X-Git-Url: https://git.camperquake.de/gitweb.cgi?a=blobdiff_plain;f=module%2Fzfs%2Fzfs_znode.c;h=aaf17e18f4951969b932acb7989d75b7fc33552d;hb=refs%2Fheads%2Frertzinger%2Ffeature-zpool-get--p;hp=f99e72f1d5f4dad2df19fa2e872d0965605a6661;hpb=45d1cae3b8c949ecc391dd7a5b81963b34c71c29;p=zfs.git diff --git a/module/zfs/zfs_znode.c b/module/zfs/zfs_znode.c index f99e72f..aaf17e1 100644 --- a/module/zfs/zfs_znode.c +++ b/module/zfs/zfs_znode.c @@ -19,8 +19,8 @@ * CDDL HEADER END */ /* - * Copyright 2009 Sun Microsystems, Inc. All rights reserved. - * Use is subject to license terms. + * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 2012 by Delphix. All rights reserved. */ /* Portions Copyright 2007 Jeremy Teo */ @@ -52,8 +52,12 @@ #include #include #include +#include +#include +#include #include #include +#include #endif /* _KERNEL */ #include @@ -61,8 +65,12 @@ #include #include #include +#include +#include +#include #include "zfs_prop.h" +#include "zfs_comutil.h" /* * Define ZNODE_STATS to turn on statistic gathering. By default, it is only @@ -78,62 +86,39 @@ #define ZNODE_STAT_ADD(stat) /* nothing */ #endif /* ZNODE_STATS */ -#define POINTER_IS_VALID(p) (!((uintptr_t)(p) & 0x3)) -#define POINTER_INVALIDATE(pp) (*(pp) = (void *)((uintptr_t)(*(pp)) | 0x1)) - /* * Functions needed for userland (ie: libzpool) are not put under * #ifdef_KERNEL; the rest of the functions have dependencies * (such as VFS logic) that will not compile easily in userland. */ #ifdef _KERNEL -/* - * Needed to close a small window in zfs_znode_move() that allows the zfsvfs to - * be freed before it can be safely accessed. - */ -krwlock_t zfsvfs_lock; static kmem_cache_t *znode_cache = NULL; /*ARGSUSED*/ -static void -znode_evict_error(dmu_buf_t *dbuf, void *user_ptr) -{ - /* - * We should never drop all dbuf refs without first clearing - * the eviction callback. - */ - panic("evicting znode %p\n", user_ptr); -} - -/*ARGSUSED*/ static int zfs_znode_cache_constructor(void *buf, void *arg, int kmflags) { znode_t *zp = buf; - ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs)); - - zp->z_vnode = vn_alloc(kmflags); - if (zp->z_vnode == NULL) { - return (-1); - } - ZTOV(zp)->v_data = zp; - + inode_init_once(ZTOI(zp)); list_link_init(&zp->z_link_node); mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL); rw_init(&zp->z_parent_lock, NULL, RW_DEFAULT, NULL); rw_init(&zp->z_name_lock, NULL, RW_DEFAULT, NULL); mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL); + rw_init(&zp->z_xattr_lock, NULL, RW_DEFAULT, NULL); mutex_init(&zp->z_range_lock, NULL, MUTEX_DEFAULT, NULL); avl_create(&zp->z_range_avl, zfs_range_compare, sizeof (rl_t), offsetof(rl_t, r_node)); - zp->z_dbuf = NULL; zp->z_dirlocks = NULL; zp->z_acl_cached = NULL; + zp->z_xattr_cached = NULL; + zp->z_xattr_parent = NULL; + zp->z_moved = 0; return (0); } @@ -143,196 +128,19 @@ zfs_znode_cache_destructor(void *buf, void *arg) { znode_t *zp = buf; - ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs)); - ASSERT(ZTOV(zp)->v_data == zp); - vn_free(ZTOV(zp)); ASSERT(!list_link_active(&zp->z_link_node)); mutex_destroy(&zp->z_lock); rw_destroy(&zp->z_parent_lock); rw_destroy(&zp->z_name_lock); mutex_destroy(&zp->z_acl_lock); + rw_destroy(&zp->z_xattr_lock); avl_destroy(&zp->z_range_avl); mutex_destroy(&zp->z_range_lock); - ASSERT(zp->z_dbuf == NULL); ASSERT(zp->z_dirlocks == NULL); ASSERT(zp->z_acl_cached == NULL); -} - -#ifdef ZNODE_STATS -static struct { - uint64_t zms_zfsvfs_invalid; - uint64_t zms_zfsvfs_recheck1; - uint64_t zms_zfsvfs_unmounted; - uint64_t zms_zfsvfs_recheck2; - uint64_t zms_obj_held; - uint64_t zms_vnode_locked; - uint64_t zms_not_only_dnlc; -} znode_move_stats; -#endif /* ZNODE_STATS */ - -static void -zfs_znode_move_impl(znode_t *ozp, znode_t *nzp) -{ - vnode_t *vp; - - /* Copy fields. */ - nzp->z_zfsvfs = ozp->z_zfsvfs; - - /* Swap vnodes. */ - vp = nzp->z_vnode; - nzp->z_vnode = ozp->z_vnode; - ozp->z_vnode = vp; /* let destructor free the overwritten vnode */ - ZTOV(ozp)->v_data = ozp; - ZTOV(nzp)->v_data = nzp; - - nzp->z_id = ozp->z_id; - ASSERT(ozp->z_dirlocks == NULL); /* znode not in use */ - ASSERT(avl_numnodes(&ozp->z_range_avl) == 0); - nzp->z_unlinked = ozp->z_unlinked; - nzp->z_atime_dirty = ozp->z_atime_dirty; - nzp->z_zn_prefetch = ozp->z_zn_prefetch; - nzp->z_blksz = ozp->z_blksz; - nzp->z_seq = ozp->z_seq; - nzp->z_mapcnt = ozp->z_mapcnt; - nzp->z_last_itx = ozp->z_last_itx; - nzp->z_gen = ozp->z_gen; - nzp->z_sync_cnt = ozp->z_sync_cnt; - nzp->z_phys = ozp->z_phys; - nzp->z_dbuf = ozp->z_dbuf; - - /* - * Release any cached ACL, since it *may* have - * zfs_acl_node_t's that directly references an - * embedded ACL in the zp_acl of the old znode_phys_t - * - * It will be recached the next time the ACL is needed. - */ - if (ozp->z_acl_cached) { - zfs_acl_free(ozp->z_acl_cached); - ozp->z_acl_cached = NULL; - } - - /* Update back pointers. */ - (void) dmu_buf_update_user(nzp->z_dbuf, ozp, nzp, &nzp->z_phys, - znode_evict_error); - - /* - * Invalidate the original znode by clearing fields that provide a - * pointer back to the znode. Set the low bit of the vfs pointer to - * ensure that zfs_znode_move() recognizes the znode as invalid in any - * subsequent callback. - */ - ozp->z_dbuf = NULL; - POINTER_INVALIDATE(&ozp->z_zfsvfs); -} - -/*ARGSUSED*/ -static kmem_cbrc_t -zfs_znode_move(void *buf, void *newbuf, size_t size, void *arg) -{ - znode_t *ozp = buf, *nzp = newbuf; - zfsvfs_t *zfsvfs; - vnode_t *vp; - - /* - * The znode is on the file system's list of known znodes if the vfs - * pointer is valid. We set the low bit of the vfs pointer when freeing - * the znode to invalidate it, and the memory patterns written by kmem - * (baddcafe and deadbeef) set at least one of the two low bits. A newly - * created znode sets the vfs pointer last of all to indicate that the - * znode is known and in a valid state to be moved by this function. - */ - zfsvfs = ozp->z_zfsvfs; - if (!POINTER_IS_VALID(zfsvfs)) { - ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_invalid); - return (KMEM_CBRC_DONT_KNOW); - } - - /* - * Close a small window in which it's possible that the filesystem could - * be unmounted and freed, and zfsvfs, though valid in the previous - * statement, could point to unrelated memory by the time we try to - * prevent the filesystem from being unmounted. - */ - rw_enter(&zfsvfs_lock, RW_WRITER); - if (zfsvfs != ozp->z_zfsvfs) { - rw_exit(&zfsvfs_lock); - ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_recheck1); - return (KMEM_CBRC_DONT_KNOW); - } - - /* - * If the znode is still valid, then so is the file system. We know that - * no valid file system can be freed while we hold zfsvfs_lock, so we - * can safely ensure that the filesystem is not and will not be - * unmounted. The next statement is equivalent to ZFS_ENTER(). - */ - rrw_enter(&zfsvfs->z_teardown_lock, RW_READER, FTAG); - if (zfsvfs->z_unmounted) { - ZFS_EXIT(zfsvfs); - rw_exit(&zfsvfs_lock); - ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_unmounted); - return (KMEM_CBRC_DONT_KNOW); - } - rw_exit(&zfsvfs_lock); - - mutex_enter(&zfsvfs->z_znodes_lock); - /* - * Recheck the vfs pointer in case the znode was removed just before - * acquiring the lock. - */ - if (zfsvfs != ozp->z_zfsvfs) { - mutex_exit(&zfsvfs->z_znodes_lock); - ZFS_EXIT(zfsvfs); - ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_recheck2); - return (KMEM_CBRC_DONT_KNOW); - } - - /* - * At this point we know that as long as we hold z_znodes_lock, the - * znode cannot be freed and fields within the znode can be safely - * accessed. Now, prevent a race with zfs_zget(). - */ - if (ZFS_OBJ_HOLD_TRYENTER(zfsvfs, ozp->z_id) == 0) { - mutex_exit(&zfsvfs->z_znodes_lock); - ZFS_EXIT(zfsvfs); - ZNODE_STAT_ADD(znode_move_stats.zms_obj_held); - return (KMEM_CBRC_LATER); - } - - vp = ZTOV(ozp); - if (mutex_tryenter(&vp->v_lock) == 0) { - ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id); - mutex_exit(&zfsvfs->z_znodes_lock); - ZFS_EXIT(zfsvfs); - ZNODE_STAT_ADD(znode_move_stats.zms_vnode_locked); - return (KMEM_CBRC_LATER); - } - - /* Only move znodes that are referenced _only_ by the DNLC. */ - if (vp->v_count != 1 || !vn_in_dnlc(vp)) { - mutex_exit(&vp->v_lock); - ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id); - mutex_exit(&zfsvfs->z_znodes_lock); - ZFS_EXIT(zfsvfs); - ZNODE_STAT_ADD(znode_move_stats.zms_not_only_dnlc); - return (KMEM_CBRC_LATER); - } - - /* - * The znode is known and in a valid state to move. We're holding the - * locks needed to execute the critical section. - */ - zfs_znode_move_impl(ozp, nzp); - mutex_exit(&vp->v_lock); - ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id); - - list_link_replace(&ozp->z_link_node, &nzp->z_link_node); - mutex_exit(&zfsvfs->z_znodes_lock); - ZFS_EXIT(zfsvfs); - - return (KMEM_CBRC_YES); + ASSERT(zp->z_xattr_cached == NULL); + ASSERT(zp->z_xattr_parent == NULL); } void @@ -341,126 +149,27 @@ zfs_znode_init(void) /* * Initialize zcache */ - rw_init(&zfsvfs_lock, NULL, RW_DEFAULT, NULL); ASSERT(znode_cache == NULL); znode_cache = kmem_cache_create("zfs_znode_cache", sizeof (znode_t), 0, zfs_znode_cache_constructor, - zfs_znode_cache_destructor, NULL, NULL, NULL, 0); - kmem_cache_set_move(znode_cache, zfs_znode_move); + zfs_znode_cache_destructor, NULL, NULL, NULL, KMC_KMEM); } void zfs_znode_fini(void) { /* - * Cleanup vfs & vnode ops - */ - zfs_remove_op_tables(); - - /* * Cleanup zcache */ if (znode_cache) kmem_cache_destroy(znode_cache); znode_cache = NULL; - rw_destroy(&zfsvfs_lock); -} - -struct vnodeops *zfs_dvnodeops; -struct vnodeops *zfs_fvnodeops; -struct vnodeops *zfs_symvnodeops; -struct vnodeops *zfs_xdvnodeops; -struct vnodeops *zfs_evnodeops; -struct vnodeops *zfs_sharevnodeops; - -void -zfs_remove_op_tables() -{ - /* - * Remove vfs ops - */ - ASSERT(zfsfstype); - (void) vfs_freevfsops_by_type(zfsfstype); - zfsfstype = 0; - - /* - * Remove vnode ops - */ - if (zfs_dvnodeops) - vn_freevnodeops(zfs_dvnodeops); - if (zfs_fvnodeops) - vn_freevnodeops(zfs_fvnodeops); - if (zfs_symvnodeops) - vn_freevnodeops(zfs_symvnodeops); - if (zfs_xdvnodeops) - vn_freevnodeops(zfs_xdvnodeops); - if (zfs_evnodeops) - vn_freevnodeops(zfs_evnodeops); - if (zfs_sharevnodeops) - vn_freevnodeops(zfs_sharevnodeops); - - zfs_dvnodeops = NULL; - zfs_fvnodeops = NULL; - zfs_symvnodeops = NULL; - zfs_xdvnodeops = NULL; - zfs_evnodeops = NULL; - zfs_sharevnodeops = NULL; } -extern const fs_operation_def_t zfs_dvnodeops_template[]; -extern const fs_operation_def_t zfs_fvnodeops_template[]; -extern const fs_operation_def_t zfs_xdvnodeops_template[]; -extern const fs_operation_def_t zfs_symvnodeops_template[]; -extern const fs_operation_def_t zfs_evnodeops_template[]; -extern const fs_operation_def_t zfs_sharevnodeops_template[]; - int -zfs_create_op_tables() -{ - int error; - - /* - * zfs_dvnodeops can be set if mod_remove() calls mod_installfs() - * due to a failure to remove the the 2nd modlinkage (zfs_modldrv). - * In this case we just return as the ops vectors are already set up. - */ - if (zfs_dvnodeops) - return (0); - - error = vn_make_ops(MNTTYPE_ZFS, zfs_dvnodeops_template, - &zfs_dvnodeops); - if (error) - return (error); - - error = vn_make_ops(MNTTYPE_ZFS, zfs_fvnodeops_template, - &zfs_fvnodeops); - if (error) - return (error); - - error = vn_make_ops(MNTTYPE_ZFS, zfs_symvnodeops_template, - &zfs_symvnodeops); - if (error) - return (error); - - error = vn_make_ops(MNTTYPE_ZFS, zfs_xdvnodeops_template, - &zfs_xdvnodeops); - if (error) - return (error); - - error = vn_make_ops(MNTTYPE_ZFS, zfs_evnodeops_template, - &zfs_evnodeops); - if (error) - return (error); - - error = vn_make_ops(MNTTYPE_ZFS, zfs_sharevnodeops_template, - &zfs_sharevnodeops); - - return (error); -} - -int -zfs_create_share_dir(zfsvfs_t *zfsvfs, dmu_tx_t *tx) +zfs_create_share_dir(zfs_sb_t *zsb, dmu_tx_t *tx) { +#ifdef HAVE_SMB_SHARE zfs_acl_ids_t acl_ids; vattr_t vattr; znode_t *sharezp; @@ -469,15 +178,16 @@ zfs_create_share_dir(zfsvfs_t *zfsvfs, dmu_tx_t *tx) int error; vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE; - vattr.va_type = VDIR; - vattr.va_mode = S_IFDIR|0555; + vattr.va_mode = S_IFDIR | 0555; vattr.va_uid = crgetuid(kcred); vattr.va_gid = crgetgid(kcred); - sharezp = kmem_cache_alloc(znode_cache, KM_SLEEP); + sharezp = kmem_cache_alloc(znode_cache, KM_PUSHPAGE); + sharezp->z_moved = 0; sharezp->z_unlinked = 0; sharezp->z_atime_dirty = 0; sharezp->z_zfsvfs = zfsvfs; + sharezp->z_is_sa = zfsvfs->z_use_sa; vp = ZTOV(sharezp); vn_reinit(vp); @@ -485,8 +195,7 @@ zfs_create_share_dir(zfsvfs_t *zfsvfs, dmu_tx_t *tx) VERIFY(0 == zfs_acl_ids_create(sharezp, IS_ROOT_NODE, &vattr, kcred, NULL, &acl_ids)); - zfs_mknode(sharezp, &vattr, tx, kcred, IS_ROOT_NODE, - &zp, 0, &acl_ids); + zfs_mknode(sharezp, &vattr, tx, kcred, IS_ROOT_NODE, &zp, &acl_ids); ASSERT3P(zp, ==, sharezp); ASSERT(!vn_in_dnlc(ZTOV(sharezp))); /* not valid to move */ POINTER_INVALIDATE(&sharezp->z_zfsvfs); @@ -495,206 +204,293 @@ zfs_create_share_dir(zfsvfs_t *zfsvfs, dmu_tx_t *tx) zfsvfs->z_shares_dir = sharezp->z_id; zfs_acl_ids_free(&acl_ids); - ZTOV(sharezp)->v_count = 0; - dmu_buf_rele(sharezp->z_dbuf, NULL); - sharezp->z_dbuf = NULL; + // ZTOV(sharezp)->v_count = 0; + sa_handle_destroy(sharezp->z_sa_hdl); kmem_cache_free(znode_cache, sharezp); return (error); +#else + return (0); +#endif /* HAVE_SMB_SHARE */ } -/* - * define a couple of values we need available - * for both 64 and 32 bit environments. - */ -#ifndef NBITSMINOR64 -#define NBITSMINOR64 32 -#endif -#ifndef MAXMAJ64 -#define MAXMAJ64 0xffffffffUL -#endif -#ifndef MAXMIN64 -#define MAXMIN64 0xffffffffUL -#endif +static void +zfs_znode_sa_init(zfs_sb_t *zsb, znode_t *zp, + dmu_buf_t *db, dmu_object_type_t obj_type, sa_handle_t *sa_hdl) +{ + ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zsb, zp->z_id))); + + mutex_enter(&zp->z_lock); + + ASSERT(zp->z_sa_hdl == NULL); + ASSERT(zp->z_acl_cached == NULL); + if (sa_hdl == NULL) { + VERIFY(0 == sa_handle_get_from_db(zsb->z_os, db, zp, + SA_HDL_SHARED, &zp->z_sa_hdl)); + } else { + zp->z_sa_hdl = sa_hdl; + sa_set_userp(sa_hdl, zp); + } + + zp->z_is_sa = (obj_type == DMU_OT_SA) ? B_TRUE : B_FALSE; + + mutex_exit(&zp->z_lock); +} + +void +zfs_znode_dmu_fini(znode_t *zp) +{ + ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(ZTOZSB(zp), zp->z_id)) || + zp->z_unlinked || + RW_WRITE_HELD(&ZTOZSB(zp)->z_teardown_inactive_lock)); + + sa_handle_destroy(zp->z_sa_hdl); + zp->z_sa_hdl = NULL; +} /* - * Create special expldev for ZFS private use. - * Can't use standard expldev since it doesn't do - * what we want. The standard expldev() takes a - * dev32_t in LP64 and expands it to a long dev_t. - * We need an interface that takes a dev32_t in ILP32 - * and expands it to a long dev_t. + * Called by new_inode() to allocate a new inode. */ -static uint64_t -zfs_expldev(dev_t dev) +int +zfs_inode_alloc(struct super_block *sb, struct inode **ip) { -#ifndef _LP64 - major_t major = (major_t)dev >> NBITSMINOR32 & MAXMAJ32; - return (((uint64_t)major << NBITSMINOR64) | - ((minor_t)dev & MAXMIN32)); -#else - return (dev); -#endif + znode_t *zp; + + zp = kmem_cache_alloc(znode_cache, KM_PUSHPAGE); + *ip = ZTOI(zp); + + return (0); } /* - * Special cmpldev for ZFS private use. - * Can't use standard cmpldev since it takes - * a long dev_t and compresses it to dev32_t in - * LP64. We need to do a compaction of a long dev_t - * to a dev32_t in ILP32. + * Called in multiple places when an inode should be destroyed. */ -dev_t -zfs_cmpldev(uint64_t dev) +void +zfs_inode_destroy(struct inode *ip) { -#ifndef _LP64 - minor_t minor = (minor_t)dev & MAXMIN64; - major_t major = (major_t)(dev >> NBITSMINOR64) & MAXMAJ64; + znode_t *zp = ITOZ(ip); + zfs_sb_t *zsb = ZTOZSB(zp); - if (major > MAXMAJ32 || minor > MAXMIN32) - return (NODEV32); + if (zfsctl_is_node(ip)) + zfsctl_inode_destroy(ip); - return (((dev32_t)major << NBITSMINOR32) | minor); -#else - return (dev); -#endif + mutex_enter(&zsb->z_znodes_lock); + if (list_link_active(&zp->z_link_node)) { + list_remove(&zsb->z_all_znodes, zp); + zsb->z_nr_znodes--; + } + mutex_exit(&zsb->z_znodes_lock); + + if (zp->z_acl_cached) { + zfs_acl_free(zp->z_acl_cached); + zp->z_acl_cached = NULL; + } + + if (zp->z_xattr_cached) { + nvlist_free(zp->z_xattr_cached); + zp->z_xattr_cached = NULL; + } + + if (zp->z_xattr_parent) { + iput(ZTOI(zp->z_xattr_parent)); + zp->z_xattr_parent = NULL; + } + + kmem_cache_free(znode_cache, zp); } static void -zfs_znode_dmu_init(zfsvfs_t *zfsvfs, znode_t *zp, dmu_buf_t *db) +zfs_inode_set_ops(zfs_sb_t *zsb, struct inode *ip) { - znode_t *nzp; + uint64_t rdev = 0; - ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs) || (zfsvfs == zp->z_zfsvfs)); - ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id))); - - mutex_enter(&zp->z_lock); + switch (ip->i_mode & S_IFMT) { + case S_IFREG: + ip->i_op = &zpl_inode_operations; + ip->i_fop = &zpl_file_operations; + ip->i_mapping->a_ops = &zpl_address_space_operations; + break; - ASSERT(zp->z_dbuf == NULL); - zp->z_dbuf = db; - nzp = dmu_buf_set_user_ie(db, zp, &zp->z_phys, znode_evict_error); + case S_IFDIR: + ip->i_op = &zpl_dir_inode_operations; + ip->i_fop = &zpl_dir_file_operations; + ITOZ(ip)->z_zn_prefetch = B_TRUE; + break; - /* - * there should be no - * concurrent zgets on this object. - */ - if (nzp != NULL) - panic("existing znode %p for dbuf %p", (void *)nzp, (void *)db); + case S_IFLNK: + ip->i_op = &zpl_symlink_inode_operations; + break; /* - * Slap on VROOT if we are the root znode + * rdev is only stored in a SA only for device files. */ - if (zp->z_id == zfsvfs->z_root) - ZTOV(zp)->v_flag |= VROOT; - - mutex_exit(&zp->z_lock); - vn_exists(ZTOV(zp)); -} + case S_IFCHR: + case S_IFBLK: + VERIFY(sa_lookup(ITOZ(ip)->z_sa_hdl, SA_ZPL_RDEV(zsb), + &rdev, sizeof (rdev)) == 0); + /*FALLTHROUGH*/ + case S_IFIFO: + case S_IFSOCK: + init_special_inode(ip, ip->i_mode, rdev); + ip->i_op = &zpl_special_inode_operations; + break; -void -zfs_znode_dmu_fini(znode_t *zp) -{ - dmu_buf_t *db = zp->z_dbuf; - ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zp->z_zfsvfs, zp->z_id)) || - zp->z_unlinked || - RW_WRITE_HELD(&zp->z_zfsvfs->z_teardown_inactive_lock)); - ASSERT(zp->z_dbuf != NULL); - zp->z_dbuf = NULL; - VERIFY(zp == dmu_buf_update_user(db, zp, NULL, NULL, NULL)); - dmu_buf_rele(db, NULL); + default: + printk("ZFS: Invalid mode: 0x%x\n", ip->i_mode); + VERIFY(0); + } } /* - * Construct a new znode/vnode and intialize. + * Construct a znode+inode and initialize. * * This does not do a call to dmu_set_user() that is * up to the caller to do, in case you don't want to * return the znode */ static znode_t * -zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz) +zfs_znode_alloc(zfs_sb_t *zsb, dmu_buf_t *db, int blksz, + dmu_object_type_t obj_type, uint64_t obj, sa_handle_t *hdl, + struct inode *dip) { znode_t *zp; - vnode_t *vp; + struct inode *ip; + uint64_t parent; + sa_bulk_attr_t bulk[9]; + int count = 0; - zp = kmem_cache_alloc(znode_cache, KM_SLEEP); + ASSERT(zsb != NULL); - ASSERT(zp->z_dirlocks == NULL); - ASSERT(zp->z_dbuf == NULL); - ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs)); + ip = new_inode(zsb->z_sb); + if (ip == NULL) + return (NULL); - /* - * Defer setting z_zfsvfs until the znode is ready to be a candidate for - * the zfs_znode_move() callback. - */ - zp->z_phys = NULL; + zp = ITOZ(ip); + ASSERT(zp->z_dirlocks == NULL); + ASSERT3P(zp->z_acl_cached, ==, NULL); + ASSERT3P(zp->z_xattr_cached, ==, NULL); + ASSERT3P(zp->z_xattr_parent, ==, NULL); + zp->z_moved = 0; + zp->z_sa_hdl = NULL; zp->z_unlinked = 0; zp->z_atime_dirty = 0; zp->z_mapcnt = 0; - zp->z_last_itx = 0; zp->z_id = db->db_object; zp->z_blksz = blksz; zp->z_seq = 0x7A4653; zp->z_sync_cnt = 0; + zp->z_is_zvol = B_FALSE; + zp->z_is_mapped = B_FALSE; + zp->z_is_ctldir = B_FALSE; + zp->z_is_stale = B_FALSE; + + zfs_znode_sa_init(zsb, zp, db, obj_type, hdl); + + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL, &zp->z_mode, 8); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zsb), NULL, &zp->z_gen, 8); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL, &zp->z_size, 8); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL, &zp->z_links, 8); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL, + &zp->z_pflags, 8); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zsb), NULL, + &parent, 8); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL, + &zp->z_atime, 16); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL, &zp->z_uid, 8); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL, &zp->z_gid, 8); + + if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0) { + if (hdl == NULL) + sa_handle_destroy(zp->z_sa_hdl); + + goto error; + } - vp = ZTOV(zp); - vn_reinit(vp); - - zfs_znode_dmu_init(zfsvfs, zp, db); - - zp->z_gen = zp->z_phys->zp_gen; - - vp->v_vfsp = zfsvfs->z_parent->z_vfs; - vp->v_type = IFTOVT((mode_t)zp->z_phys->zp_mode); - - switch (vp->v_type) { - case VDIR: - if (zp->z_phys->zp_flags & ZFS_XATTR) { - vn_setops(vp, zfs_xdvnodeops); - vp->v_flag |= V_XATTRDIR; - } else { - vn_setops(vp, zfs_dvnodeops); - } - zp->z_zn_prefetch = B_TRUE; /* z_prefetch default is enabled */ - break; - case VBLK: - case VCHR: - vp->v_rdev = zfs_cmpldev(zp->z_phys->zp_rdev); - /*FALLTHROUGH*/ - case VFIFO: - case VSOCK: - case VDOOR: - vn_setops(vp, zfs_fvnodeops); - break; - case VREG: - vp->v_flag |= VMODSORT; - if (zp->z_phys->zp_parent == zfsvfs->z_shares_dir) - vn_setops(vp, zfs_sharevnodeops); - else - vn_setops(vp, zfs_fvnodeops); - break; - case VLNK: - vn_setops(vp, zfs_symvnodeops); - break; - default: - vn_setops(vp, zfs_evnodeops); - break; + /* + * xattr znodes hold a reference on their unique parent + */ + if (dip && zp->z_pflags & ZFS_XATTR) { + igrab(dip); + zp->z_xattr_parent = ITOZ(dip); } - mutex_enter(&zfsvfs->z_znodes_lock); - list_insert_tail(&zfsvfs->z_all_znodes, zp); - membar_producer(); + ip->i_ino = obj; + zfs_inode_update(zp); + zfs_inode_set_ops(zsb, ip); + /* - * Everything else must be valid before assigning z_zfsvfs makes the - * znode eligible for zfs_znode_move(). + * The only way insert_inode_locked() can fail is if the ip->i_ino + * number is already hashed for this super block. This can never + * happen because the inode numbers map 1:1 with the object numbers. + * + * The one exception is rolling back a mounted file system, but in + * this case all the active inode are unhashed during the rollback. */ - zp->z_zfsvfs = zfsvfs; - mutex_exit(&zfsvfs->z_znodes_lock); + VERIFY3S(insert_inode_locked(ip), ==, 0); - VFS_HOLD(zfsvfs->z_vfs); + mutex_enter(&zsb->z_znodes_lock); + list_insert_tail(&zsb->z_all_znodes, zp); + zsb->z_nr_znodes++; + membar_producer(); + mutex_exit(&zsb->z_znodes_lock); + + unlock_new_inode(ip); return (zp); + +error: + unlock_new_inode(ip); + iput(ip); + return NULL; +} + +/* + * Update the embedded inode given the znode. We should work toward + * eliminating this function as soon as possible by removing values + * which are duplicated between the znode and inode. If the generic + * inode has the correct field it should be used, and the ZFS code + * updated to access the inode. This can be done incrementally. + */ +void +zfs_inode_update(znode_t *zp) +{ + zfs_sb_t *zsb; + struct inode *ip; + uint32_t blksize; + uint64_t atime[2], mtime[2], ctime[2]; + + ASSERT(zp != NULL); + zsb = ZTOZSB(zp); + ip = ZTOI(zp); + + /* Skip .zfs control nodes which do not exist on disk. */ + if (zfsctl_is_node(ip)) + return; + + sa_lookup(zp->z_sa_hdl, SA_ZPL_ATIME(zsb), &atime, 16); + sa_lookup(zp->z_sa_hdl, SA_ZPL_MTIME(zsb), &mtime, 16); + sa_lookup(zp->z_sa_hdl, SA_ZPL_CTIME(zsb), &ctime, 16); + + spin_lock(&ip->i_lock); + ip->i_generation = zp->z_gen; + ip->i_uid = SUID_TO_KUID(zp->z_uid); + ip->i_gid = SGID_TO_KGID(zp->z_gid); + set_nlink(ip, zp->z_links); + ip->i_mode = zp->z_mode; + ip->i_blkbits = SPA_MINBLOCKSHIFT; + dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &blksize, + (u_longlong_t *)&ip->i_blocks); + + ZFS_TIME_DECODE(&ip->i_atime, atime); + ZFS_TIME_DECODE(&ip->i_mtime, mtime); + ZFS_TIME_DECODE(&ip->i_ctime, ctime); + + i_size_write(ip, zp->z_size); + spin_unlock(&ip->i_lock); } +static uint64_t empty_xattr; +static uint64_t pad[4]; +static zfs_acl_phys_t acl_phys; /* * Create a new DMU object to hold a zfs znode. * @@ -705,7 +501,6 @@ zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz) * flag - flags: * IS_ROOT_NODE - new object will be root * IS_XATTR - new object is an attribute - * IS_REPLAY - intent log replay * bonuslen - length of bonus buffer * setaclp - File/Dir initial ACL * fuidp - Tracks fuid allocation. @@ -715,20 +510,26 @@ zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz) */ void zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr, - uint_t flag, znode_t **zpp, int bonuslen, zfs_acl_ids_t *acl_ids) + uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids) { + uint64_t crtime[2], atime[2], mtime[2], ctime[2]; + uint64_t mode, size, links, parent, pflags; + uint64_t dzp_pflags = 0; + uint64_t rdev = 0; + zfs_sb_t *zsb = ZTOZSB(dzp); dmu_buf_t *db; - znode_phys_t *pzp; - zfsvfs_t *zfsvfs = dzp->z_zfsvfs; timestruc_t now; uint64_t gen, obj; int err; - - ASSERT(vap && (vap->va_mask & (AT_TYPE|AT_MODE)) == (AT_TYPE|AT_MODE)); - - if (zfsvfs->z_replay) { + int bonuslen; + sa_handle_t *sa_hdl; + dmu_object_type_t obj_type; + sa_bulk_attr_t *sa_attrs; + int cnt = 0; + zfs_acl_locator_cb_t locate = { 0 }; + + if (zsb->z_replay) { obj = vap->va_nodeid; - flag |= IS_REPLAY; now = vap->va_ctime; /* see zfs_replay_create() */ gen = vap->va_nblocks; /* ditto */ } else { @@ -737,119 +538,228 @@ zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr, gen = dmu_tx_get_txg(tx); } + obj_type = zsb->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE; + bonuslen = (obj_type == DMU_OT_SA) ? + DN_MAX_BONUSLEN : ZFS_OLD_ZNODE_PHYS_SIZE; + /* * Create a new DMU object. */ /* * There's currently no mechanism for pre-reading the blocks that will - * be to needed allocate a new object, so we accept the small chance + * be needed to allocate a new object, so we accept the small chance * that there will be an i/o error and we will fail one of the * assertions below. */ - if (vap->va_type == VDIR) { - if (flag & IS_REPLAY) { - err = zap_create_claim_norm(zfsvfs->z_os, obj, - zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS, - DMU_OT_ZNODE, sizeof (znode_phys_t) + bonuslen, tx); - ASSERT3U(err, ==, 0); + if (S_ISDIR(vap->va_mode)) { + if (zsb->z_replay) { + err = zap_create_claim_norm(zsb->z_os, obj, + zsb->z_norm, DMU_OT_DIRECTORY_CONTENTS, + obj_type, bonuslen, tx); + ASSERT0(err); } else { - obj = zap_create_norm(zfsvfs->z_os, - zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS, - DMU_OT_ZNODE, sizeof (znode_phys_t) + bonuslen, tx); + obj = zap_create_norm(zsb->z_os, + zsb->z_norm, DMU_OT_DIRECTORY_CONTENTS, + obj_type, bonuslen, tx); } } else { - if (flag & IS_REPLAY) { - err = dmu_object_claim(zfsvfs->z_os, obj, + if (zsb->z_replay) { + err = dmu_object_claim(zsb->z_os, obj, DMU_OT_PLAIN_FILE_CONTENTS, 0, - DMU_OT_ZNODE, sizeof (znode_phys_t) + bonuslen, tx); - ASSERT3U(err, ==, 0); + obj_type, bonuslen, tx); + ASSERT0(err); } else { - obj = dmu_object_alloc(zfsvfs->z_os, + obj = dmu_object_alloc(zsb->z_os, DMU_OT_PLAIN_FILE_CONTENTS, 0, - DMU_OT_ZNODE, sizeof (znode_phys_t) + bonuslen, tx); + obj_type, bonuslen, tx); } } - VERIFY(0 == dmu_bonus_hold(zfsvfs->z_os, obj, NULL, &db)); - dmu_buf_will_dirty(db, tx); - /* - * Initialize the znode physical data to zero. - */ - ASSERT(db->db_size >= sizeof (znode_phys_t)); - bzero(db->db_data, db->db_size); - pzp = db->db_data; + ZFS_OBJ_HOLD_ENTER(zsb, obj); + VERIFY(0 == sa_buf_hold(zsb->z_os, obj, NULL, &db)); /* * If this is the root, fix up the half-initialized parent pointer * to reference the just-allocated physical data area. */ if (flag & IS_ROOT_NODE) { - dzp->z_dbuf = db; - dzp->z_phys = pzp; dzp->z_id = obj; + } else { + dzp_pflags = dzp->z_pflags; } /* * If parent is an xattr, so am I. */ - if (dzp->z_phys->zp_flags & ZFS_XATTR) + if (dzp_pflags & ZFS_XATTR) { flag |= IS_XATTR; - - if (vap->va_type == VBLK || vap->va_type == VCHR) { - pzp->zp_rdev = zfs_expldev(vap->va_rdev); } - if (zfsvfs->z_use_fuids) - pzp->zp_flags = ZFS_ARCHIVE | ZFS_AV_MODIFIED; + if (zsb->z_use_fuids) + pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED; + else + pflags = 0; - if (vap->va_type == VDIR) { - pzp->zp_size = 2; /* contents ("." and "..") */ - pzp->zp_links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1; + if (S_ISDIR(vap->va_mode)) { + size = 2; /* contents ("." and "..") */ + links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1; + } else { + size = links = 0; } - pzp->zp_parent = dzp->z_id; + if (S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode)) + rdev = vap->va_rdev; + + parent = dzp->z_id; + mode = acl_ids->z_mode; if (flag & IS_XATTR) - pzp->zp_flags |= ZFS_XATTR; + pflags |= ZFS_XATTR; + + /* + * No execs denied will be deterimed when zfs_mode_compute() is called. + */ + pflags |= acl_ids->z_aclp->z_hints & + (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT| + ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED); - pzp->zp_gen = gen; + ZFS_TIME_ENCODE(&now, crtime); + ZFS_TIME_ENCODE(&now, ctime); - ZFS_TIME_ENCODE(&now, pzp->zp_crtime); - ZFS_TIME_ENCODE(&now, pzp->zp_ctime); + if (vap->va_mask & ATTR_ATIME) { + ZFS_TIME_ENCODE(&vap->va_atime, atime); + } else { + ZFS_TIME_ENCODE(&now, atime); + } - if (vap->va_mask & AT_ATIME) { - ZFS_TIME_ENCODE(&vap->va_atime, pzp->zp_atime); + if (vap->va_mask & ATTR_MTIME) { + ZFS_TIME_ENCODE(&vap->va_mtime, mtime); } else { - ZFS_TIME_ENCODE(&now, pzp->zp_atime); + ZFS_TIME_ENCODE(&now, mtime); } - if (vap->va_mask & AT_MTIME) { - ZFS_TIME_ENCODE(&vap->va_mtime, pzp->zp_mtime); + /* Now add in all of the "SA" attributes */ + VERIFY(0 == sa_handle_get_from_db(zsb->z_os, db, NULL, SA_HDL_SHARED, + &sa_hdl)); + + /* + * Setup the array of attributes to be replaced/set on the new file + * + * order for DMU_OT_ZNODE is critical since it needs to be constructed + * in the old znode_phys_t format. Don't change this ordering + */ + sa_attrs = kmem_alloc(sizeof(sa_bulk_attr_t) * ZPL_END, KM_PUSHPAGE); + + if (obj_type == DMU_OT_ZNODE) { + SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zsb), + NULL, &atime, 16); + SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zsb), + NULL, &mtime, 16); + SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zsb), + NULL, &ctime, 16); + SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zsb), + NULL, &crtime, 16); + SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zsb), + NULL, &gen, 8); + SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zsb), + NULL, &mode, 8); + SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zsb), + NULL, &size, 8); + SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zsb), + NULL, &parent, 8); } else { - ZFS_TIME_ENCODE(&now, pzp->zp_mtime); + SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zsb), + NULL, &mode, 8); + SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zsb), + NULL, &size, 8); + SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zsb), + NULL, &gen, 8); + SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zsb), + NULL, &acl_ids->z_fuid, 8); + SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zsb), + NULL, &acl_ids->z_fgid, 8); + SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zsb), + NULL, &parent, 8); + SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zsb), + NULL, &pflags, 8); + SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zsb), + NULL, &atime, 16); + SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zsb), + NULL, &mtime, 16); + SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zsb), + NULL, &ctime, 16); + SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zsb), + NULL, &crtime, 16); + } + + SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zsb), NULL, &links, 8); + + if (obj_type == DMU_OT_ZNODE) { + SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zsb), NULL, + &empty_xattr, 8); + } + if (obj_type == DMU_OT_ZNODE || + (S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode))) { + SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zsb), + NULL, &rdev, 8); } + if (obj_type == DMU_OT_ZNODE) { + SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zsb), + NULL, &pflags, 8); + SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zsb), NULL, + &acl_ids->z_fuid, 8); + SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zsb), NULL, + &acl_ids->z_fgid, 8); + SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zsb), NULL, pad, + sizeof (uint64_t) * 4); + SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zsb), NULL, + &acl_phys, sizeof (zfs_acl_phys_t)); + } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) { + SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zsb), NULL, + &acl_ids->z_aclp->z_acl_count, 8); + locate.cb_aclp = acl_ids->z_aclp; + SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zsb), + zfs_acl_data_locator, &locate, + acl_ids->z_aclp->z_acl_bytes); + mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags, + acl_ids->z_fuid, acl_ids->z_fgid); + } + + VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0); - pzp->zp_mode = MAKEIMODE(vap->va_type, vap->va_mode); if (!(flag & IS_ROOT_NODE)) { - ZFS_OBJ_HOLD_ENTER(zfsvfs, obj); - *zpp = zfs_znode_alloc(zfsvfs, db, 0); - ZFS_OBJ_HOLD_EXIT(zfsvfs, obj); + *zpp = zfs_znode_alloc(zsb, db, 0, obj_type, obj, sa_hdl, + ZTOI(dzp)); + VERIFY(*zpp != NULL); + VERIFY(dzp != NULL); } else { /* * If we are creating the root node, the "parent" we * passed in is the znode for the root. */ *zpp = dzp; + + (*zpp)->z_sa_hdl = sa_hdl; + } + + (*zpp)->z_pflags = pflags; + (*zpp)->z_mode = mode; + + if (obj_type == DMU_OT_ZNODE || + acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) { + err = zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx); + ASSERT0(err); } - pzp->zp_uid = acl_ids->z_fuid; - pzp->zp_gid = acl_ids->z_fgid; - pzp->zp_mode = acl_ids->z_mode; - VERIFY(0 == zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx)); - if (vap->va_mask & AT_XVATTR) - zfs_xvattr_set(*zpp, (xvattr_t *)vap); + kmem_free(sa_attrs, sizeof(sa_bulk_attr_t) * ZPL_END); + ZFS_OBJ_HOLD_EXIT(zsb, obj); } +/* + * zfs_xvattr_set only updates the in-core attributes + * it is assumed the caller will be doing an sa_bulk_update + * to push the changes out + */ void -zfs_xvattr_set(znode_t *zp, xvattr_t *xvap) +zfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx) { xoptattr_t *xoap; @@ -857,157 +767,278 @@ zfs_xvattr_set(znode_t *zp, xvattr_t *xvap) ASSERT(xoap); if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) { - ZFS_TIME_ENCODE(&xoap->xoa_createtime, zp->z_phys->zp_crtime); + uint64_t times[2]; + ZFS_TIME_ENCODE(&xoap->xoa_createtime, times); + (void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(ZTOZSB(zp)), + ×, sizeof (times), tx); XVA_SET_RTN(xvap, XAT_CREATETIME); } if (XVA_ISSET_REQ(xvap, XAT_READONLY)) { - ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly); + ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly, + zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_READONLY); } if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) { - ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden); + ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden, + zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_HIDDEN); } if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) { - ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system); + ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system, + zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_SYSTEM); } if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) { - ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive); + ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive, + zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_ARCHIVE); } if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { - ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable); + ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable, + zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_IMMUTABLE); } if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { - ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink); + ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink, + zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_NOUNLINK); } if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { - ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly); + ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly, + zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_APPENDONLY); } if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { - ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump); + ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump, + zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_NODUMP); } if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) { - ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque); + ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque, + zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_OPAQUE); } if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED, - xoap->xoa_av_quarantined); + xoap->xoa_av_quarantined, zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_AV_QUARANTINED); } if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { - ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified); + ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified, + zp->z_pflags, tx); XVA_SET_RTN(xvap, XAT_AV_MODIFIED); } if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) { - (void) memcpy(zp->z_phys + 1, xoap->xoa_av_scanstamp, - sizeof (xoap->xoa_av_scanstamp)); - zp->z_phys->zp_flags |= ZFS_BONUS_SCANSTAMP; + zfs_sa_set_scanstamp(zp, xvap, tx); XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP); } + if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { + ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse, + zp->z_pflags, tx); + XVA_SET_RTN(xvap, XAT_REPARSE); + } + if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) { + ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline, + zp->z_pflags, tx); + XVA_SET_RTN(xvap, XAT_OFFLINE); + } + if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) { + ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse, + zp->z_pflags, tx); + XVA_SET_RTN(xvap, XAT_SPARSE); + } } int -zfs_zget(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp) +zfs_zget(zfs_sb_t *zsb, uint64_t obj_num, znode_t **zpp) { dmu_object_info_t doi; dmu_buf_t *db; znode_t *zp; int err; + sa_handle_t *hdl; + struct inode *ip; *zpp = NULL; - ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num); +again: + ip = ilookup(zsb->z_sb, obj_num); + + ZFS_OBJ_HOLD_ENTER(zsb, obj_num); - err = dmu_bonus_hold(zfsvfs->z_os, obj_num, NULL, &db); + err = sa_buf_hold(zsb->z_os, obj_num, NULL, &db); if (err) { - ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); + ZFS_OBJ_HOLD_EXIT(zsb, obj_num); + iput(ip); return (err); } dmu_object_info_from_db(db, &doi); - if (doi.doi_bonus_type != DMU_OT_ZNODE || - doi.doi_bonus_size < sizeof (znode_phys_t)) { - dmu_buf_rele(db, NULL); - ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); + if (doi.doi_bonus_type != DMU_OT_SA && + (doi.doi_bonus_type != DMU_OT_ZNODE || + (doi.doi_bonus_type == DMU_OT_ZNODE && + doi.doi_bonus_size < sizeof (znode_phys_t)))) { + sa_buf_rele(db, NULL); + ZFS_OBJ_HOLD_EXIT(zsb, obj_num); + iput(ip); return (EINVAL); } - zp = dmu_buf_get_user(db); - if (zp != NULL) { - mutex_enter(&zp->z_lock); + hdl = dmu_buf_get_user(db); + if (hdl != NULL) { + if (ip == NULL) { + /* + * ilookup returned NULL, which means + * the znode is dying - but the SA handle isn't + * quite dead yet, we need to drop any locks + * we're holding, re-schedule the task and try again. + */ + sa_buf_rele(db, NULL); + ZFS_OBJ_HOLD_EXIT(zsb, obj_num); + + schedule(); + goto again; + } + + zp = sa_get_userdata(hdl); /* - * Since we do immediate eviction of the z_dbuf, we - * should never find a dbuf with a znode that doesn't - * know about the dbuf. + * Since "SA" does immediate eviction we + * should never find a sa handle that doesn't + * know about the znode. */ - ASSERT3P(zp->z_dbuf, ==, db); + + ASSERT3P(zp, !=, NULL); + + mutex_enter(&zp->z_lock); ASSERT3U(zp->z_id, ==, obj_num); if (zp->z_unlinked) { err = ENOENT; } else { - VN_HOLD(ZTOV(zp)); + igrab(ZTOI(zp)); *zpp = zp; err = 0; } - dmu_buf_rele(db, NULL); + sa_buf_rele(db, NULL); mutex_exit(&zp->z_lock); - ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); + ZFS_OBJ_HOLD_EXIT(zsb, obj_num); + iput(ip); return (err); } + ASSERT3P(ip, ==, NULL); + /* - * Not found create new znode/vnode + * Not found create new znode/vnode but only if file exists. + * + * There is a small window where zfs_vget() could + * find this object while a file create is still in + * progress. This is checked for in zfs_znode_alloc() + * + * if zfs_znode_alloc() fails it will drop the hold on the + * bonus buffer. */ - zp = zfs_znode_alloc(zfsvfs, db, doi.doi_data_block_size); - ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); - *zpp = zp; - return (0); + zp = zfs_znode_alloc(zsb, db, doi.doi_data_block_size, + doi.doi_bonus_type, obj_num, NULL, NULL); + if (zp == NULL) { + err = ENOENT; + } else { + *zpp = zp; + } + ZFS_OBJ_HOLD_EXIT(zsb, obj_num); + return (err); } int zfs_rezget(znode_t *zp) { - zfsvfs_t *zfsvfs = zp->z_zfsvfs; + zfs_sb_t *zsb = ZTOZSB(zp); dmu_object_info_t doi; dmu_buf_t *db; uint64_t obj_num = zp->z_id; + uint64_t mode; + sa_bulk_attr_t bulk[8]; int err; + int count = 0; + uint64_t gen; - ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num); + ZFS_OBJ_HOLD_ENTER(zsb, obj_num); - err = dmu_bonus_hold(zfsvfs->z_os, obj_num, NULL, &db); + mutex_enter(&zp->z_acl_lock); + if (zp->z_acl_cached) { + zfs_acl_free(zp->z_acl_cached); + zp->z_acl_cached = NULL; + } + mutex_exit(&zp->z_acl_lock); + + rw_enter(&zp->z_xattr_lock, RW_WRITER); + if (zp->z_xattr_cached) { + nvlist_free(zp->z_xattr_cached); + zp->z_xattr_cached = NULL; + } + + if (zp->z_xattr_parent) { + iput(ZTOI(zp->z_xattr_parent)); + zp->z_xattr_parent = NULL; + } + rw_exit(&zp->z_xattr_lock); + + ASSERT(zp->z_sa_hdl == NULL); + err = sa_buf_hold(zsb->z_os, obj_num, NULL, &db); if (err) { - ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); + ZFS_OBJ_HOLD_EXIT(zsb, obj_num); return (err); } dmu_object_info_from_db(db, &doi); - if (doi.doi_bonus_type != DMU_OT_ZNODE || - doi.doi_bonus_size < sizeof (znode_phys_t)) { - dmu_buf_rele(db, NULL); - ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); + if (doi.doi_bonus_type != DMU_OT_SA && + (doi.doi_bonus_type != DMU_OT_ZNODE || + (doi.doi_bonus_type == DMU_OT_ZNODE && + doi.doi_bonus_size < sizeof (znode_phys_t)))) { + sa_buf_rele(db, NULL); + ZFS_OBJ_HOLD_EXIT(zsb, obj_num); return (EINVAL); } - if (((znode_phys_t *)db->db_data)->zp_gen != zp->z_gen) { - dmu_buf_rele(db, NULL); - ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); + zfs_znode_sa_init(zsb, zp, db, doi.doi_bonus_type, NULL); + + /* reload cached values */ + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zsb), NULL, + &gen, sizeof (gen)); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL, + &zp->z_size, sizeof (zp->z_size)); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL, + &zp->z_links, sizeof (zp->z_links)); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL, + &zp->z_pflags, sizeof (zp->z_pflags)); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL, + &zp->z_atime, sizeof (zp->z_atime)); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL, + &zp->z_uid, sizeof (zp->z_uid)); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL, + &zp->z_gid, sizeof (zp->z_gid)); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL, + &mode, sizeof (mode)); + + if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) { + zfs_znode_dmu_fini(zp); + ZFS_OBJ_HOLD_EXIT(zsb, obj_num); + return (EIO); + } + + zp->z_mode = mode; + + if (gen != zp->z_gen) { + zfs_znode_dmu_fini(zp); + ZFS_OBJ_HOLD_EXIT(zsb, obj_num); return (EIO); } - zfs_znode_dmu_init(zfsvfs, zp, db); - zp->z_unlinked = (zp->z_phys->zp_links == 0); + zp->z_unlinked = (zp->z_links == 0); zp->z_blksz = doi.doi_data_block_size; + zfs_inode_update(zp); - ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); + ZFS_OBJ_HOLD_EXIT(zsb, obj_num); return (0); } @@ -1015,55 +1046,45 @@ zfs_rezget(znode_t *zp) void zfs_znode_delete(znode_t *zp, dmu_tx_t *tx) { - zfsvfs_t *zfsvfs = zp->z_zfsvfs; - objset_t *os = zfsvfs->z_os; + zfs_sb_t *zsb = ZTOZSB(zp); + objset_t *os = zsb->z_os; uint64_t obj = zp->z_id; - uint64_t acl_obj = zp->z_phys->zp_acl.z_acl_extern_obj; + uint64_t acl_obj = zfs_external_acl(zp); - ZFS_OBJ_HOLD_ENTER(zfsvfs, obj); - if (acl_obj) + ZFS_OBJ_HOLD_ENTER(zsb, obj); + if (acl_obj) { + VERIFY(!zp->z_is_sa); VERIFY(0 == dmu_object_free(os, acl_obj, tx)); + } VERIFY(0 == dmu_object_free(os, obj, tx)); zfs_znode_dmu_fini(zp); - ZFS_OBJ_HOLD_EXIT(zfsvfs, obj); - zfs_znode_free(zp); + ZFS_OBJ_HOLD_EXIT(zsb, obj); } void zfs_zinactive(znode_t *zp) { - vnode_t *vp = ZTOV(zp); - zfsvfs_t *zfsvfs = zp->z_zfsvfs; + zfs_sb_t *zsb = ZTOZSB(zp); uint64_t z_id = zp->z_id; + boolean_t drop_mutex = 0; - ASSERT(zp->z_dbuf && zp->z_phys); + ASSERT(zp->z_sa_hdl); /* - * Don't allow a zfs_zget() while were trying to release this znode + * Don't allow a zfs_zget() while were trying to release this znode. + * + * Linux allows direct memory reclaim which means that any KM_SLEEP + * allocation may trigger inode eviction. This can lead to a deadlock + * through the ->shrink_icache_memory()->evict()->zfs_inactive()-> + * zfs_zinactive() call path. To avoid this deadlock the process + * must not reacquire the mutex when it is already holding it. */ - ZFS_OBJ_HOLD_ENTER(zfsvfs, z_id); + if (!ZFS_OBJ_HOLD_OWNED(zsb, z_id)) { + ZFS_OBJ_HOLD_ENTER(zsb, z_id); + drop_mutex = 1; + } mutex_enter(&zp->z_lock); - mutex_enter(&vp->v_lock); - vp->v_count--; - if (vp->v_count > 0 || vn_has_cached_data(vp)) { - /* - * If the hold count is greater than zero, somebody has - * obtained a new reference on this znode while we were - * processing it here, so we are done. If we still have - * mapped pages then we are also done, since we don't - * want to inactivate the znode until the pages get pushed. - * - * XXX - if vn_has_cached_data(vp) is true, but count == 0, - * this seems like it would leave the znode hanging with - * no chance to go inactive... - */ - mutex_exit(&vp->v_lock); - mutex_exit(&zp->z_lock); - ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id); - return; - } - mutex_exit(&vp->v_lock); /* * If this was the last reference to a file with no links, @@ -1071,91 +1092,53 @@ zfs_zinactive(znode_t *zp) */ if (zp->z_unlinked) { mutex_exit(&zp->z_lock); - ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id); + + if (drop_mutex) + ZFS_OBJ_HOLD_EXIT(zsb, z_id); + zfs_rmnode(zp); return; } + mutex_exit(&zp->z_lock); zfs_znode_dmu_fini(zp); - ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id); - zfs_znode_free(zp); -} - -void -zfs_znode_free(znode_t *zp) -{ - zfsvfs_t *zfsvfs = zp->z_zfsvfs; - - vn_invalid(ZTOV(zp)); - - ASSERT(ZTOV(zp)->v_count == 0); - - mutex_enter(&zfsvfs->z_znodes_lock); - POINTER_INVALIDATE(&zp->z_zfsvfs); - list_remove(&zfsvfs->z_all_znodes, zp); - mutex_exit(&zfsvfs->z_znodes_lock); - - if (zp->z_acl_cached) { - zfs_acl_free(zp->z_acl_cached); - zp->z_acl_cached = NULL; - } - kmem_cache_free(znode_cache, zp); - - VFS_RELE(zfsvfs->z_vfs); + if (drop_mutex) + ZFS_OBJ_HOLD_EXIT(zsb, z_id); } void -zfs_time_stamper_locked(znode_t *zp, uint_t flag, dmu_tx_t *tx) +zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2], + uint64_t ctime[2], boolean_t have_tx) { timestruc_t now; - ASSERT(MUTEX_HELD(&zp->z_lock)); - gethrestime(&now); - if (tx) { - dmu_buf_will_dirty(zp->z_dbuf, tx); + if (have_tx) { /* will sa_bulk_update happen really soon? */ zp->z_atime_dirty = 0; zp->z_seq++; } else { zp->z_atime_dirty = 1; } - if (flag & AT_ATIME) - ZFS_TIME_ENCODE(&now, zp->z_phys->zp_atime); - - if (flag & AT_MTIME) { - ZFS_TIME_ENCODE(&now, zp->z_phys->zp_mtime); - if (zp->z_zfsvfs->z_use_fuids) - zp->z_phys->zp_flags |= (ZFS_ARCHIVE | ZFS_AV_MODIFIED); + if (flag & ATTR_ATIME) { + ZFS_TIME_ENCODE(&now, zp->z_atime); } - if (flag & AT_CTIME) { - ZFS_TIME_ENCODE(&now, zp->z_phys->zp_ctime); - if (zp->z_zfsvfs->z_use_fuids) - zp->z_phys->zp_flags |= ZFS_ARCHIVE; + if (flag & ATTR_MTIME) { + ZFS_TIME_ENCODE(&now, mtime); + if (ZTOZSB(zp)->z_use_fuids) { + zp->z_pflags |= (ZFS_ARCHIVE | + ZFS_AV_MODIFIED); + } } -} -/* - * Update the requested znode timestamps with the current time. - * If we are in a transaction, then go ahead and mark the znode - * dirty in the transaction so the timestamps will go to disk. - * Otherwise, we will get pushed next time the znode is updated - * in a transaction, or when this znode eventually goes inactive. - * - * Why is this OK? - * 1 - Only the ACCESS time is ever updated outside of a transaction. - * 2 - Multiple consecutive updates will be collapsed into a single - * znode update by the transaction grouping semantics of the DMU. - */ -void -zfs_time_stamper(znode_t *zp, uint_t flag, dmu_tx_t *tx) -{ - mutex_enter(&zp->z_lock); - zfs_time_stamper_locked(zp, flag, tx); - mutex_exit(&zp->z_lock); + if (flag & ATTR_CTIME) { + ZFS_TIME_ENCODE(&now, ctime); + if (ZTOZSB(zp)->z_use_fuids) + zp->z_pflags |= ZFS_ARCHIVE; + } } /* @@ -1180,31 +1163,18 @@ zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx) * we will not grow. If there is more than one block in a file, * the blocksize cannot change. */ - if (zp->z_blksz && zp->z_phys->zp_size > zp->z_blksz) + if (zp->z_blksz && zp->z_size > zp->z_blksz) return; - error = dmu_object_set_blocksize(zp->z_zfsvfs->z_os, zp->z_id, + error = dmu_object_set_blocksize(ZTOZSB(zp)->z_os, zp->z_id, size, 0, tx); + if (error == ENOTSUP) return; - ASSERT3U(error, ==, 0); + ASSERT0(error); /* What blocksize did we actually get? */ - dmu_object_size_from_db(zp->z_dbuf, &zp->z_blksz, &dummy); -} - -/* - * This is a dummy interface used when pvn_vplist_dirty() should *not* - * be calling back into the fs for a putpage(). E.g.: when truncating - * a file, the pages being "thrown away* don't need to be written out. - */ -/* ARGSUSED */ -static int -zfs_no_putpage(vnode_t *vp, page_t *pp, u_offset_t *offp, size_t *lenp, - int flags, cred_t *cr) -{ - ASSERT(0); - return (0); + dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy); } /* @@ -1219,7 +1189,7 @@ zfs_no_putpage(vnode_t *vp, page_t *pp, u_offset_t *offp, size_t *lenp, static int zfs_extend(znode_t *zp, uint64_t end) { - zfsvfs_t *zfsvfs = zp->z_zfsvfs; + zfs_sb_t *zsb = ZTOZSB(zp); dmu_tx_t *tx; rl_t *rl; uint64_t newblksz; @@ -1233,23 +1203,24 @@ zfs_extend(znode_t *zp, uint64_t end) /* * Nothing to do if file already at desired length. */ - if (end <= zp->z_phys->zp_size) { + if (end <= zp->z_size) { zfs_range_unlock(rl); return (0); } top: - tx = dmu_tx_create(zfsvfs->z_os); - dmu_tx_hold_bonus(tx, zp->z_id); + tx = dmu_tx_create(zsb->z_os); + dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); + zfs_sa_upgrade_txholds(tx, zp); if (end > zp->z_blksz && - (!ISP2(zp->z_blksz) || zp->z_blksz < zfsvfs->z_max_blksz)) { + (!ISP2(zp->z_blksz) || zp->z_blksz < zsb->z_max_blksz)) { /* * We are growing the file past the current block size. */ - if (zp->z_blksz > zp->z_zfsvfs->z_max_blksz) { + if (zp->z_blksz > ZTOZSB(zp)->z_max_blksz) { ASSERT(!ISP2(zp->z_blksz)); newblksz = MIN(end, SPA_MAXBLOCKSIZE); } else { - newblksz = MIN(end, zp->z_zfsvfs->z_max_blksz); + newblksz = MIN(end, ZTOZSB(zp)->z_max_blksz); } dmu_tx_hold_write(tx, zp->z_id, 0, newblksz); } else { @@ -1267,12 +1238,14 @@ top: zfs_range_unlock(rl); return (error); } - dmu_buf_will_dirty(zp->z_dbuf, tx); if (newblksz) zfs_grow_blocksize(zp, newblksz, tx); - zp->z_phys->zp_size = end; + zp->z_size = end; + + VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(ZTOZSB(zp)), + &zp->z_size, sizeof (zp->z_size), tx)); zfs_range_unlock(rl); @@ -1288,13 +1261,13 @@ top: * off - start of section to free. * len - length of section to free. * - * RETURN: 0 if success + * RETURN: 0 if success * error code if failure */ static int zfs_free_range(znode_t *zp, uint64_t off, uint64_t len) { - zfsvfs_t *zfsvfs = zp->z_zfsvfs; + zfs_sb_t *zsb = ZTOZSB(zp); rl_t *rl; int error; @@ -1306,15 +1279,15 @@ zfs_free_range(znode_t *zp, uint64_t off, uint64_t len) /* * Nothing to do if file already at desired length. */ - if (off >= zp->z_phys->zp_size) { + if (off >= zp->z_size) { zfs_range_unlock(rl); return (0); } - if (off + len > zp->z_phys->zp_size) - len = zp->z_phys->zp_size - off; + if (off + len > zp->z_size) + len = zp->z_size - off; - error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, off, len); + error = dmu_free_long_range(zsb->z_os, zp->z_id, off, len); zfs_range_unlock(rl); @@ -1327,17 +1300,18 @@ zfs_free_range(znode_t *zp, uint64_t off, uint64_t len) * IN: zp - znode of file to free data in. * end - new end-of-file. * - * RETURN: 0 if success + * RETURN: 0 if success * error code if failure */ static int zfs_trunc(znode_t *zp, uint64_t end) { - zfsvfs_t *zfsvfs = zp->z_zfsvfs; - vnode_t *vp = ZTOV(zp); + zfs_sb_t *zsb = ZTOZSB(zp); dmu_tx_t *tx; rl_t *rl; int error; + sa_bulk_attr_t bulk[2]; + int count = 0; /* * We will change zp_size, lock the whole file. @@ -1347,19 +1321,20 @@ zfs_trunc(znode_t *zp, uint64_t end) /* * Nothing to do if file already at desired length. */ - if (end >= zp->z_phys->zp_size) { + if (end >= zp->z_size) { zfs_range_unlock(rl); return (0); } - error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, end, -1); + error = dmu_free_long_range(zsb->z_os, zp->z_id, end, -1); if (error) { zfs_range_unlock(rl); return (error); } top: - tx = dmu_tx_create(zfsvfs->z_os); - dmu_tx_hold_bonus(tx, zp->z_id); + tx = dmu_tx_create(zsb->z_os); + dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); + zfs_sa_upgrade_txholds(tx, zp); error = dmu_tx_assign(tx, TXG_NOWAIT); if (error) { if (error == ERESTART) { @@ -1371,35 +1346,19 @@ top: zfs_range_unlock(rl); return (error); } - dmu_buf_will_dirty(zp->z_dbuf, tx); - zp->z_phys->zp_size = end; + zp->z_size = end; + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), + NULL, &zp->z_size, sizeof (zp->z_size)); - dmu_tx_commit(tx); - - /* - * Clear any mapped pages in the truncated region. This has to - * happen outside of the transaction to avoid the possibility of - * a deadlock with someone trying to push a page that we are - * about to invalidate. - */ - if (vn_has_cached_data(vp)) { - page_t *pp; - uint64_t start = end & PAGEMASK; - int poff = end & PAGEOFFSET; - - if (poff != 0 && (pp = page_lookup(vp, start, SE_SHARED))) { - /* - * We need to zero a partial page. - */ - pagezero(pp, poff, PAGESIZE - poff); - start += PAGESIZE; - page_unlock(pp); - } - error = pvn_vplist_dirty(vp, start, zfs_no_putpage, - B_INVAL | B_TRUNC, NULL); - ASSERT(error == 0); + if (end == 0) { + zp->z_pflags &= ~ZFS_SPARSE; + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), + NULL, &zp->z_pflags, 8); } + VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0); + + dmu_tx_commit(tx); zfs_range_unlock(rl); @@ -1415,19 +1374,27 @@ top: * flag - current file open mode flags. * log - TRUE if this action should be logged * - * RETURN: 0 if success + * RETURN: 0 if success * error code if failure */ int zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log) { - vnode_t *vp = ZTOV(zp); + struct inode *ip = ZTOI(zp); dmu_tx_t *tx; - zfsvfs_t *zfsvfs = zp->z_zfsvfs; - zilog_t *zilog = zfsvfs->z_log; + zfs_sb_t *zsb = ZTOZSB(zp); + zilog_t *zilog = zsb->z_log; + uint64_t mode; + uint64_t mtime[2], ctime[2]; + sa_bulk_attr_t bulk[3]; + int count = 0; int error; - if (off > zp->z_phys->zp_size) { + if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zsb), &mode, + sizeof (mode))) != 0) + return (error); + + if (off > zp->z_size) { error = zfs_extend(zp, off+len); if (error == 0 && log) goto log; @@ -1438,24 +1405,25 @@ zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log) /* * Check for any locks in the region to be freed. */ - if (MANDLOCK(vp, (mode_t)zp->z_phys->zp_mode)) { - uint64_t length = (len ? len : zp->z_phys->zp_size - off); - if (error = chklock(vp, FWRITE, off, length, flag, NULL)) - return (error); + if (ip->i_flock && mandatory_lock(ip)) { + uint64_t length = (len ? len : zp->z_size - off); + if (!lock_may_write(ip, off, length)) + return (EAGAIN); } if (len == 0) { error = zfs_trunc(zp, off); } else { if ((error = zfs_free_range(zp, off, len)) == 0 && - off + len > zp->z_phys->zp_size) + off + len > zp->z_size) error = zfs_extend(zp, off+len); } if (error || !log) return (error); log: - tx = dmu_tx_create(zfsvfs->z_os); - dmu_tx_hold_bonus(tx, zp->z_id); + tx = dmu_tx_create(zsb->z_os); + dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); + zfs_sa_upgrade_txholds(tx, zp); error = dmu_tx_assign(tx, TXG_NOWAIT); if (error) { if (error == ERESTART) { @@ -1467,24 +1435,33 @@ log: return (error); } - zfs_time_stamper(zp, CONTENT_MODIFIED, tx); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, mtime, 16); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, ctime, 16); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), + NULL, &zp->z_pflags, 8); + zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, B_TRUE); + error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); + ASSERT(error == 0); + zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len); dmu_tx_commit(tx); + zfs_inode_update(zp); return (0); } void zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx) { - zfsvfs_t zfsvfs; - uint64_t moid, obj, version; + struct super_block *sb; + zfs_sb_t *zsb; + uint64_t moid, obj, sa_obj, version; uint64_t sense = ZFS_CASE_SENSITIVE; uint64_t norm = 0; nvpair_t *elem; int error; + int i; znode_t *rootzp = NULL; - vnode_t *vp; vattr_t vattr; znode_t *zp; zfs_acl_ids_t acl_ids; @@ -1504,12 +1481,7 @@ zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx) /* * Set starting attributes. */ - if (spa_version(dmu_objset_spa(os)) >= SPA_VERSION_USERSPACE) - version = ZPL_VERSION; - else if (spa_version(dmu_objset_spa(os)) >= SPA_VERSION_FUID) - version = ZPL_VERSION_USERSPACE - 1; - else - version = ZPL_VERSION_FUID - 1; + version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os))); elem = NULL; while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) { /* For the moment we expect all zpl props to be uint64_ts */ @@ -1535,6 +1507,18 @@ zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx) error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx); /* + * Create zap object used for SA attribute registration + */ + + if (version >= ZPL_VERSION_SA) { + sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE, + DMU_OT_NONE, 0, tx); + error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx); + ASSERT(error == 0); + } else { + sa_obj = 0; + } + /* * Create a delete queue. */ obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx); @@ -1543,106 +1527,193 @@ zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx) ASSERT(error == 0); /* - * Create root znode. Create minimal znode/vnode/zfsvfs + * Create root znode. Create minimal znode/inode/zsb/sb * to allow zfs_mknode to work. */ - vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE; - vattr.va_type = VDIR; + vattr.va_mask = ATTR_MODE|ATTR_UID|ATTR_GID; vattr.va_mode = S_IFDIR|0755; vattr.va_uid = crgetuid(cr); vattr.va_gid = crgetgid(cr); - rootzp = kmem_cache_alloc(znode_cache, KM_SLEEP); + rootzp = kmem_cache_alloc(znode_cache, KM_PUSHPAGE); + rootzp->z_moved = 0; rootzp->z_unlinked = 0; rootzp->z_atime_dirty = 0; + rootzp->z_is_sa = USE_SA(version, os); - vp = ZTOV(rootzp); - vn_reinit(vp); - vp->v_type = VDIR; + zsb = kmem_zalloc(sizeof (zfs_sb_t), KM_PUSHPAGE | KM_NODEBUG); + zsb->z_os = os; + zsb->z_parent = zsb; + zsb->z_version = version; + zsb->z_use_fuids = USE_FUIDS(version, os); + zsb->z_use_sa = USE_SA(version, os); + zsb->z_norm = norm; - bzero(&zfsvfs, sizeof (zfsvfs_t)); + sb = kmem_zalloc(sizeof (struct super_block), KM_PUSHPAGE); + sb->s_fs_info = zsb; + + ZTOI(rootzp)->i_sb = sb; + + error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END, + &zsb->z_attr_table); + + ASSERT(error == 0); - zfsvfs.z_os = os; - zfsvfs.z_parent = &zfsvfs; - zfsvfs.z_version = version; - zfsvfs.z_use_fuids = USE_FUIDS(version, os); - zfsvfs.z_norm = norm; /* * Fold case on file systems that are always or sometimes case * insensitive. */ if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED) - zfsvfs.z_norm |= U8_TEXTPREP_TOUPPER; + zsb->z_norm |= U8_TEXTPREP_TOUPPER; - mutex_init(&zfsvfs.z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); - list_create(&zfsvfs.z_all_znodes, sizeof (znode_t), + mutex_init(&zsb->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); + list_create(&zsb->z_all_znodes, sizeof (znode_t), offsetof(znode_t, z_link_node)); - ASSERT(!POINTER_IS_VALID(rootzp->z_zfsvfs)); - rootzp->z_zfsvfs = &zfsvfs; + for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) + mutex_init(&zsb->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL); + VERIFY(0 == zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr, cr, NULL, &acl_ids)); - zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, 0, &acl_ids); + zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, &acl_ids); ASSERT3P(zp, ==, rootzp); - ASSERT(!vn_in_dnlc(ZTOV(rootzp))); /* not valid to move */ error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx); ASSERT(error == 0); zfs_acl_ids_free(&acl_ids); - POINTER_INVALIDATE(&rootzp->z_zfsvfs); - ZTOV(rootzp)->v_count = 0; - dmu_buf_rele(rootzp->z_dbuf, NULL); - rootzp->z_dbuf = NULL; + atomic_set(&ZTOI(rootzp)->i_count, 0); + sa_handle_destroy(rootzp->z_sa_hdl); kmem_cache_free(znode_cache, rootzp); /* * Create shares directory */ + error = zfs_create_share_dir(zsb, tx); + ASSERT(error == 0); - error = zfs_create_share_dir(&zfsvfs, tx); + for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) + mutex_destroy(&zsb->z_hold_mtx[i]); - ASSERT(error == 0); + kmem_free(sb, sizeof (struct super_block)); + kmem_free(zsb, sizeof (zfs_sb_t)); } - #endif /* _KERNEL */ + +static int +zfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table) +{ + uint64_t sa_obj = 0; + int error; + + error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj); + if (error != 0 && error != ENOENT) + return (error); + + error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table); + return (error); +} + +static int +zfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp, + dmu_buf_t **db, void *tag) +{ + dmu_object_info_t doi; + int error; + + if ((error = sa_buf_hold(osp, obj, tag, db)) != 0) + return (error); + + dmu_object_info_from_db(*db, &doi); + if ((doi.doi_bonus_type != DMU_OT_SA && + doi.doi_bonus_type != DMU_OT_ZNODE) || + (doi.doi_bonus_type == DMU_OT_ZNODE && + doi.doi_bonus_size < sizeof (znode_phys_t))) { + sa_buf_rele(*db, tag); + return (ENOTSUP); + } + + error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp); + if (error != 0) { + sa_buf_rele(*db, tag); + return (error); + } + + return (0); +} + +void +zfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db, void *tag) +{ + sa_handle_destroy(hdl); + sa_buf_rele(db, tag); +} + /* * Given an object number, return its parent object number and whether * or not the object is an extended attribute directory. */ static int -zfs_obj_to_pobj(objset_t *osp, uint64_t obj, uint64_t *pobjp, int *is_xattrdir) +zfs_obj_to_pobj(sa_handle_t *hdl, sa_attr_type_t *sa_table, uint64_t *pobjp, + int *is_xattrdir) { - dmu_buf_t *db; - dmu_object_info_t doi; - znode_phys_t *zp; + uint64_t parent; + uint64_t pflags; + uint64_t mode; + sa_bulk_attr_t bulk[3]; + int count = 0; int error; - if ((error = dmu_bonus_hold(osp, obj, FTAG, &db)) != 0) - return (error); + SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL, + &parent, sizeof (parent)); + SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL, + &pflags, sizeof (pflags)); + SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL, + &mode, sizeof (mode)); - dmu_object_info_from_db(db, &doi); - if (doi.doi_bonus_type != DMU_OT_ZNODE || - doi.doi_bonus_size < sizeof (znode_phys_t)) { - dmu_buf_rele(db, FTAG); - return (EINVAL); - } + if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0) + return (error); - zp = db->db_data; - *pobjp = zp->zp_parent; - *is_xattrdir = ((zp->zp_flags & ZFS_XATTR) != 0) && - S_ISDIR(zp->zp_mode); - dmu_buf_rele(db, FTAG); + *pobjp = parent; + *is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode); return (0); } -int -zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len) +/* + * Given an object number, return some zpl level statistics + */ +static int +zfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table, + zfs_stat_t *sb) { + sa_bulk_attr_t bulk[4]; + int count = 0; + + SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL, + &sb->zs_mode, sizeof (sb->zs_mode)); + SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL, + &sb->zs_gen, sizeof (sb->zs_gen)); + SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL, + &sb->zs_links, sizeof (sb->zs_links)); + SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL, + &sb->zs_ctime, sizeof (sb->zs_ctime)); + + return (sa_bulk_lookup(hdl, bulk, count)); +} + +static int +zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl, + sa_attr_type_t *sa_table, char *buf, int len) +{ + sa_handle_t *sa_hdl; + sa_handle_t *prevhdl = NULL; + dmu_buf_t *prevdb = NULL; + dmu_buf_t *sa_db = NULL; char *path = buf + len - 1; int error; *path = '\0'; + sa_hdl = hdl; for (;;) { uint64_t pobj; @@ -1650,7 +1721,10 @@ zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len) size_t complen; int is_xattrdir; - if ((error = zfs_obj_to_pobj(osp, obj, &pobj, + if (prevdb) + zfs_release_sa_handle(prevhdl, prevdb, FTAG); + + if ((error = zfs_obj_to_pobj(sa_hdl, sa_table, &pobj, &is_xattrdir)) != 0) break; @@ -1675,9 +1749,85 @@ zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len) ASSERT(path >= buf); bcopy(component, path, complen); obj = pobj; + + if (sa_hdl != hdl) { + prevhdl = sa_hdl; + prevdb = sa_db; + } + error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db, FTAG); + if (error != 0) { + sa_hdl = prevhdl; + sa_db = prevdb; + break; + } + } + + if (sa_hdl != NULL && sa_hdl != hdl) { + ASSERT(sa_db != NULL); + zfs_release_sa_handle(sa_hdl, sa_db, FTAG); } if (error == 0) (void) memmove(buf, path, buf + len - path); + return (error); } + +int +zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len) +{ + sa_attr_type_t *sa_table; + sa_handle_t *hdl; + dmu_buf_t *db; + int error; + + error = zfs_sa_setup(osp, &sa_table); + if (error != 0) + return (error); + + error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG); + if (error != 0) + return (error); + + error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len); + + zfs_release_sa_handle(hdl, db, FTAG); + return (error); +} + +int +zfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb, + char *buf, int len) +{ + char *path = buf + len - 1; + sa_attr_type_t *sa_table; + sa_handle_t *hdl; + dmu_buf_t *db; + int error; + + *path = '\0'; + + error = zfs_sa_setup(osp, &sa_table); + if (error != 0) + return (error); + + error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG); + if (error != 0) + return (error); + + error = zfs_obj_to_stats_impl(hdl, sa_table, sb); + if (error != 0) { + zfs_release_sa_handle(hdl, db, FTAG); + return (error); + } + + error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len); + + zfs_release_sa_handle(hdl, db, FTAG); + return (error); +} + +#if defined(_KERNEL) && defined(HAVE_SPL) +EXPORT_SYMBOL(zfs_create_fs); +EXPORT_SYMBOL(zfs_obj_to_path); +#endif