DMU_OT_SA); + hdrsize = sa_hdrsize(data); + + 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(zfsvfs_t *zfsvfs, const char *fuidstr, + char *domainbuf, int buflen, uid_t *ridp) +{ + uint64_t fuid; + const char *domain; + + fuid = strtonum(fuidstr, NULL); + + domain = zfs_fuid_find_by_idx(zfsvfs, 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(zfsvfs_t *zfsvfs, 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 (zfsvfs->z_userquota_obj); + case ZFS_PROP_GROUPQUOTA: + return (zfsvfs->z_groupquota_obj); + } + return (0); +} + +int +zfs_userspace_many(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type, + uint64_t *cookiep, void *vbuf, uint64_t *bufsizep) +{ + int error; + zap_cursor_t zc; + zap_attribute_t za; + zfs_useracct_t *buf = vbuf; + uint64_t obj; + + if (!dmu_objset_userspace_present(zfsvfs->z_os)) + return (ENOTSUP); + + obj = zfs_userquota_prop_to_obj(zfsvfs, type); + if (obj == 0) { + *bufsizep = 0; + return (0); + } + + for (zap_cursor_init_serialized(&zc, zfsvfs->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(zfsvfs, 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); +} + +/* + * buf must be big enough (eg, 32 bytes) + */ +static int +id_to_fuidstr(zfsvfs_t *zfsvfs, const char *domain, uid_t rid, + char *buf, boolean_t addok) +{ + uint64_t fuid; + int domainid = 0; + + if (domain && domain[0]) { + domainid = zfs_fuid_find_by_domain(zfsvfs, domain, NULL, addok); + if (domainid == -1) + return (ENOENT); + } + fuid = FUID_ENCODE(domainid, rid); + (void) sprintf(buf, "%llx", (longlong_t)fuid); + return (0); +} + +int +zfs_userspace_one(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type, + const char *domain, uint64_t rid, uint64_t *valp) +{ + char buf[32]; + int err; + uint64_t obj; + + *valp = 0; + + if (!dmu_objset_userspace_present(zfsvfs->z_os)) + return (ENOTSUP); + + obj = zfs_userquota_prop_to_obj(zfsvfs, type); + if (obj == 0) + return (0); + + err = id_to_fuidstr(zfsvfs, domain, rid, buf, B_FALSE); + if (err) + return (err); + + err = zap_lookup(zfsvfs->z_os, obj, buf, 8, 1, valp); + if (err == ENOENT) + err = 0; + return (err); +} + +int +zfs_set_userquota(zfsvfs_t *zfsvfs, 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; + + if (type != ZFS_PROP_USERQUOTA && type != ZFS_PROP_GROUPQUOTA) + return (EINVAL); + + if (zfsvfs->z_version < ZPL_VERSION_USERSPACE) + return (ENOTSUP); + + objp = (type == ZFS_PROP_USERQUOTA) ? &zfsvfs->z_userquota_obj : + &zfsvfs->z_groupquota_obj; + + err = id_to_fuidstr(zfsvfs, domain, rid, buf, B_TRUE); + if (err) + return (err); + fuid_dirtied = zfsvfs->z_fuid_dirty; + + tx = dmu_tx_create(zfsvfs->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 (fuid_dirtied) + zfs_fuid_txhold(zfsvfs, tx); + err = dmu_tx_assign(tx, TXG_WAIT); + if (err) { + dmu_tx_abort(tx); + return (err); + } + + mutex_enter(&zfsvfs->z_lock); + if (*objp == 0) { + *objp = zap_create(zfsvfs->z_os, DMU_OT_USERGROUP_QUOTA, + DMU_OT_NONE, 0, tx); + VERIFY(0 == zap_add(zfsvfs->z_os, MASTER_NODE_OBJ, + zfs_userquota_prop_prefixes[type], 8, 1, objp, tx)); + } + mutex_exit(&zfsvfs->z_lock); + + if (quota == 0) { + err = zap_remove(zfsvfs->z_os, *objp, buf, tx); + if (err == ENOENT) + err = 0; + } else { + err = zap_update(zfsvfs->z_os, *objp, buf, 8, 1, "a, tx); + } + ASSERT(err == 0); + if (fuid_dirtied) + zfs_fuid_sync(zfsvfs, tx); + dmu_tx_commit(tx); + return (err); +} + +boolean_t +zfs_fuid_overquota(zfsvfs_t *zfsvfs, boolean_t isgroup, uint64_t fuid) +{ + char buf[32]; + uint64_t used, quota, usedobj, quotaobj; + int err; + + usedobj = isgroup ? DMU_GROUPUSED_OBJECT : DMU_USERUSED_OBJECT; + quotaobj = isgroup ? zfsvfs->z_groupquota_obj : zfsvfs->z_userquota_obj; + + if (quotaobj == 0 || zfsvfs->z_replay) + return (B_FALSE); + + (void) sprintf(buf, "%llx", (longlong_t)fuid); + err = zap_lookup(zfsvfs->z_os, quotaobj, buf, 8, 1, "a); + if (err != 0) + return (B_FALSE); + + err = zap_lookup(zfsvfs->z_os, usedobj, buf, 8, 1, &used); + if (err != 0) + return (B_FALSE); + return (used >= quota); +} + +boolean_t +zfs_owner_overquota(zfsvfs_t *zfsvfs, znode_t *zp, boolean_t isgroup) +{ + uint64_t fuid; + uint64_t quotaobj; + uid_t id; + + quotaobj = isgroup ? zfsvfs->z_groupquota_obj : zfsvfs->z_userquota_obj; + + id = isgroup ? zp->z_gid : zp->z_uid; + + if (quotaobj == 0 || zfsvfs->z_replay) + return (B_FALSE); + + if (IS_EPHEMERAL(id)) { + VERIFY(0 == sa_lookup(zp->z_sa_hdl, + isgroup ? SA_ZPL_GID(zfsvfs) : SA_ZPL_UID(zfsvfs), + &fuid, sizeof (fuid))); + } else { + fuid = (uint64_t)id; + } + + return (zfs_fuid_overquota(zfsvfs, isgroup, fuid)); +} + +int +zfsvfs_create(const char *osname, zfsvfs_t **zfvp) +{ + objset_t *os; + zfsvfs_t *zfsvfs; + uint64_t zval; + int i, error; + uint64_t sa_obj; + + zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP); + + /* + * 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, zfsvfs, &os); + if (error) { + kmem_free(zfsvfs, sizeof (zfsvfs_t)); + return (error); + } + + /* + * Initialize the zfs-specific filesystem structure. + * Should probably make this a kmem cache, shuffle fields, + * and just bzero up to z_hold_mtx[]. + */ + zfsvfs->z_vfs = NULL; + zfsvfs->z_parent = zfsvfs; + zfsvfs->z_max_blksz = SPA_MAXBLOCKSIZE; + zfsvfs->z_show_ctldir = ZFS_SNAPDIR_VISIBLE; + zfsvfs->z_os = os; + + error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zfsvfs->z_version); + if (error) { + goto out; + } else if (zfsvfs->z_version > + zfs_zpl_version_map(spa_version(dmu_objset_spa(os)))) { + (void) printf("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)zfsvfs->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; + zfsvfs->z_norm = (int)zval; + + if ((error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &zval)) != 0) + goto out; + zfsvfs->z_utf8 = (zval != 0); + + if ((error = zfs_get_zplprop(os, ZFS_PROP_CASE, &zval)) != 0) + goto out; + zfsvfs->z_case = (uint_t)zval; + + /* + * Fold case on file systems that are always or sometimes case + * insensitive. + */ + if (zfsvfs->z_case == ZFS_CASE_INSENSITIVE || + zfsvfs->z_case == ZFS_CASE_MIXED) + zfsvfs->z_norm |= U8_TEXTPREP_TOUPPER; + + zfsvfs->z_use_fuids = USE_FUIDS(zfsvfs->z_version, zfsvfs->z_os); + zfsvfs->z_use_sa = USE_SA(zfsvfs->z_version, zfsvfs->z_os); + + if (zfsvfs->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; + } + + zfsvfs->z_attr_table = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END); + + if (zfsvfs->z_version >= ZPL_VERSION_SA) + sa_register_update_callback(os, zfs_sa_upgrade); + + error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1, + &zfsvfs->z_root); + if (error) + goto out; + ASSERT(zfsvfs->z_root != 0); + + error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1, + &zfsvfs->z_unlinkedobj); + if (error) + goto out; + + error = zap_lookup(os, MASTER_NODE_OBJ, + zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA], + 8, 1, &zfsvfs->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, &zfsvfs->z_groupquota_obj); + if (error && error != ENOENT) + goto out; + + error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1, + &zfsvfs->z_fuid_obj); + if (error && error != ENOENT) + goto out; + + error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SHARES_DIR, 8, 1, + &zfsvfs->z_shares_dir); + if (error && error != ENOENT) + goto out; + + mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); + mutex_init(&zfsvfs->z_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); + for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) + mutex_init(&zfsvfs->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL); + + *zfvp = zfsvfs; + return (0); + +out: + dmu_objset_disown(os, zfsvfs); + *zfvp = NULL; + kmem_free(zfsvfs, sizeof (zfsvfs_t)); + return (error); +} + +static int zfsvfs_setup(zfsvfs_t *zfsvfs, boolean_t mounting) { int error; @@ -566,9 +992,11 @@ zfsvfs_setup(zfsvfs_t *zfsvfs, boolean_t mounting) /* * Set the objset user_ptr to track its zfsvfs. */ - mutex_enter(&zfsvfs->z_os->os->os_user_ptr_lock); + mutex_enter(&zfsvfs->z_os->os_user_ptr_lock); dmu_objset_set_user(zfsvfs->z_os, zfsvfs); - mutex_exit(&zfsvfs->z_os->os->os_user_ptr_lock); + mutex_exit(&zfsvfs->z_os->os_user_ptr_lock); + + zfsvfs->z_log = zil_open(zfsvfs->z_os, zfs_get_data); /* * If we are not mounting (ie: online recv), then we don't @@ -583,68 +1011,109 @@ zfsvfs_setup(zfsvfs_t *zfsvfs, boolean_t mounting) * allow replays to succeed. */ readonly = zfsvfs->z_vfs->vfs_flag & VFS_RDONLY; - zfsvfs->z_vfs->vfs_flag &= ~VFS_RDONLY; + if (readonly != 0) + zfsvfs->z_vfs->vfs_flag &= ~VFS_RDONLY; + else + zfs_unlinked_drain(zfsvfs); /* * 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. */ - zil_replay(zfsvfs->z_os, zfsvfs, &zfsvfs->z_assign, - zfs_replay_vector, zfs_unlinked_drain); - - zfs_unlinked_drain(zfsvfs); + if (zil_replay_disable) { + zil_destroy(zfsvfs->z_log, B_FALSE); + } else { + zfsvfs->z_replay = B_TRUE; + zil_replay(zfsvfs->z_os, zfsvfs, zfs_replay_vector); + zfsvfs->z_replay = B_FALSE; + } zfsvfs->z_vfs->vfs_flag |= readonly; /* restore readonly bit */ } - if (!zil_disable) - zfsvfs->z_log = zil_open(zfsvfs->z_os, zfs_get_data); - return (0); } -static void -zfs_freezfsvfs(zfsvfs_t *zfsvfs) +void +zfsvfs_free(zfsvfs_t *zfsvfs) { + int i; + extern krwlock_t zfsvfs_lock; /* in zfs_znode.c */ + + /* + * This is a barrier to prevent the filesystem from going away in + * zfs_znode_move() until we can safely ensure that the filesystem is + * not unmounted. We consider the filesystem valid before the barrier + * and invalid after the barrier. + */ + rw_enter(&zfsvfs_lock, RW_READER); + rw_exit(&zfsvfs_lock); + + zfs_fuid_destroy(zfsvfs); + mutex_destroy(&zfsvfs->z_znodes_lock); - mutex_destroy(&zfsvfs->z_online_recv_lock); + mutex_destroy(&zfsvfs->z_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); + for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) + mutex_destroy(&zfsvfs->z_hold_mtx[i]); kmem_free(zfsvfs, sizeof (zfsvfs_t)); } +static void +zfs_set_fuid_feature(zfsvfs_t *zfsvfs) +{ + zfsvfs->z_use_fuids = USE_FUIDS(zfsvfs->z_version, zfsvfs->z_os); + if (zfsvfs->z_use_fuids && zfsvfs->z_vfs) { + vfs_set_feature(zfsvfs->z_vfs, VFSFT_XVATTR); + vfs_set_feature(zfsvfs->z_vfs, VFSFT_SYSATTR_VIEWS); + vfs_set_feature(zfsvfs->z_vfs, VFSFT_ACEMASKONACCESS); + vfs_set_feature(zfsvfs->z_vfs, VFSFT_ACLONCREATE); + vfs_set_feature(zfsvfs->z_vfs, VFSFT_ACCESS_FILTER); + vfs_set_feature(zfsvfs->z_vfs, VFSFT_REPARSE); + } + zfsvfs->z_use_sa = USE_SA(zfsvfs->z_version, zfsvfs->z_os); +} + static int zfs_domount(vfs_t *vfsp, char *osname) { dev_t mount_dev; - uint64_t recordsize, readonly; + uint64_t recordsize, fsid_guid; int error = 0; - int mode; zfsvfs_t *zfsvfs; - znode_t *zp = NULL; ASSERT(vfsp); ASSERT(osname); - /* - * 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); + error = zfsvfs_create(osname, &zfsvfs); + if (error) + return (error); 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); /* Initialize the generic filesystem structure. */ vfsp->vfs_bcount = 0; @@ -666,39 +1135,24 @@ zfs_domount(vfs_t *vfsp, char *osname) vfsp->vfs_flag |= VFS_NOTRUNC; vfsp->vfs_data = zfsvfs; - if (error = dsl_prop_get_integer(osname, "readonly", &readonly, NULL)) - goto out; - - mode = DS_MODE_OWNER; - if (readonly) - mode |= DS_MODE_READONLY; - - 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 (error) - goto out; - - if (error = zfs_init_fs(zfsvfs, &zp)) - goto out; - - /* The call to zfs_init_fs leaves the vnode held, release it here. */ - VN_RELE(ZTOV(zp)); + /* + * The fsid is 64 bits, composed of an 8-bit fs type, which + * separates our fsid from any other filesystem types, and a + * 56-bit objset unique ID. The objset unique ID is unique to + * all objsets open on this system, provided by unique_create(). + * The 8-bit fs type must be put in the low bits of fsid[1] + * because that's where other Solaris filesystems put it. + */ + fsid_guid = dmu_objset_fsid_guid(zfsvfs->z_os); + ASSERT((fsid_guid & ~((1ULL<<56)-1)) == 0); + vfsp->vfs_fsid.val[0] = fsid_guid; + vfsp->vfs_fsid.val[1] = ((fsid_guid>>32) << 8) | + zfsfstype & 0xFF; /* * 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); - } + zfs_set_fuid_feature(zfsvfs); if (zfsvfs->z_case == ZFS_CASE_INSENSITIVE) { vfs_set_feature(vfsp, VFSFT_DIRENTFLAGS); vfs_set_feature(vfsp, VFSFT_CASEINSENSITIVE); @@ -707,17 +1161,21 @@ zfs_domount(vfs_t *vfsp, char *osname) vfs_set_feature(vfsp, VFSFT_DIRENTFLAGS); vfs_set_feature(vfsp, VFSFT_CASEINSENSITIVE); } + vfs_set_feature(vfsp, VFSFT_ZEROCOPY_SUPPORTED); 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; + + mutex_enter(&zfsvfs->z_os->os_user_ptr_lock); + dmu_objset_set_user(zfsvfs->z_os, zfsvfs); + mutex_exit(&zfsvfs->z_os->os_user_ptr_lock); } else { error = zfsvfs_setup(zfsvfs, B_TRUE); } @@ -726,9 +1184,8 @@ zfs_domount(vfs_t *vfsp, char *osname) zfsctl_create(zfsvfs); out: if (error) { - if (zfsvfs->z_os) - dmu_objset_close(zfsvfs->z_os); - zfs_freezfsvfs(zfsvfs); + dmu_objset_disown(zfsvfs->z_os, zfsvfs); + zfsvfs_free(zfsvfs); } else { atomic_add_32(&zfs_active_fs_count, 1); } @@ -771,9 +1228,6 @@ zfs_unregister_callbacks(zfsvfs_t *zfsvfs) VERIFY(dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb, zfsvfs) == 0); - VERIFY(dsl_prop_unregister(ds, "aclmode", acl_mode_changed_cb, - zfsvfs) == 0); - VERIFY(dsl_prop_unregister(ds, "aclinherit", acl_inherit_changed_cb, zfsvfs) == 0); @@ -837,6 +1291,139 @@ zfs_parse_bootfs(char *bpath, char *outpath) return (error); } +/* + * 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); +} + +/* + * zfs_mount_label_policy: + * Determine whether the mount is allowed according to MAC check. + * by comparing (where appropriate) label of the dataset against + * the label of the zone being mounted into. If the dataset has + * no label, create one. + * + * Returns: + * 0 : access allowed + * >0 : error code, such as EACCES + */ +static int +zfs_mount_label_policy(vfs_t *vfsp, char *osname) +{ + int error, retv; + zone_t *mntzone = NULL; + ts_label_t *mnt_tsl; + bslabel_t *mnt_sl; + bslabel_t ds_sl; + char ds_hexsl[MAXNAMELEN]; + + retv = EACCES; /* assume the worst */ + + /* + * Start by getting the dataset label if it exists. + */ + error = dsl_prop_get(osname, zfs_prop_to_name(ZFS_PROP_MLSLABEL), + 1, sizeof (ds_hexsl), &ds_hexsl, NULL); + if (error) + return (EACCES); + + /* + * If labeling is NOT enabled, then disallow the mount of datasets + * which have a non-default label already. No other label checks + * are needed. + */ + if (!is_system_labeled()) { + if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) == 0) + return (0); + return (EACCES); + } + + /* + * Get the label of the mountpoint. If mounting into the global + * zone (i.e. mountpoint is not within an active zone and the + * zoned property is off), the label must be default or + * admin_low/admin_high only; no other checks are needed. + */ + mntzone = zone_find_by_any_path(refstr_value(vfsp->vfs_mntpt), B_FALSE); + if (mntzone->zone_id == GLOBAL_ZONEID) { + uint64_t zoned; + + zone_rele(mntzone); + + if (dsl_prop_get_integer(osname, + zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL)) + return (EACCES); + if (!zoned) + return (zfs_check_global_label(osname, ds_hexsl)); + else + /* + * This is the case of a zone dataset being mounted + * initially, before the zone has been fully created; + * allow this mount into global zone. + */ + return (0); + } + + mnt_tsl = mntzone->zone_slabel; + ASSERT(mnt_tsl != NULL); + label_hold(mnt_tsl); + mnt_sl = label2bslabel(mnt_tsl); + + if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) == 0) { + /* + * The dataset doesn't have a real label, so fabricate one. + */ + char *str = NULL; + + if (l_to_str_internal(mnt_sl, &str) == 0 && + dsl_prop_set(osname, zfs_prop_to_name(ZFS_PROP_MLSLABEL), + ZPROP_SRC_LOCAL, 1, strlen(str) + 1, str) == 0) + retv = 0; + if (str != NULL) + kmem_free(str, strlen(str) + 1); + } else if (hexstr_to_label(ds_hexsl, &ds_sl) == 0) { + /* + * Now compare labels to complete the MAC check. If the + * labels are equal then allow access. If the mountpoint + * label dominates the dataset label, allow readonly access. + * Otherwise, access is denied. + */ + if (blequal(mnt_sl, &ds_sl)) + retv = 0; + else if (bldominates(mnt_sl, &ds_sl)) { + vfs_setmntopt(vfsp, MNTOPT_RO, NULL, 0); + retv = 0; + } + } + + label_rele(mnt_tsl); + zone_rele(mntzone); + return (retv); +} + static int zfs_mountroot(vfs_t *vfsp, enum whymountroot why) { @@ -989,8 +1576,7 @@ zfs_mount(vfs_t *vfsp, vnode_t *mvp, struct mounta *uap, cred_t *cr) */ error = secpolicy_fs_mount(cr, mvp, vfsp); if (error) { - error = dsl_deleg_access(osname, ZFS_DELEG_PERM_MOUNT, cr); - if (error == 0) { + if (dsl_deleg_access(osname, ZFS_DELEG_PERM_MOUNT, cr) == 0) { vattr_t vattr; /* @@ -1000,16 +1586,14 @@ zfs_mount(vfs_t *vfsp, vnode_t *mvp, struct mounta *uap, cred_t *cr) vattr.va_mask = AT_UID; - if (error = VOP_GETATTR(mvp, &vattr, 0, cr, NULL)) { + if (VOP_GETATTR(mvp, &vattr, 0, cr, NULL)) { goto out; } if (secpolicy_vnode_owner(cr, vattr.va_uid) != 0 && VOP_ACCESS(mvp, VWRITE, 0, cr, NULL) != 0) { - error = EPERM; goto out; } - secpolicy_fs_mount_clearopts(cr, vfsp); } else { goto out; @@ -1026,6 +1610,10 @@ zfs_mount(vfs_t *vfsp, vnode_t *mvp, struct mounta *uap, cred_t *cr) goto out; } + error = zfs_mount_label_policy(vfsp, osname); + if (error) + goto out; + /* * When doing a remount, we simply refresh our temporary properties * according to those options set in the current VFS options. @@ -1039,6 +1627,13 @@ zfs_mount(vfs_t *vfsp, vnode_t *mvp, struct mounta *uap, cred_t *cr) error = zfs_domount(vfsp, osname); + /* + * Add an extra VFS_HOLD on our parent vfs so that it can't + * disappear due to a forced unmount. + */ + if (error == 0 && ((zfsvfs_t *)vfsp->vfs_data)->z_issnap) + VFS_HOLD(mvp->v_vfsp); + out: pn_free(&spn); return (error); @@ -1180,7 +1775,7 @@ zfsvfs_teardown(zfsvfs_t *zfsvfs, boolean_t unmounting) 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) { + if (zp->z_sa_hdl) { ASSERT(ZTOV(zp)->v_count > 0); zfs_znode_dmu_fini(zp); } @@ -1231,9 +1826,8 @@ zfs_umount(vfs_t *vfsp, int fflag, cred_t *cr) 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) + if (dsl_deleg_access((char *)refstr_value(vfsp->vfs_resource), + ZFS_DELEG_PERM_MOUNT, cr)) return (ret); } @@ -1288,14 +1882,14 @@ zfs_umount(vfs_t *vfsp, int fflag, cred_t *cr) /* * 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, zfsvfs); } /* @@ -1376,7 +1970,9 @@ zfs_vget(vfs_t *vfsp, vnode_t **vpp, fid_t *fidp) ZFS_EXIT(zfsvfs); return (err); } - zp_gen = zp->z_phys->zp_gen & gen_mask; + (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zfsvfs), &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) { @@ -1398,16 +1994,13 @@ zfs_vget(vfs_t *vfsp, vnode_t **vpp, fid_t *fidp) * 'z_teardown_inactive_lock' write held. */ int -zfs_suspend_fs(zfsvfs_t *zfsvfs, char *name, int *mode) +zfs_suspend_fs(zfsvfs_t *zfsvfs) { int error; if ((error = zfsvfs_teardown(zfsvfs, 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(zfsvfs->z_os, zfsvfs); return (0); } @@ -1416,18 +2009,30 @@ zfs_suspend_fs(zfsvfs_t *zfsvfs, char *name, int *mode) * Reopen zfsvfs_t::z_os and release VOPs. */ int -zfs_resume_fs(zfsvfs_t *zfsvfs, const char *osname, int mode) +zfs_resume_fs(zfsvfs_t *zfsvfs, 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)); - err = dmu_objset_open(osname, DMU_OST_ZFS, mode, &zfsvfs->z_os); + err = dmu_objset_own(osname, DMU_OST_ZFS, B_FALSE, zfsvfs, + &zfsvfs->z_os); if (err) { zfsvfs->z_os = NULL; } else { znode_t *zp; + uint64_t sa_obj = 0; + + err2 = zap_lookup(zfsvfs->z_os, MASTER_NODE_OBJ, + ZFS_SA_ATTRS, 8, 1, &sa_obj); + + if ((err || err2) && zfsvfs->z_version >= ZPL_VERSION_SA) + goto bail; + + + zfsvfs->z_attr_table = sa_setup(zfsvfs->z_os, sa_obj, + zfs_attr_table, ZPL_END); VERIFY(zfsvfs_setup(zfsvfs, B_FALSE) == 0); @@ -1446,6 +2051,7 @@ zfs_resume_fs(zfsvfs_t *zfsvfs, const char *osname, int mode) } +bail: /* release the VOPs */ rw_exit(&zfsvfs->z_teardown_inactive_lock); rrw_exit(&zfsvfs->z_teardown_lock, FTAG); @@ -1465,13 +2071,17 @@ static void zfs_freevfs(vfs_t *vfsp) { zfsvfs_t *zfsvfs = vfsp->vfs_data; - int i; - for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) - mutex_destroy(&zfsvfs->z_hold_mtx[i]); + /* + * If this is a snapshot, we have an extra VFS_HOLD on our parent + * from zfs_mount(). Release it here. If we came through + * zfs_mountroot() instead, we didn't grab an extra hold, so + * skip the VFS_RELE for rootvfs. + */ + if (zfsvfs->z_issnap && (vfsp != rootvfs)) + VFS_RELE(zfsvfs->z_parent->z_vfs); - zfs_fuid_destroy(zfsvfs); - zfs_freezfsvfs(zfsvfs); + zfsvfs_free(zfsvfs); atomic_add_32(&zfs_active_fs_count, -1); } @@ -1530,6 +2140,8 @@ zfs_init(void) * Initialize znode cache, vnode ops, etc... */ zfs_znode_init(); + + dmu_objset_register_type(DMU_OST_ZFS, zfs_space_delta_cb); } void @@ -1546,54 +2158,71 @@ zfs_busy(void) } int -zfs_set_version(const char *name, uint64_t newvers) +zfs_set_version(zfsvfs_t *zfsvfs, uint64_t newvers) { int error; - objset_t *os; + objset_t *os = zfsvfs->z_os; dmu_tx_t *tx; - uint64_t curvers; - - /* - * 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_INITIAL || newvers > ZPL_VERSION) return (EINVAL); - error = dmu_objset_open(name, DMU_OST_ZFS, DS_MODE_OWNER, &os); - if (error) - return (error); + if (newvers < zfsvfs->z_version) + return (EINVAL); - error = zap_lookup(os, MASTER_NODE_OBJ, ZPL_VERSION_STR, - 8, 1, &curvers); - if (error) - goto out; - if (newvers < curvers) { - error = EINVAL; - goto out; - } + if (zfs_spa_version_map(newvers) > + spa_version(dmu_objset_spa(zfsvfs->z_os))) + return (ENOTSUP); tx = dmu_tx_create(os); - dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, 0, ZPL_VERSION_STR); + dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_FALSE, ZPL_VERSION_STR); + if (newvers >= ZPL_VERSION_SA && !zfsvfs->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 = dmu_tx_assign(tx, TXG_WAIT); if (error) { dmu_tx_abort(tx); - goto out; + return (error); + } + + error = zap_update(os, MASTER_NODE_OBJ, ZPL_VERSION_STR, + 8, 1, &newvers, tx); + + if (error) { + dmu_tx_commit(tx); + return (error); } - error = zap_update(os, MASTER_NODE_OBJ, ZPL_VERSION_STR, 8, 1, - &newvers, tx); - 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)); + if (newvers >= ZPL_VERSION_SA && !zfsvfs->z_use_sa) { + uint64_t sa_obj; + + ASSERT3U(spa_version(dmu_objset_spa(zfsvfs->z_os)), >=, + SPA_VERSION_SA); + sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE, + DMU_OT_NONE, 0, tx); + + error = zap_add(os, MASTER_NODE_OBJ, + ZFS_SA_ATTRS, 8, 1, &sa_obj, tx); + ASSERT3U(error, ==, 0); + + VERIFY(0 == sa_set_sa_object(os, sa_obj)); + sa_register_update_callback(os, zfs_sa_upgrade); + } + + spa_history_log_internal(LOG_DS_UPGRADE, + dmu_objset_spa(os), tx, "oldver=%llu newver=%llu dataset = %llu", + zfsvfs->z_version, newvers, dmu_objset_id(os)); + dmu_tx_commit(tx); -out: - dmu_objset_close(os); - return (error); + zfsvfs->z_version = newvers; + + if (zfsvfs->z_version >= ZPL_VERSION_FUID) + zfs_set_fuid_feature(zfsvfs); + + return (0); } /*