Index: vendor-sys/illumos/dist/uts/common/fs/zfs/dmu_tx.c =================================================================== --- vendor-sys/illumos/dist/uts/common/fs/zfs/dmu_tx.c (revision 329714) +++ vendor-sys/illumos/dist/uts/common/fs/zfs/dmu_tx.c (revision 329715) @@ -1,1328 +1,1331 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright 2011 Nexenta Systems, Inc. All rights reserved. * Copyright (c) 2012, 2017 by Delphix. All rights reserved. * Copyright (c) 2014 Integros [integros.com] */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include typedef void (*dmu_tx_hold_func_t)(dmu_tx_t *tx, struct dnode *dn, uint64_t arg1, uint64_t arg2); dmu_tx_t * dmu_tx_create_dd(dsl_dir_t *dd) { dmu_tx_t *tx = kmem_zalloc(sizeof (dmu_tx_t), KM_SLEEP); tx->tx_dir = dd; if (dd != NULL) tx->tx_pool = dd->dd_pool; list_create(&tx->tx_holds, sizeof (dmu_tx_hold_t), offsetof(dmu_tx_hold_t, txh_node)); list_create(&tx->tx_callbacks, sizeof (dmu_tx_callback_t), offsetof(dmu_tx_callback_t, dcb_node)); tx->tx_start = gethrtime(); return (tx); } dmu_tx_t * dmu_tx_create(objset_t *os) { dmu_tx_t *tx = dmu_tx_create_dd(os->os_dsl_dataset->ds_dir); tx->tx_objset = os; return (tx); } dmu_tx_t * dmu_tx_create_assigned(struct dsl_pool *dp, uint64_t txg) { dmu_tx_t *tx = dmu_tx_create_dd(NULL); txg_verify(dp->dp_spa, txg); tx->tx_pool = dp; tx->tx_txg = txg; tx->tx_anyobj = TRUE; return (tx); } int dmu_tx_is_syncing(dmu_tx_t *tx) { return (tx->tx_anyobj); } int dmu_tx_private_ok(dmu_tx_t *tx) { return (tx->tx_anyobj); } static dmu_tx_hold_t * dmu_tx_hold_dnode_impl(dmu_tx_t *tx, dnode_t *dn, enum dmu_tx_hold_type type, uint64_t arg1, uint64_t arg2) { dmu_tx_hold_t *txh; if (dn != NULL) { (void) refcount_add(&dn->dn_holds, tx); if (tx->tx_txg != 0) { mutex_enter(&dn->dn_mtx); /* * dn->dn_assigned_txg == tx->tx_txg doesn't pose a * problem, but there's no way for it to happen (for * now, at least). */ ASSERT(dn->dn_assigned_txg == 0); dn->dn_assigned_txg = tx->tx_txg; (void) refcount_add(&dn->dn_tx_holds, tx); mutex_exit(&dn->dn_mtx); } } txh = kmem_zalloc(sizeof (dmu_tx_hold_t), KM_SLEEP); txh->txh_tx = tx; txh->txh_dnode = dn; refcount_create(&txh->txh_space_towrite); refcount_create(&txh->txh_memory_tohold); txh->txh_type = type; txh->txh_arg1 = arg1; txh->txh_arg2 = arg2; list_insert_tail(&tx->tx_holds, txh); return (txh); } static dmu_tx_hold_t * dmu_tx_hold_object_impl(dmu_tx_t *tx, objset_t *os, uint64_t object, enum dmu_tx_hold_type type, uint64_t arg1, uint64_t arg2) { dnode_t *dn = NULL; dmu_tx_hold_t *txh; int err; if (object != DMU_NEW_OBJECT) { err = dnode_hold(os, object, FTAG, &dn); if (err != 0) { tx->tx_err = err; return (NULL); } } txh = dmu_tx_hold_dnode_impl(tx, dn, type, arg1, arg2); if (dn != NULL) dnode_rele(dn, FTAG); return (txh); } void dmu_tx_add_new_object(dmu_tx_t *tx, dnode_t *dn) { /* * If we're syncing, they can manipulate any object anyhow, and * the hold on the dnode_t can cause problems. */ if (!dmu_tx_is_syncing(tx)) (void) dmu_tx_hold_dnode_impl(tx, dn, THT_NEWOBJECT, 0, 0); } /* * This function reads specified data from disk. The specified data will * be needed to perform the transaction -- i.e, it will be read after * we do dmu_tx_assign(). There are two reasons that we read the data now * (before dmu_tx_assign()): * * 1. Reading it now has potentially better performance. The transaction * has not yet been assigned, so the TXG is not held open, and also the * caller typically has less locks held when calling dmu_tx_hold_*() than * after the transaction has been assigned. This reduces the lock (and txg) * hold times, thus reducing lock contention. * * 2. It is easier for callers (primarily the ZPL) to handle i/o errors * that are detected before they start making changes to the DMU state * (i.e. now). Once the transaction has been assigned, and some DMU * state has been changed, it can be difficult to recover from an i/o * error (e.g. to undo the changes already made in memory at the DMU * layer). Typically code to do so does not exist in the caller -- it * assumes that the data has already been cached and thus i/o errors are * not possible. * * It has been observed that the i/o initiated here can be a performance * problem, and it appears to be optional, because we don't look at the * data which is read. However, removing this read would only serve to * move the work elsewhere (after the dmu_tx_assign()), where it may * have a greater impact on performance (in addition to the impact on * fault tolerance noted above). */ static int dmu_tx_check_ioerr(zio_t *zio, dnode_t *dn, int level, uint64_t blkid) { int err; dmu_buf_impl_t *db; rw_enter(&dn->dn_struct_rwlock, RW_READER); db = dbuf_hold_level(dn, level, blkid, FTAG); rw_exit(&dn->dn_struct_rwlock); if (db == NULL) return (SET_ERROR(EIO)); err = dbuf_read(db, zio, DB_RF_CANFAIL | DB_RF_NOPREFETCH); dbuf_rele(db, FTAG); return (err); } /* ARGSUSED */ static void dmu_tx_count_write(dmu_tx_hold_t *txh, uint64_t off, uint64_t len) { dnode_t *dn = txh->txh_dnode; int err = 0; if (len == 0) return; (void) refcount_add_many(&txh->txh_space_towrite, len, FTAG); if (refcount_count(&txh->txh_space_towrite) > 2 * DMU_MAX_ACCESS) err = SET_ERROR(EFBIG); if (dn == NULL) return; /* * For i/o error checking, read the blocks that will be needed * to perform the write: the first and last level-0 blocks (if * they are not aligned, i.e. if they are partial-block writes), * and all the level-1 blocks. */ if (dn->dn_maxblkid == 0) { if (off < dn->dn_datablksz && (off > 0 || len < dn->dn_datablksz)) { err = dmu_tx_check_ioerr(NULL, dn, 0, 0); if (err != 0) { txh->txh_tx->tx_err = err; } } } else { zio_t *zio = zio_root(dn->dn_objset->os_spa, NULL, NULL, ZIO_FLAG_CANFAIL); /* first level-0 block */ uint64_t start = off >> dn->dn_datablkshift; if (P2PHASE(off, dn->dn_datablksz) || len < dn->dn_datablksz) { err = dmu_tx_check_ioerr(zio, dn, 0, start); if (err != 0) { txh->txh_tx->tx_err = err; } } /* last level-0 block */ uint64_t end = (off + len - 1) >> dn->dn_datablkshift; if (end != start && end <= dn->dn_maxblkid && P2PHASE(off + len, dn->dn_datablksz)) { err = dmu_tx_check_ioerr(zio, dn, 0, end); if (err != 0) { txh->txh_tx->tx_err = err; } } /* level-1 blocks */ if (dn->dn_nlevels > 1) { int shft = dn->dn_indblkshift - SPA_BLKPTRSHIFT; for (uint64_t i = (start >> shft) + 1; i < end >> shft; i++) { err = dmu_tx_check_ioerr(zio, dn, 1, i); if (err != 0) { txh->txh_tx->tx_err = err; } } } err = zio_wait(zio); if (err != 0) { txh->txh_tx->tx_err = err; } } } static void dmu_tx_count_dnode(dmu_tx_hold_t *txh) { (void) refcount_add_many(&txh->txh_space_towrite, DNODE_SIZE, FTAG); } void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len) { dmu_tx_hold_t *txh; ASSERT0(tx->tx_txg); ASSERT3U(len, <=, DMU_MAX_ACCESS); ASSERT(len == 0 || UINT64_MAX - off >= len - 1); txh = dmu_tx_hold_object_impl(tx, tx->tx_objset, object, THT_WRITE, off, len); if (txh != NULL) { dmu_tx_count_write(txh, off, len); dmu_tx_count_dnode(txh); } } void dmu_tx_hold_remap_l1indirect(dmu_tx_t *tx, uint64_t object) { dmu_tx_hold_t *txh; ASSERT(tx->tx_txg == 0); txh = dmu_tx_hold_object_impl(tx, tx->tx_objset, object, THT_WRITE, 0, 0); if (txh == NULL) return; dnode_t *dn = txh->txh_dnode; (void) refcount_add_many(&txh->txh_space_towrite, 1ULL << dn->dn_indblkshift, FTAG); dmu_tx_count_dnode(txh); } void dmu_tx_hold_write_by_dnode(dmu_tx_t *tx, dnode_t *dn, uint64_t off, int len) { dmu_tx_hold_t *txh; ASSERT0(tx->tx_txg); ASSERT3U(len, <=, DMU_MAX_ACCESS); ASSERT(len == 0 || UINT64_MAX - off >= len - 1); txh = dmu_tx_hold_dnode_impl(tx, dn, THT_WRITE, off, len); if (txh != NULL) { dmu_tx_count_write(txh, off, len); dmu_tx_count_dnode(txh); } } /* * This function marks the transaction as being a "net free". The end * result is that refquotas will be disabled for this transaction, and * this transaction will be able to use half of the pool space overhead * (see dsl_pool_adjustedsize()). Therefore this function should only * be called for transactions that we expect will not cause a net increase * in the amount of space used (but it's OK if that is occasionally not true). */ void dmu_tx_mark_netfree(dmu_tx_t *tx) { tx->tx_netfree = B_TRUE; } static void dmu_tx_hold_free_impl(dmu_tx_hold_t *txh, uint64_t off, uint64_t len) { dmu_tx_t *tx; dnode_t *dn; int err; tx = txh->txh_tx; ASSERT(tx->tx_txg == 0); dn = txh->txh_dnode; dmu_tx_count_dnode(txh); if (off >= (dn->dn_maxblkid + 1) * dn->dn_datablksz) return; if (len == DMU_OBJECT_END) len = (dn->dn_maxblkid + 1) * dn->dn_datablksz - off; /* * For i/o error checking, we read the first and last level-0 * blocks if they are not aligned, and all the level-1 blocks. * * Note: dbuf_free_range() assumes that we have not instantiated * any level-0 dbufs that will be completely freed. Therefore we must * exercise care to not read or count the first and last blocks * if they are blocksize-aligned. */ if (dn->dn_datablkshift == 0) { if (off != 0 || len < dn->dn_datablksz) dmu_tx_count_write(txh, 0, dn->dn_datablksz); } else { /* first block will be modified if it is not aligned */ if (!IS_P2ALIGNED(off, 1 << dn->dn_datablkshift)) dmu_tx_count_write(txh, off, 1); /* last block will be modified if it is not aligned */ if (!IS_P2ALIGNED(off + len, 1 << dn->dn_datablkshift)) dmu_tx_count_write(txh, off + len, 1); } /* * Check level-1 blocks. */ if (dn->dn_nlevels > 1) { int shift = dn->dn_datablkshift + dn->dn_indblkshift - SPA_BLKPTRSHIFT; uint64_t start = off >> shift; uint64_t end = (off + len) >> shift; ASSERT(dn->dn_indblkshift != 0); /* * dnode_reallocate() can result in an object with indirect * blocks having an odd data block size. In this case, * just check the single block. */ if (dn->dn_datablkshift == 0) start = end = 0; zio_t *zio = zio_root(tx->tx_pool->dp_spa, NULL, NULL, ZIO_FLAG_CANFAIL); for (uint64_t i = start; i <= end; i++) { uint64_t ibyte = i << shift; err = dnode_next_offset(dn, 0, &ibyte, 2, 1, 0); i = ibyte >> shift; if (err == ESRCH || i > end) break; if (err != 0) { tx->tx_err = err; (void) zio_wait(zio); return; } (void) refcount_add_many(&txh->txh_memory_tohold, 1 << dn->dn_indblkshift, FTAG); err = dmu_tx_check_ioerr(zio, dn, 1, i); if (err != 0) { tx->tx_err = err; (void) zio_wait(zio); return; } } err = zio_wait(zio); if (err != 0) { tx->tx_err = err; return; } } } void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off, uint64_t len) { dmu_tx_hold_t *txh; txh = dmu_tx_hold_object_impl(tx, tx->tx_objset, object, THT_FREE, off, len); if (txh != NULL) (void) dmu_tx_hold_free_impl(txh, off, len); } void dmu_tx_hold_free_by_dnode(dmu_tx_t *tx, dnode_t *dn, uint64_t off, uint64_t len) { dmu_tx_hold_t *txh; txh = dmu_tx_hold_dnode_impl(tx, dn, THT_FREE, off, len); if (txh != NULL) (void) dmu_tx_hold_free_impl(txh, off, len); } static void dmu_tx_hold_zap_impl(dmu_tx_hold_t *txh, const char *name) { dmu_tx_t *tx = txh->txh_tx; dnode_t *dn; int err; ASSERT(tx->tx_txg == 0); dn = txh->txh_dnode; dmu_tx_count_dnode(txh); /* * Modifying a almost-full microzap is around the worst case (128KB) * * If it is a fat zap, the worst case would be 7*16KB=112KB: * - 3 blocks overwritten: target leaf, ptrtbl block, header block * - 4 new blocks written if adding: * - 2 blocks for possibly split leaves, * - 2 grown ptrtbl blocks */ (void) refcount_add_many(&txh->txh_space_towrite, MZAP_MAX_BLKSZ, FTAG); if (dn == NULL) return; ASSERT3P(DMU_OT_BYTESWAP(dn->dn_type), ==, DMU_BSWAP_ZAP); if (dn->dn_maxblkid == 0 || name == NULL) { /* * This is a microzap (only one block), or we don't know * the name. Check the first block for i/o errors. */ err = dmu_tx_check_ioerr(NULL, dn, 0, 0); if (err != 0) { tx->tx_err = err; } } else { /* * Access the name so that we'll check for i/o errors to * the leaf blocks, etc. We ignore ENOENT, as this name * may not yet exist. */ err = zap_lookup_by_dnode(dn, name, 8, 0, NULL); if (err == EIO || err == ECKSUM || err == ENXIO) { tx->tx_err = err; } } } void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name) { dmu_tx_hold_t *txh; ASSERT0(tx->tx_txg); txh = dmu_tx_hold_object_impl(tx, tx->tx_objset, object, THT_ZAP, add, (uintptr_t)name); if (txh != NULL) dmu_tx_hold_zap_impl(txh, name); } void dmu_tx_hold_zap_by_dnode(dmu_tx_t *tx, dnode_t *dn, int add, const char *name) { dmu_tx_hold_t *txh; ASSERT0(tx->tx_txg); ASSERT(dn != NULL); txh = dmu_tx_hold_dnode_impl(tx, dn, THT_ZAP, add, (uintptr_t)name); if (txh != NULL) dmu_tx_hold_zap_impl(txh, name); } void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object) { dmu_tx_hold_t *txh; ASSERT(tx->tx_txg == 0); txh = dmu_tx_hold_object_impl(tx, tx->tx_objset, object, THT_BONUS, 0, 0); if (txh) dmu_tx_count_dnode(txh); } void dmu_tx_hold_bonus_by_dnode(dmu_tx_t *tx, dnode_t *dn) { dmu_tx_hold_t *txh; ASSERT0(tx->tx_txg); txh = dmu_tx_hold_dnode_impl(tx, dn, THT_BONUS, 0, 0); if (txh) dmu_tx_count_dnode(txh); } void dmu_tx_hold_space(dmu_tx_t *tx, uint64_t space) { dmu_tx_hold_t *txh; ASSERT(tx->tx_txg == 0); txh = dmu_tx_hold_object_impl(tx, tx->tx_objset, DMU_NEW_OBJECT, THT_SPACE, space, 0); (void) refcount_add_many(&txh->txh_space_towrite, space, FTAG); } #ifdef ZFS_DEBUG void dmu_tx_dirty_buf(dmu_tx_t *tx, dmu_buf_impl_t *db) { boolean_t match_object = B_FALSE; boolean_t match_offset = B_FALSE; DB_DNODE_ENTER(db); dnode_t *dn = DB_DNODE(db); ASSERT(tx->tx_txg != 0); ASSERT(tx->tx_objset == NULL || dn->dn_objset == tx->tx_objset); ASSERT3U(dn->dn_object, ==, db->db.db_object); if (tx->tx_anyobj) { DB_DNODE_EXIT(db); return; } /* XXX No checking on the meta dnode for now */ if (db->db.db_object == DMU_META_DNODE_OBJECT) { DB_DNODE_EXIT(db); return; } for (dmu_tx_hold_t *txh = list_head(&tx->tx_holds); txh != NULL; txh = list_next(&tx->tx_holds, txh)) { ASSERT(dn == NULL || dn->dn_assigned_txg == tx->tx_txg); if (txh->txh_dnode == dn && txh->txh_type != THT_NEWOBJECT) match_object = TRUE; if (txh->txh_dnode == NULL || txh->txh_dnode == dn) { int datablkshift = dn->dn_datablkshift ? dn->dn_datablkshift : SPA_MAXBLOCKSHIFT; int epbs = dn->dn_indblkshift - SPA_BLKPTRSHIFT; int shift = datablkshift + epbs * db->db_level; uint64_t beginblk = shift >= 64 ? 0 : (txh->txh_arg1 >> shift); uint64_t endblk = shift >= 64 ? 0 : ((txh->txh_arg1 + txh->txh_arg2 - 1) >> shift); uint64_t blkid = db->db_blkid; /* XXX txh_arg2 better not be zero... */ dprintf("found txh type %x beginblk=%llx endblk=%llx\n", txh->txh_type, beginblk, endblk); switch (txh->txh_type) { case THT_WRITE: if (blkid >= beginblk && blkid <= endblk) match_offset = TRUE; /* * We will let this hold work for the bonus * or spill buffer so that we don't need to * hold it when creating a new object. */ if (blkid == DMU_BONUS_BLKID || blkid == DMU_SPILL_BLKID) match_offset = TRUE; /* * They might have to increase nlevels, * thus dirtying the new TLIBs. Or the * might have to change the block size, * thus dirying the new lvl=0 blk=0. */ if (blkid == 0) match_offset = TRUE; break; case THT_FREE: /* * We will dirty all the level 1 blocks in * the free range and perhaps the first and * last level 0 block. */ if (blkid >= beginblk && (blkid <= endblk || txh->txh_arg2 == DMU_OBJECT_END)) match_offset = TRUE; break; case THT_SPILL: if (blkid == DMU_SPILL_BLKID) match_offset = TRUE; break; case THT_BONUS: if (blkid == DMU_BONUS_BLKID) match_offset = TRUE; break; case THT_ZAP: match_offset = TRUE; break; case THT_NEWOBJECT: match_object = TRUE; break; default: ASSERT(!"bad txh_type"); } } if (match_object && match_offset) { DB_DNODE_EXIT(db); return; } } DB_DNODE_EXIT(db); panic("dirtying dbuf obj=%llx lvl=%u blkid=%llx but not tx_held\n", (u_longlong_t)db->db.db_object, db->db_level, (u_longlong_t)db->db_blkid); } #endif /* * If we can't do 10 iops, something is wrong. Let us go ahead * and hit zfs_dirty_data_max. */ hrtime_t zfs_delay_max_ns = MSEC2NSEC(100); int zfs_delay_resolution_ns = 100 * 1000; /* 100 microseconds */ /* * We delay transactions when we've determined that the backend storage * isn't able to accommodate the rate of incoming writes. * * If there is already a transaction waiting, we delay relative to when * that transaction finishes waiting. This way the calculated min_time * is independent of the number of threads concurrently executing * transactions. * * If we are the only waiter, wait relative to when the transaction * started, rather than the current time. This credits the transaction for * "time already served", e.g. reading indirect blocks. * * The minimum time for a transaction to take is calculated as: * min_time = scale * (dirty - min) / (max - dirty) * min_time is then capped at zfs_delay_max_ns. * * The delay has two degrees of freedom that can be adjusted via tunables. * The percentage of dirty data at which we start to delay is defined by * zfs_delay_min_dirty_percent. This should typically be at or above * zfs_vdev_async_write_active_max_dirty_percent so that we only start to * delay after writing at full speed has failed to keep up with the incoming * write rate. The scale of the curve is defined by zfs_delay_scale. Roughly * speaking, this variable determines the amount of delay at the midpoint of * the curve. * * delay * 10ms +-------------------------------------------------------------*+ * | *| * 9ms + *+ * | *| * 8ms + *+ * | * | * 7ms + * + * | * | * 6ms + * + * | * | * 5ms + * + * | * | * 4ms + * + * | * | * 3ms + * + * | * | * 2ms + (midpoint) * + * | | ** | * 1ms + v *** + * | zfs_delay_scale ----------> ******** | * 0 +-------------------------------------*********----------------+ * 0% <- zfs_dirty_data_max -> 100% * * Note that since the delay is added to the outstanding time remaining on the * most recent transaction, the delay is effectively the inverse of IOPS. * Here the midpoint of 500us translates to 2000 IOPS. The shape of the curve * was chosen such that small changes in the amount of accumulated dirty data * in the first 3/4 of the curve yield relatively small differences in the * amount of delay. * * The effects can be easier to understand when the amount of delay is * represented on a log scale: * * delay * 100ms +-------------------------------------------------------------++ * + + * | | * + *+ * 10ms + *+ * + ** + * | (midpoint) ** | * + | ** + * 1ms + v **** + * + zfs_delay_scale ----------> ***** + * | **** | * + **** + * 100us + ** + * + * + * | * | * + * + * 10us + * + * + + * | | * + + * +--------------------------------------------------------------+ * 0% <- zfs_dirty_data_max -> 100% * * Note here that only as the amount of dirty data approaches its limit does * the delay start to increase rapidly. The goal of a properly tuned system * should be to keep the amount of dirty data out of that range by first * ensuring that the appropriate limits are set for the I/O scheduler to reach * optimal throughput on the backend storage, and then by changing the value * of zfs_delay_scale to increase the steepness of the curve. */ static void dmu_tx_delay(dmu_tx_t *tx, uint64_t dirty) { dsl_pool_t *dp = tx->tx_pool; uint64_t delay_min_bytes = zfs_dirty_data_max * zfs_delay_min_dirty_percent / 100; hrtime_t wakeup, min_tx_time, now; if (dirty <= delay_min_bytes) return; /* * The caller has already waited until we are under the max. * We make them pass us the amount of dirty data so we don't * have to handle the case of it being >= the max, which could * cause a divide-by-zero if it's == the max. */ ASSERT3U(dirty, <, zfs_dirty_data_max); now = gethrtime(); min_tx_time = zfs_delay_scale * (dirty - delay_min_bytes) / (zfs_dirty_data_max - dirty); if (now > tx->tx_start + min_tx_time) return; min_tx_time = MIN(min_tx_time, zfs_delay_max_ns); DTRACE_PROBE3(delay__mintime, dmu_tx_t *, tx, uint64_t, dirty, uint64_t, min_tx_time); mutex_enter(&dp->dp_lock); wakeup = MAX(tx->tx_start + min_tx_time, dp->dp_last_wakeup + min_tx_time); dp->dp_last_wakeup = wakeup; mutex_exit(&dp->dp_lock); #ifdef _KERNEL mutex_enter(&curthread->t_delay_lock); while (cv_timedwait_hires(&curthread->t_delay_cv, &curthread->t_delay_lock, wakeup, zfs_delay_resolution_ns, CALLOUT_FLAG_ABSOLUTE | CALLOUT_FLAG_ROUNDUP) > 0) continue; mutex_exit(&curthread->t_delay_lock); #else hrtime_t delta = wakeup - gethrtime(); struct timespec ts; ts.tv_sec = delta / NANOSEC; ts.tv_nsec = delta % NANOSEC; (void) nanosleep(&ts, NULL); #endif } /* * This routine attempts to assign the transaction to a transaction group. * To do so, we must determine if there is sufficient free space on disk. * * If this is a "netfree" transaction (i.e. we called dmu_tx_mark_netfree() * on it), then it is assumed that there is sufficient free space, * unless there's insufficient slop space in the pool (see the comment * above spa_slop_shift in spa_misc.c). * * If it is not a "netfree" transaction, then if the data already on disk * is over the allowed usage (e.g. quota), this will fail with EDQUOT or * ENOSPC. Otherwise, if the current rough estimate of pending changes, * plus the rough estimate of this transaction's changes, may exceed the * allowed usage, then this will fail with ERESTART, which will cause the * caller to wait for the pending changes to be written to disk (by waiting * for the next TXG to open), and then check the space usage again. * * The rough estimate of pending changes is comprised of the sum of: * * - this transaction's holds' txh_space_towrite * * - dd_tempreserved[], which is the sum of in-flight transactions' * holds' txh_space_towrite (i.e. those transactions that have called * dmu_tx_assign() but not yet called dmu_tx_commit()). * * - dd_space_towrite[], which is the amount of dirtied dbufs. * * Note that all of these values are inflated by spa_get_worst_case_asize(), * which means that we may get ERESTART well before we are actually in danger * of running out of space, but this also mitigates any small inaccuracies * in the rough estimate (e.g. txh_space_towrite doesn't take into account * indirect blocks, and dd_space_towrite[] doesn't take into account changes * to the MOS). * * Note that due to this algorithm, it is possible to exceed the allowed * usage by one transaction. Also, as we approach the allowed usage, * we will allow a very limited amount of changes into each TXG, thus * decreasing performance. */ static int -dmu_tx_try_assign(dmu_tx_t *tx, txg_how_t txg_how) +dmu_tx_try_assign(dmu_tx_t *tx, uint64_t txg_how) { spa_t *spa = tx->tx_pool->dp_spa; ASSERT0(tx->tx_txg); if (tx->tx_err) return (tx->tx_err); if (spa_suspended(spa)) { /* * If the user has indicated a blocking failure mode * then return ERESTART which will block in dmu_tx_wait(). * Otherwise, return EIO so that an error can get * propagated back to the VOP calls. * * Note that we always honor the txg_how flag regardless * of the failuremode setting. */ if (spa_get_failmode(spa) == ZIO_FAILURE_MODE_CONTINUE && - txg_how != TXG_WAIT) + !(txg_how & TXG_WAIT)) return (SET_ERROR(EIO)); return (SET_ERROR(ERESTART)); } - if (!tx->tx_waited && + if (!tx->tx_dirty_delayed && dsl_pool_need_dirty_delay(tx->tx_pool)) { tx->tx_wait_dirty = B_TRUE; return (SET_ERROR(ERESTART)); } tx->tx_txg = txg_hold_open(tx->tx_pool, &tx->tx_txgh); tx->tx_needassign_txh = NULL; /* * NB: No error returns are allowed after txg_hold_open, but * before processing the dnode holds, due to the * dmu_tx_unassign() logic. */ uint64_t towrite = 0; uint64_t tohold = 0; for (dmu_tx_hold_t *txh = list_head(&tx->tx_holds); txh != NULL; txh = list_next(&tx->tx_holds, txh)) { dnode_t *dn = txh->txh_dnode; if (dn != NULL) { mutex_enter(&dn->dn_mtx); if (dn->dn_assigned_txg == tx->tx_txg - 1) { mutex_exit(&dn->dn_mtx); tx->tx_needassign_txh = txh; return (SET_ERROR(ERESTART)); } if (dn->dn_assigned_txg == 0) dn->dn_assigned_txg = tx->tx_txg; ASSERT3U(dn->dn_assigned_txg, ==, tx->tx_txg); (void) refcount_add(&dn->dn_tx_holds, tx); mutex_exit(&dn->dn_mtx); } towrite += refcount_count(&txh->txh_space_towrite); tohold += refcount_count(&txh->txh_memory_tohold); } /* needed allocation: worst-case estimate of write space */ uint64_t asize = spa_get_worst_case_asize(tx->tx_pool->dp_spa, towrite); /* calculate memory footprint estimate */ uint64_t memory = towrite + tohold; if (tx->tx_dir != NULL && asize != 0) { int err = dsl_dir_tempreserve_space(tx->tx_dir, memory, asize, tx->tx_netfree, &tx->tx_tempreserve_cookie, tx); if (err != 0) return (err); } return (0); } static void dmu_tx_unassign(dmu_tx_t *tx) { if (tx->tx_txg == 0) return; txg_rele_to_quiesce(&tx->tx_txgh); /* * Walk the transaction's hold list, removing the hold on the * associated dnode, and notifying waiters if the refcount drops to 0. */ for (dmu_tx_hold_t *txh = list_head(&tx->tx_holds); txh != tx->tx_needassign_txh; txh = list_next(&tx->tx_holds, txh)) { dnode_t *dn = txh->txh_dnode; if (dn == NULL) continue; mutex_enter(&dn->dn_mtx); ASSERT3U(dn->dn_assigned_txg, ==, tx->tx_txg); if (refcount_remove(&dn->dn_tx_holds, tx) == 0) { dn->dn_assigned_txg = 0; cv_broadcast(&dn->dn_notxholds); } mutex_exit(&dn->dn_mtx); } txg_rele_to_sync(&tx->tx_txgh); tx->tx_lasttried_txg = tx->tx_txg; tx->tx_txg = 0; } /* - * Assign tx to a transaction group. txg_how can be one of: + * Assign tx to a transaction group; txg_how is a bitmask: * - * (1) TXG_WAIT. If the current open txg is full, waits until there's - * a new one. This should be used when you're not holding locks. - * It will only fail if we're truly out of space (or over quota). + * If TXG_WAIT is set and the currently open txg is full, this function + * will wait until there's a new txg. This should be used when no locks + * are being held. With this bit set, this function will only fail if + * we're truly out of space (or over quota). * - * (2) TXG_NOWAIT. If we can't assign into the current open txg without - * blocking, returns immediately with ERESTART. This should be used - * whenever you're holding locks. On an ERESTART error, the caller - * should drop locks, do a dmu_tx_wait(tx), and try again. + * If TXG_WAIT is *not* set and we can't assign into the currently open + * txg without blocking, this function will return immediately with + * ERESTART. This should be used whenever locks are being held. On an + * ERESTART error, the caller should drop all locks, call dmu_tx_wait(), + * and try again. * - * (3) TXG_WAITED. Like TXG_NOWAIT, but indicates that dmu_tx_wait() - * has already been called on behalf of this operation (though - * most likely on a different tx). + * If TXG_NOTHROTTLE is set, this indicates that this tx should not be + * delayed due on the ZFS Write Throttle (see comments in dsl_pool.c for + * details on the throttle). This is used by the VFS operations, after + * they have already called dmu_tx_wait() (though most likely on a + * different tx). */ int -dmu_tx_assign(dmu_tx_t *tx, txg_how_t txg_how) +dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how) { int err; ASSERT(tx->tx_txg == 0); - ASSERT(txg_how == TXG_WAIT || txg_how == TXG_NOWAIT || - txg_how == TXG_WAITED); + ASSERT0(txg_how & ~(TXG_WAIT | TXG_NOTHROTTLE)); ASSERT(!dsl_pool_sync_context(tx->tx_pool)); /* If we might wait, we must not hold the config lock. */ - ASSERT(txg_how != TXG_WAIT || !dsl_pool_config_held(tx->tx_pool)); + IMPLY((txg_how & TXG_WAIT), !dsl_pool_config_held(tx->tx_pool)); - if (txg_how == TXG_WAITED) - tx->tx_waited = B_TRUE; + if ((txg_how & TXG_NOTHROTTLE)) + tx->tx_dirty_delayed = B_TRUE; while ((err = dmu_tx_try_assign(tx, txg_how)) != 0) { dmu_tx_unassign(tx); - if (err != ERESTART || txg_how != TXG_WAIT) + if (err != ERESTART || !(txg_how & TXG_WAIT)) return (err); dmu_tx_wait(tx); } txg_rele_to_quiesce(&tx->tx_txgh); return (0); } void dmu_tx_wait(dmu_tx_t *tx) { spa_t *spa = tx->tx_pool->dp_spa; dsl_pool_t *dp = tx->tx_pool; ASSERT(tx->tx_txg == 0); ASSERT(!dsl_pool_config_held(tx->tx_pool)); if (tx->tx_wait_dirty) { /* * dmu_tx_try_assign() has determined that we need to wait * because we've consumed much or all of the dirty buffer * space. */ mutex_enter(&dp->dp_lock); while (dp->dp_dirty_total >= zfs_dirty_data_max) cv_wait(&dp->dp_spaceavail_cv, &dp->dp_lock); uint64_t dirty = dp->dp_dirty_total; mutex_exit(&dp->dp_lock); dmu_tx_delay(tx, dirty); tx->tx_wait_dirty = B_FALSE; /* - * Note: setting tx_waited only has effect if the caller - * used TX_WAIT. Otherwise they are going to destroy - * this tx and try again. The common case, zfs_write(), - * uses TX_WAIT. + * Note: setting tx_dirty_delayed only has effect if the + * caller used TX_WAIT. Otherwise they are going to + * destroy this tx and try again. The common case, + * zfs_write(), uses TX_WAIT. */ - tx->tx_waited = B_TRUE; + tx->tx_dirty_delayed = B_TRUE; } else if (spa_suspended(spa) || tx->tx_lasttried_txg == 0) { /* * If the pool is suspended we need to wait until it * is resumed. Note that it's possible that the pool * has become active after this thread has tried to * obtain a tx. If that's the case then tx_lasttried_txg * would not have been set. */ txg_wait_synced(dp, spa_last_synced_txg(spa) + 1); } else if (tx->tx_needassign_txh) { /* * A dnode is assigned to the quiescing txg. Wait for its * transaction to complete. */ dnode_t *dn = tx->tx_needassign_txh->txh_dnode; mutex_enter(&dn->dn_mtx); while (dn->dn_assigned_txg == tx->tx_lasttried_txg - 1) cv_wait(&dn->dn_notxholds, &dn->dn_mtx); mutex_exit(&dn->dn_mtx); tx->tx_needassign_txh = NULL; } else { txg_wait_open(tx->tx_pool, tx->tx_lasttried_txg + 1); } } static void dmu_tx_destroy(dmu_tx_t *tx) { dmu_tx_hold_t *txh; while ((txh = list_head(&tx->tx_holds)) != NULL) { dnode_t *dn = txh->txh_dnode; list_remove(&tx->tx_holds, txh); refcount_destroy_many(&txh->txh_space_towrite, refcount_count(&txh->txh_space_towrite)); refcount_destroy_many(&txh->txh_memory_tohold, refcount_count(&txh->txh_memory_tohold)); kmem_free(txh, sizeof (dmu_tx_hold_t)); if (dn != NULL) dnode_rele(dn, tx); } list_destroy(&tx->tx_callbacks); list_destroy(&tx->tx_holds); kmem_free(tx, sizeof (dmu_tx_t)); } void dmu_tx_commit(dmu_tx_t *tx) { ASSERT(tx->tx_txg != 0); /* * Go through the transaction's hold list and remove holds on * associated dnodes, notifying waiters if no holds remain. */ for (dmu_tx_hold_t *txh = list_head(&tx->tx_holds); txh != NULL; txh = list_next(&tx->tx_holds, txh)) { dnode_t *dn = txh->txh_dnode; if (dn == NULL) continue; mutex_enter(&dn->dn_mtx); ASSERT3U(dn->dn_assigned_txg, ==, tx->tx_txg); if (refcount_remove(&dn->dn_tx_holds, tx) == 0) { dn->dn_assigned_txg = 0; cv_broadcast(&dn->dn_notxholds); } mutex_exit(&dn->dn_mtx); } if (tx->tx_tempreserve_cookie) dsl_dir_tempreserve_clear(tx->tx_tempreserve_cookie, tx); if (!list_is_empty(&tx->tx_callbacks)) txg_register_callbacks(&tx->tx_txgh, &tx->tx_callbacks); if (tx->tx_anyobj == FALSE) txg_rele_to_sync(&tx->tx_txgh); dmu_tx_destroy(tx); } void dmu_tx_abort(dmu_tx_t *tx) { ASSERT(tx->tx_txg == 0); /* * Call any registered callbacks with an error code. */ if (!list_is_empty(&tx->tx_callbacks)) dmu_tx_do_callbacks(&tx->tx_callbacks, ECANCELED); dmu_tx_destroy(tx); } uint64_t dmu_tx_get_txg(dmu_tx_t *tx) { ASSERT(tx->tx_txg != 0); return (tx->tx_txg); } dsl_pool_t * dmu_tx_pool(dmu_tx_t *tx) { ASSERT(tx->tx_pool != NULL); return (tx->tx_pool); } void dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *func, void *data) { dmu_tx_callback_t *dcb; dcb = kmem_alloc(sizeof (dmu_tx_callback_t), KM_SLEEP); dcb->dcb_func = func; dcb->dcb_data = data; list_insert_tail(&tx->tx_callbacks, dcb); } /* * Call all the commit callbacks on a list, with a given error code. */ void dmu_tx_do_callbacks(list_t *cb_list, int error) { dmu_tx_callback_t *dcb; while ((dcb = list_head(cb_list)) != NULL) { list_remove(cb_list, dcb); dcb->dcb_func(dcb->dcb_data, error); kmem_free(dcb, sizeof (dmu_tx_callback_t)); } } /* * Interface to hold a bunch of attributes. * used for creating new files. * attrsize is the total size of all attributes * to be added during object creation * * For updating/adding a single attribute dmu_tx_hold_sa() should be used. */ /* * hold necessary attribute name for attribute registration. * should be a very rare case where this is needed. If it does * happen it would only happen on the first write to the file system. */ static void dmu_tx_sa_registration_hold(sa_os_t *sa, dmu_tx_t *tx) { if (!sa->sa_need_attr_registration) return; for (int i = 0; i != sa->sa_num_attrs; i++) { if (!sa->sa_attr_table[i].sa_registered) { if (sa->sa_reg_attr_obj) dmu_tx_hold_zap(tx, sa->sa_reg_attr_obj, B_TRUE, sa->sa_attr_table[i].sa_name); else dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, B_TRUE, sa->sa_attr_table[i].sa_name); } } } void dmu_tx_hold_spill(dmu_tx_t *tx, uint64_t object) { dmu_tx_hold_t *txh = dmu_tx_hold_object_impl(tx, tx->tx_objset, object, THT_SPILL, 0, 0); (void) refcount_add_many(&txh->txh_space_towrite, SPA_OLD_MAXBLOCKSIZE, FTAG); } void dmu_tx_hold_sa_create(dmu_tx_t *tx, int attrsize) { sa_os_t *sa = tx->tx_objset->os_sa; dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT); if (tx->tx_objset->os_sa->sa_master_obj == 0) return; if (tx->tx_objset->os_sa->sa_layout_attr_obj) { dmu_tx_hold_zap(tx, sa->sa_layout_attr_obj, B_TRUE, NULL); } else { dmu_tx_hold_zap(tx, sa->sa_master_obj, B_TRUE, SA_LAYOUTS); dmu_tx_hold_zap(tx, sa->sa_master_obj, B_TRUE, SA_REGISTRY); dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, B_TRUE, NULL); dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, B_TRUE, NULL); } dmu_tx_sa_registration_hold(sa, tx); if (attrsize <= DN_MAX_BONUSLEN && !sa->sa_force_spill) return; (void) dmu_tx_hold_object_impl(tx, tx->tx_objset, DMU_NEW_OBJECT, THT_SPILL, 0, 0); } /* * Hold SA attribute * * dmu_tx_hold_sa(dmu_tx_t *tx, sa_handle_t *, attribute, add, size) * * variable_size is the total size of all variable sized attributes * passed to this function. It is not the total size of all * variable size attributes that *may* exist on this object. */ void dmu_tx_hold_sa(dmu_tx_t *tx, sa_handle_t *hdl, boolean_t may_grow) { uint64_t object; sa_os_t *sa = tx->tx_objset->os_sa; ASSERT(hdl != NULL); object = sa_handle_object(hdl); dmu_tx_hold_bonus(tx, object); if (tx->tx_objset->os_sa->sa_master_obj == 0) return; if (tx->tx_objset->os_sa->sa_reg_attr_obj == 0 || tx->tx_objset->os_sa->sa_layout_attr_obj == 0) { dmu_tx_hold_zap(tx, sa->sa_master_obj, B_TRUE, SA_LAYOUTS); dmu_tx_hold_zap(tx, sa->sa_master_obj, B_TRUE, SA_REGISTRY); dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, B_TRUE, NULL); dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, B_TRUE, NULL); } dmu_tx_sa_registration_hold(sa, tx); if (may_grow && tx->tx_objset->os_sa->sa_layout_attr_obj) dmu_tx_hold_zap(tx, sa->sa_layout_attr_obj, B_TRUE, NULL); if (sa->sa_force_spill || may_grow || hdl->sa_spill) { ASSERT(tx->tx_txg == 0); dmu_tx_hold_spill(tx, object); } else { dmu_buf_impl_t *db = (dmu_buf_impl_t *)hdl->sa_bonus; dnode_t *dn; DB_DNODE_ENTER(db); dn = DB_DNODE(db); if (dn->dn_have_spill) { ASSERT(tx->tx_txg == 0); dmu_tx_hold_spill(tx, object); } DB_DNODE_EXIT(db); } } Index: vendor-sys/illumos/dist/uts/common/fs/zfs/sys/dmu.h =================================================================== --- vendor-sys/illumos/dist/uts/common/fs/zfs/sys/dmu.h (revision 329714) +++ vendor-sys/illumos/dist/uts/common/fs/zfs/sys/dmu.h (revision 329715) @@ -1,971 +1,974 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2011, 2017 by Delphix. All rights reserved. * Copyright 2011 Nexenta Systems, Inc. All rights reserved. * Copyright (c) 2012, Joyent, Inc. All rights reserved. * Copyright 2013 DEY Storage Systems, Inc. * Copyright 2014 HybridCluster. All rights reserved. * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved. * Copyright 2013 Saso Kiselkov. All rights reserved. * Copyright (c) 2014 Integros [integros.com] */ /* Portions Copyright 2010 Robert Milkowski */ #ifndef _SYS_DMU_H #define _SYS_DMU_H /* * This file describes the interface that the DMU provides for its * consumers. * * The DMU also interacts with the SPA. That interface is described in * dmu_spa.h. */ #include #include #include #include #include #include #ifdef __cplusplus extern "C" { #endif struct uio; struct xuio; struct page; struct vnode; struct spa; struct zilog; struct zio; struct blkptr; struct zap_cursor; struct dsl_dataset; struct dsl_pool; struct dnode; struct drr_begin; struct drr_end; struct zbookmark_phys; struct spa; struct nvlist; struct arc_buf; struct zio_prop; struct sa_handle; typedef struct objset objset_t; typedef struct dmu_tx dmu_tx_t; typedef struct dsl_dir dsl_dir_t; typedef struct dnode dnode_t; typedef enum dmu_object_byteswap { DMU_BSWAP_UINT8, DMU_BSWAP_UINT16, DMU_BSWAP_UINT32, DMU_BSWAP_UINT64, DMU_BSWAP_ZAP, DMU_BSWAP_DNODE, DMU_BSWAP_OBJSET, DMU_BSWAP_ZNODE, DMU_BSWAP_OLDACL, DMU_BSWAP_ACL, /* * Allocating a new byteswap type number makes the on-disk format * incompatible with any other format that uses the same number. * * Data can usually be structured to work with one of the * DMU_BSWAP_UINT* or DMU_BSWAP_ZAP types. */ DMU_BSWAP_NUMFUNCS } dmu_object_byteswap_t; #define DMU_OT_NEWTYPE 0x80 #define DMU_OT_METADATA 0x40 #define DMU_OT_BYTESWAP_MASK 0x3f /* * Defines a uint8_t object type. Object types specify if the data * in the object is metadata (boolean) and how to byteswap the data * (dmu_object_byteswap_t). */ #define DMU_OT(byteswap, metadata) \ (DMU_OT_NEWTYPE | \ ((metadata) ? DMU_OT_METADATA : 0) | \ ((byteswap) & DMU_OT_BYTESWAP_MASK)) #define DMU_OT_IS_VALID(ot) (((ot) & DMU_OT_NEWTYPE) ? \ ((ot) & DMU_OT_BYTESWAP_MASK) < DMU_BSWAP_NUMFUNCS : \ (ot) < DMU_OT_NUMTYPES) #define DMU_OT_IS_METADATA(ot) (((ot) & DMU_OT_NEWTYPE) ? \ ((ot) & DMU_OT_METADATA) : \ dmu_ot[(ot)].ot_metadata) /* * These object types use bp_fill != 1 for their L0 bp's. Therefore they can't * have their data embedded (i.e. use a BP_IS_EMBEDDED() bp), because bp_fill * is repurposed for embedded BPs. */ #define DMU_OT_HAS_FILL(ot) \ ((ot) == DMU_OT_DNODE || (ot) == DMU_OT_OBJSET) #define DMU_OT_BYTESWAP(ot) (((ot) & DMU_OT_NEWTYPE) ? \ ((ot) & DMU_OT_BYTESWAP_MASK) : \ dmu_ot[(ot)].ot_byteswap) typedef enum dmu_object_type { DMU_OT_NONE, /* general: */ DMU_OT_OBJECT_DIRECTORY, /* ZAP */ DMU_OT_OBJECT_ARRAY, /* UINT64 */ DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */ DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */ DMU_OT_BPOBJ, /* UINT64 */ DMU_OT_BPOBJ_HDR, /* UINT64 */ /* spa: */ DMU_OT_SPACE_MAP_HEADER, /* UINT64 */ DMU_OT_SPACE_MAP, /* UINT64 */ /* zil: */ DMU_OT_INTENT_LOG, /* UINT64 */ /* dmu: */ DMU_OT_DNODE, /* DNODE */ DMU_OT_OBJSET, /* OBJSET */ /* dsl: */ DMU_OT_DSL_DIR, /* UINT64 */ DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */ DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */ DMU_OT_DSL_PROPS, /* ZAP */ DMU_OT_DSL_DATASET, /* UINT64 */ /* zpl: */ DMU_OT_ZNODE, /* ZNODE */ DMU_OT_OLDACL, /* Old ACL */ DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */ DMU_OT_DIRECTORY_CONTENTS, /* ZAP */ DMU_OT_MASTER_NODE, /* ZAP */ DMU_OT_UNLINKED_SET, /* ZAP */ /* zvol: */ DMU_OT_ZVOL, /* UINT8 */ DMU_OT_ZVOL_PROP, /* ZAP */ /* other; for testing only! */ DMU_OT_PLAIN_OTHER, /* UINT8 */ DMU_OT_UINT64_OTHER, /* UINT64 */ DMU_OT_ZAP_OTHER, /* ZAP */ /* new object types: */ DMU_OT_ERROR_LOG, /* ZAP */ DMU_OT_SPA_HISTORY, /* UINT8 */ DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */ DMU_OT_POOL_PROPS, /* ZAP */ DMU_OT_DSL_PERMS, /* ZAP */ DMU_OT_ACL, /* ACL */ DMU_OT_SYSACL, /* SYSACL */ DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */ DMU_OT_FUID_SIZE, /* FUID table size UINT64 */ DMU_OT_NEXT_CLONES, /* ZAP */ DMU_OT_SCAN_QUEUE, /* ZAP */ DMU_OT_USERGROUP_USED, /* ZAP */ DMU_OT_USERGROUP_QUOTA, /* ZAP */ DMU_OT_USERREFS, /* ZAP */ DMU_OT_DDT_ZAP, /* ZAP */ DMU_OT_DDT_STATS, /* ZAP */ DMU_OT_SA, /* System attr */ DMU_OT_SA_MASTER_NODE, /* ZAP */ DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */ DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */ DMU_OT_SCAN_XLATE, /* ZAP */ DMU_OT_DEDUP, /* fake dedup BP from ddt_bp_create() */ DMU_OT_DEADLIST, /* ZAP */ DMU_OT_DEADLIST_HDR, /* UINT64 */ DMU_OT_DSL_CLONES, /* ZAP */ DMU_OT_BPOBJ_SUBOBJ, /* UINT64 */ /* * Do not allocate new object types here. Doing so makes the on-disk * format incompatible with any other format that uses the same object * type number. * * When creating an object which does not have one of the above types * use the DMU_OTN_* type with the correct byteswap and metadata * values. * * The DMU_OTN_* types do not have entries in the dmu_ot table, * use the DMU_OT_IS_METDATA() and DMU_OT_BYTESWAP() macros instead * of indexing into dmu_ot directly (this works for both DMU_OT_* types * and DMU_OTN_* types). */ DMU_OT_NUMTYPES, /* * Names for valid types declared with DMU_OT(). */ DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE), DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE), DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE), DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE), DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE), DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE), DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE), DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE), DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE), DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE), } dmu_object_type_t; -typedef enum txg_how { - TXG_WAIT = 1, - TXG_NOWAIT, - TXG_WAITED, -} txg_how_t; +/* + * These flags are intended to be used to specify the "txg_how" + * parameter when calling the dmu_tx_assign() function. See the comment + * above dmu_tx_assign() for more details on the meaning of these flags. + */ +#define TXG_NOWAIT (0ULL) +#define TXG_WAIT (1ULL<<0) +#define TXG_NOTHROTTLE (1ULL<<1) void byteswap_uint64_array(void *buf, size_t size); void byteswap_uint32_array(void *buf, size_t size); void byteswap_uint16_array(void *buf, size_t size); void byteswap_uint8_array(void *buf, size_t size); void zap_byteswap(void *buf, size_t size); void zfs_oldacl_byteswap(void *buf, size_t size); void zfs_acl_byteswap(void *buf, size_t size); void zfs_znode_byteswap(void *buf, size_t size); #define DS_FIND_SNAPSHOTS (1<<0) #define DS_FIND_CHILDREN (1<<1) #define DS_FIND_SERIALIZE (1<<2) /* * The maximum number of bytes that can be accessed as part of one * operation, including metadata. */ #define DMU_MAX_ACCESS (32 * 1024 * 1024) /* 32MB */ #define DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */ #define DMU_USERUSED_OBJECT (-1ULL) #define DMU_GROUPUSED_OBJECT (-2ULL) /* * artificial blkids for bonus buffer and spill blocks */ #define DMU_BONUS_BLKID (-1ULL) #define DMU_SPILL_BLKID (-2ULL) /* * Public routines to create, destroy, open, and close objsets. */ int dmu_objset_hold(const char *name, void *tag, objset_t **osp); int dmu_objset_own(const char *name, dmu_objset_type_t type, boolean_t readonly, void *tag, objset_t **osp); void dmu_objset_rele(objset_t *os, void *tag); void dmu_objset_disown(objset_t *os, void *tag); int dmu_objset_open_ds(struct dsl_dataset *ds, objset_t **osp); void dmu_objset_evict_dbufs(objset_t *os); int dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags, void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg); int dmu_objset_clone(const char *name, const char *origin); int dsl_destroy_snapshots_nvl(struct nvlist *snaps, boolean_t defer, struct nvlist *errlist); int dmu_objset_snapshot_one(const char *fsname, const char *snapname); int dmu_objset_snapshot_tmp(const char *, const char *, int); int dmu_objset_find(char *name, int func(const char *, void *), void *arg, int flags); void dmu_objset_byteswap(void *buf, size_t size); int dsl_dataset_rename_snapshot(const char *fsname, const char *oldsnapname, const char *newsnapname, boolean_t recursive); int dmu_objset_remap_indirects(const char *fsname); typedef struct dmu_buf { uint64_t db_object; /* object that this buffer is part of */ uint64_t db_offset; /* byte offset in this object */ uint64_t db_size; /* size of buffer in bytes */ void *db_data; /* data in buffer */ } dmu_buf_t; /* * The names of zap entries in the DIRECTORY_OBJECT of the MOS. */ #define DMU_POOL_DIRECTORY_OBJECT 1 #define DMU_POOL_CONFIG "config" #define DMU_POOL_FEATURES_FOR_WRITE "features_for_write" #define DMU_POOL_FEATURES_FOR_READ "features_for_read" #define DMU_POOL_FEATURE_DESCRIPTIONS "feature_descriptions" #define DMU_POOL_FEATURE_ENABLED_TXG "feature_enabled_txg" #define DMU_POOL_ROOT_DATASET "root_dataset" #define DMU_POOL_SYNC_BPOBJ "sync_bplist" #define DMU_POOL_ERRLOG_SCRUB "errlog_scrub" #define DMU_POOL_ERRLOG_LAST "errlog_last" #define DMU_POOL_SPARES "spares" #define DMU_POOL_DEFLATE "deflate" #define DMU_POOL_HISTORY "history" #define DMU_POOL_PROPS "pool_props" #define DMU_POOL_L2CACHE "l2cache" #define DMU_POOL_TMP_USERREFS "tmp_userrefs" #define DMU_POOL_DDT "DDT-%s-%s-%s" #define DMU_POOL_DDT_STATS "DDT-statistics" #define DMU_POOL_CREATION_VERSION "creation_version" #define DMU_POOL_SCAN "scan" #define DMU_POOL_FREE_BPOBJ "free_bpobj" #define DMU_POOL_BPTREE_OBJ "bptree_obj" #define DMU_POOL_EMPTY_BPOBJ "empty_bpobj" #define DMU_POOL_CHECKSUM_SALT "org.illumos:checksum_salt" #define DMU_POOL_VDEV_ZAP_MAP "com.delphix:vdev_zap_map" #define DMU_POOL_REMOVING "com.delphix:removing" #define DMU_POOL_OBSOLETE_BPOBJ "com.delphix:obsolete_bpobj" #define DMU_POOL_CONDENSING_INDIRECT "com.delphix:condensing_indirect" /* * Allocate an object from this objset. The range of object numbers * available is (0, DN_MAX_OBJECT). Object 0 is the meta-dnode. * * The transaction must be assigned to a txg. The newly allocated * object will be "held" in the transaction (ie. you can modify the * newly allocated object in this transaction). * * dmu_object_alloc() chooses an object and returns it in *objectp. * * dmu_object_claim() allocates a specific object number. If that * number is already allocated, it fails and returns EEXIST. * * Return 0 on success, or ENOSPC or EEXIST as specified above. */ uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot, int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx); int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot, int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx); int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot, int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *txp); /* * Free an object from this objset. * * The object's data will be freed as well (ie. you don't need to call * dmu_free(object, 0, -1, tx)). * * The object need not be held in the transaction. * * If there are any holds on this object's buffers (via dmu_buf_hold()), * or tx holds on the object (via dmu_tx_hold_object()), you can not * free it; it fails and returns EBUSY. * * If the object is not allocated, it fails and returns ENOENT. * * Return 0 on success, or EBUSY or ENOENT as specified above. */ int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx); /* * Find the next allocated or free object. * * The objectp parameter is in-out. It will be updated to be the next * object which is allocated. Ignore objects which have not been * modified since txg. * * XXX Can only be called on a objset with no dirty data. * * Returns 0 on success, or ENOENT if there are no more objects. */ int dmu_object_next(objset_t *os, uint64_t *objectp, boolean_t hole, uint64_t txg); /* * Set the data blocksize for an object. * * The object cannot have any blocks allcated beyond the first. If * the first block is allocated already, the new size must be greater * than the current block size. If these conditions are not met, * ENOTSUP will be returned. * * Returns 0 on success, or EBUSY if there are any holds on the object * contents, or ENOTSUP as described above. */ int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size, int ibs, dmu_tx_t *tx); /* * Set the checksum property on a dnode. The new checksum algorithm will * apply to all newly written blocks; existing blocks will not be affected. */ void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum, dmu_tx_t *tx); /* * Set the compress property on a dnode. The new compression algorithm will * apply to all newly written blocks; existing blocks will not be affected. */ void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress, dmu_tx_t *tx); int dmu_object_remap_indirects(objset_t *os, uint64_t object, uint64_t txg); void dmu_write_embedded(objset_t *os, uint64_t object, uint64_t offset, void *data, uint8_t etype, uint8_t comp, int uncompressed_size, int compressed_size, int byteorder, dmu_tx_t *tx); /* * Decide how to write a block: checksum, compression, number of copies, etc. */ #define WP_NOFILL 0x1 #define WP_DMU_SYNC 0x2 #define WP_SPILL 0x4 void dmu_write_policy(objset_t *os, dnode_t *dn, int level, int wp, struct zio_prop *zp); /* * The bonus data is accessed more or less like a regular buffer. * You must dmu_bonus_hold() to get the buffer, which will give you a * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus * data. As with any normal buffer, you must call dmu_buf_will_dirty() * before modifying it, and the * object must be held in an assigned transaction before calling * dmu_buf_will_dirty. You may use dmu_buf_set_user() on the bonus * buffer as well. You must release your hold with dmu_buf_rele(). * * Returns ENOENT, EIO, or 0. */ int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **); int dmu_bonus_max(void); int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *); int dmu_set_bonustype(dmu_buf_t *, dmu_object_type_t, dmu_tx_t *); dmu_object_type_t dmu_get_bonustype(dmu_buf_t *); int dmu_rm_spill(objset_t *, uint64_t, dmu_tx_t *); /* * Special spill buffer support used by "SA" framework */ int dmu_spill_hold_by_bonus(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp); int dmu_spill_hold_by_dnode(dnode_t *dn, uint32_t flags, void *tag, dmu_buf_t **dbp); int dmu_spill_hold_existing(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp); /* * Obtain the DMU buffer from the specified object which contains the * specified offset. dmu_buf_hold() puts a "hold" on the buffer, so * that it will remain in memory. You must release the hold with * dmu_buf_rele(). You musn't access the dmu_buf_t after releasing your * hold. You must have a hold on any dmu_buf_t* you pass to the DMU. * * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill * on the returned buffer before reading or writing the buffer's * db_data. The comments for those routines describe what particular * operations are valid after calling them. * * The object number must be a valid, allocated object number. */ int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset, void *tag, dmu_buf_t **, int flags); int dmu_buf_hold_by_dnode(dnode_t *dn, uint64_t offset, void *tag, dmu_buf_t **dbp, int flags); /* * Add a reference to a dmu buffer that has already been held via * dmu_buf_hold() in the current context. */ void dmu_buf_add_ref(dmu_buf_t *db, void* tag); /* * Attempt to add a reference to a dmu buffer that is in an unknown state, * using a pointer that may have been invalidated by eviction processing. * The request will succeed if the passed in dbuf still represents the * same os/object/blkid, is ineligible for eviction, and has at least * one hold by a user other than the syncer. */ boolean_t dmu_buf_try_add_ref(dmu_buf_t *, objset_t *os, uint64_t object, uint64_t blkid, void *tag); void dmu_buf_rele(dmu_buf_t *db, void *tag); uint64_t dmu_buf_refcount(dmu_buf_t *db); /* * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a * range of an object. A pointer to an array of dmu_buf_t*'s is * returned (in *dbpp). * * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and * frees the array. The hold on the array of buffers MUST be released * with dmu_buf_rele_array. You can NOT release the hold on each buffer * individually with dmu_buf_rele. */ int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset, uint64_t length, boolean_t read, void *tag, int *numbufsp, dmu_buf_t ***dbpp); void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag); typedef void dmu_buf_evict_func_t(void *user_ptr); /* * A DMU buffer user object may be associated with a dbuf for the * duration of its lifetime. This allows the user of a dbuf (client) * to attach private data to a dbuf (e.g. in-core only data such as a * dnode_children_t, zap_t, or zap_leaf_t) and be optionally notified * when that dbuf has been evicted. Clients typically respond to the * eviction notification by freeing their private data, thus ensuring * the same lifetime for both dbuf and private data. * * The mapping from a dmu_buf_user_t to any client private data is the * client's responsibility. All current consumers of the API with private * data embed a dmu_buf_user_t as the first member of the structure for * their private data. This allows conversions between the two types * with a simple cast. Since the DMU buf user API never needs access * to the private data, other strategies can be employed if necessary * or convenient for the client (e.g. using container_of() to do the * conversion for private data that cannot have the dmu_buf_user_t as * its first member). * * Eviction callbacks are executed without the dbuf mutex held or any * other type of mechanism to guarantee that the dbuf is still available. * For this reason, users must assume the dbuf has already been freed * and not reference the dbuf from the callback context. * * Users requesting "immediate eviction" are notified as soon as the dbuf * is only referenced by dirty records (dirties == holds). Otherwise the * notification occurs after eviction processing for the dbuf begins. */ typedef struct dmu_buf_user { /* * Asynchronous user eviction callback state. */ taskq_ent_t dbu_tqent; /* * This instance's eviction function pointers. * * dbu_evict_func_sync is called synchronously and then * dbu_evict_func_async is executed asynchronously on a taskq. */ dmu_buf_evict_func_t *dbu_evict_func_sync; dmu_buf_evict_func_t *dbu_evict_func_async; #ifdef ZFS_DEBUG /* * Pointer to user's dbuf pointer. NULL for clients that do * not associate a dbuf with their user data. * * The dbuf pointer is cleared upon eviction so as to catch * use-after-evict bugs in clients. */ dmu_buf_t **dbu_clear_on_evict_dbufp; #endif } dmu_buf_user_t; /* * Initialize the given dmu_buf_user_t instance with the eviction function * evict_func, to be called when the user is evicted. * * NOTE: This function should only be called once on a given dmu_buf_user_t. * To allow enforcement of this, dbu must already be zeroed on entry. */ /*ARGSUSED*/ inline void dmu_buf_init_user(dmu_buf_user_t *dbu, dmu_buf_evict_func_t *evict_func_sync, dmu_buf_evict_func_t *evict_func_async, dmu_buf_t **clear_on_evict_dbufp) { ASSERT(dbu->dbu_evict_func_sync == NULL); ASSERT(dbu->dbu_evict_func_async == NULL); /* must have at least one evict func */ IMPLY(evict_func_sync == NULL, evict_func_async != NULL); dbu->dbu_evict_func_sync = evict_func_sync; dbu->dbu_evict_func_async = evict_func_async; #ifdef ZFS_DEBUG dbu->dbu_clear_on_evict_dbufp = clear_on_evict_dbufp; #endif } /* * Attach user data to a dbuf and mark it for normal (when the dbuf's * data is cleared or its reference count goes to zero) eviction processing. * * Returns NULL on success, or the existing user if another user currently * owns the buffer. */ void *dmu_buf_set_user(dmu_buf_t *db, dmu_buf_user_t *user); /* * Attach user data to a dbuf and mark it for immediate (its dirty and * reference counts are equal) eviction processing. * * Returns NULL on success, or the existing user if another user currently * owns the buffer. */ void *dmu_buf_set_user_ie(dmu_buf_t *db, dmu_buf_user_t *user); /* * Replace the current user of a dbuf. * * If given the current user of a dbuf, replaces the dbuf's user with * "new_user" and returns the user data pointer that was replaced. * Otherwise returns the current, and unmodified, dbuf user pointer. */ void *dmu_buf_replace_user(dmu_buf_t *db, dmu_buf_user_t *old_user, dmu_buf_user_t *new_user); /* * Remove the specified user data for a DMU buffer. * * Returns the user that was removed on success, or the current user if * another user currently owns the buffer. */ void *dmu_buf_remove_user(dmu_buf_t *db, dmu_buf_user_t *user); /* * Returns the user data (dmu_buf_user_t *) associated with this dbuf. */ void *dmu_buf_get_user(dmu_buf_t *db); objset_t *dmu_buf_get_objset(dmu_buf_t *db); dnode_t *dmu_buf_dnode_enter(dmu_buf_t *db); void dmu_buf_dnode_exit(dmu_buf_t *db); /* Block until any in-progress dmu buf user evictions complete. */ void dmu_buf_user_evict_wait(void); /* * Returns the blkptr associated with this dbuf, or NULL if not set. */ struct blkptr *dmu_buf_get_blkptr(dmu_buf_t *db); /* * Indicate that you are going to modify the buffer's data (db_data). * * The transaction (tx) must be assigned to a txg (ie. you've called * dmu_tx_assign()). The buffer's object must be held in the tx * (ie. you've called dmu_tx_hold_object(tx, db->db_object)). */ void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx); /* * You must create a transaction, then hold the objects which you will * (or might) modify as part of this transaction. Then you must assign * the transaction to a transaction group. Once the transaction has * been assigned, you can modify buffers which belong to held objects as * part of this transaction. You can't modify buffers before the * transaction has been assigned; you can't modify buffers which don't * belong to objects which this transaction holds; you can't hold * objects once the transaction has been assigned. You may hold an * object which you are going to free (with dmu_object_free()), but you * don't have to. * * You can abort the transaction before it has been assigned. * * Note that you may hold buffers (with dmu_buf_hold) at any time, * regardless of transaction state. */ #define DMU_NEW_OBJECT (-1ULL) #define DMU_OBJECT_END (-1ULL) dmu_tx_t *dmu_tx_create(objset_t *os); void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len); void dmu_tx_hold_write_by_dnode(dmu_tx_t *tx, dnode_t *dn, uint64_t off, int len); void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off, uint64_t len); void dmu_tx_hold_free_by_dnode(dmu_tx_t *tx, dnode_t *dn, uint64_t off, uint64_t len); void dmu_tx_hold_remap_l1indirect(dmu_tx_t *tx, uint64_t object); void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name); void dmu_tx_hold_zap_by_dnode(dmu_tx_t *tx, dnode_t *dn, int add, const char *name); void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object); void dmu_tx_hold_bonus_by_dnode(dmu_tx_t *tx, dnode_t *dn); void dmu_tx_hold_spill(dmu_tx_t *tx, uint64_t object); void dmu_tx_hold_sa(dmu_tx_t *tx, struct sa_handle *hdl, boolean_t may_grow); void dmu_tx_hold_sa_create(dmu_tx_t *tx, int total_size); void dmu_tx_abort(dmu_tx_t *tx); -int dmu_tx_assign(dmu_tx_t *tx, enum txg_how txg_how); +int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how); void dmu_tx_wait(dmu_tx_t *tx); void dmu_tx_commit(dmu_tx_t *tx); void dmu_tx_mark_netfree(dmu_tx_t *tx); /* * To register a commit callback, dmu_tx_callback_register() must be called. * * dcb_data is a pointer to caller private data that is passed on as a * callback parameter. The caller is responsible for properly allocating and * freeing it. * * When registering a callback, the transaction must be already created, but * it cannot be committed or aborted. It can be assigned to a txg or not. * * The callback will be called after the transaction has been safely written * to stable storage and will also be called if the dmu_tx is aborted. * If there is any error which prevents the transaction from being committed to * disk, the callback will be called with a value of error != 0. */ typedef void dmu_tx_callback_func_t(void *dcb_data, int error); void dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *dcb_func, void *dcb_data); /* * Free up the data blocks for a defined range of a file. If size is * -1, the range from offset to end-of-file is freed. */ int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, dmu_tx_t *tx); int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset, uint64_t size); int dmu_free_long_object(objset_t *os, uint64_t object); /* * Convenience functions. * * Canfail routines will return 0 on success, or an errno if there is a * nonrecoverable I/O error. */ #define DMU_READ_PREFETCH 0 /* prefetch */ #define DMU_READ_NO_PREFETCH 1 /* don't prefetch */ int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, void *buf, uint32_t flags); int dmu_read_by_dnode(dnode_t *dn, uint64_t offset, uint64_t size, void *buf, uint32_t flags); void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, const void *buf, dmu_tx_t *tx); void dmu_write_by_dnode(dnode_t *dn, uint64_t offset, uint64_t size, const void *buf, dmu_tx_t *tx); void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, dmu_tx_t *tx); int dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size); int dmu_read_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size); int dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size, dmu_tx_t *tx); int dmu_write_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size, dmu_tx_t *tx); int dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, struct page *pp, dmu_tx_t *tx); struct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size); void dmu_return_arcbuf(struct arc_buf *buf); void dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, struct arc_buf *buf, dmu_tx_t *tx); int dmu_xuio_init(struct xuio *uio, int niov); void dmu_xuio_fini(struct xuio *uio); int dmu_xuio_add(struct xuio *uio, struct arc_buf *abuf, offset_t off, size_t n); int dmu_xuio_cnt(struct xuio *uio); struct arc_buf *dmu_xuio_arcbuf(struct xuio *uio, int i); void dmu_xuio_clear(struct xuio *uio, int i); void xuio_stat_wbuf_copied(void); void xuio_stat_wbuf_nocopy(void); extern boolean_t zfs_prefetch_disable; extern int zfs_max_recordsize; /* * Asynchronously try to read in the data. */ void dmu_prefetch(objset_t *os, uint64_t object, int64_t level, uint64_t offset, uint64_t len, enum zio_priority pri); typedef struct dmu_object_info { /* All sizes are in bytes unless otherwise indicated. */ uint32_t doi_data_block_size; uint32_t doi_metadata_block_size; dmu_object_type_t doi_type; dmu_object_type_t doi_bonus_type; uint64_t doi_bonus_size; uint8_t doi_indirection; /* 2 = dnode->indirect->data */ uint8_t doi_checksum; uint8_t doi_compress; uint8_t doi_nblkptr; uint8_t doi_pad[4]; uint64_t doi_physical_blocks_512; /* data + metadata, 512b blks */ uint64_t doi_max_offset; uint64_t doi_fill_count; /* number of non-empty blocks */ } dmu_object_info_t; typedef void arc_byteswap_func_t(void *buf, size_t size); typedef struct dmu_object_type_info { dmu_object_byteswap_t ot_byteswap; boolean_t ot_metadata; char *ot_name; } dmu_object_type_info_t; typedef struct dmu_object_byteswap_info { arc_byteswap_func_t *ob_func; char *ob_name; } dmu_object_byteswap_info_t; extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES]; extern const dmu_object_byteswap_info_t dmu_ot_byteswap[DMU_BSWAP_NUMFUNCS]; /* * Get information on a DMU object. * * Return 0 on success or ENOENT if object is not allocated. * * If doi is NULL, just indicates whether the object exists. */ int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi); /* Like dmu_object_info, but faster if you have a held dnode in hand. */ void dmu_object_info_from_dnode(dnode_t *dn, dmu_object_info_t *doi); /* Like dmu_object_info, but faster if you have a held dbuf in hand. */ void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi); /* * Like dmu_object_info_from_db, but faster still when you only care about * the size. This is specifically optimized for zfs_getattr(). */ void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize, u_longlong_t *nblk512); typedef struct dmu_objset_stats { uint64_t dds_num_clones; /* number of clones of this */ uint64_t dds_creation_txg; uint64_t dds_guid; dmu_objset_type_t dds_type; uint8_t dds_is_snapshot; uint8_t dds_inconsistent; char dds_origin[ZFS_MAX_DATASET_NAME_LEN]; } dmu_objset_stats_t; /* * Get stats on a dataset. */ void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat); /* * Add entries to the nvlist for all the objset's properties. See * zfs_prop_table[] and zfs(1m) for details on the properties. */ void dmu_objset_stats(objset_t *os, struct nvlist *nv); /* * Get the space usage statistics for statvfs(). * * refdbytes is the amount of space "referenced" by this objset. * availbytes is the amount of space available to this objset, taking * into account quotas & reservations, assuming that no other objsets * use the space first. These values correspond to the 'referenced' and * 'available' properties, described in the zfs(1m) manpage. * * usedobjs and availobjs are the number of objects currently allocated, * and available. */ void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp, uint64_t *usedobjsp, uint64_t *availobjsp); /* * The fsid_guid is a 56-bit ID that can change to avoid collisions. * (Contrast with the ds_guid which is a 64-bit ID that will never * change, so there is a small probability that it will collide.) */ uint64_t dmu_objset_fsid_guid(objset_t *os); /* * Get the [cm]time for an objset's snapshot dir */ timestruc_t dmu_objset_snap_cmtime(objset_t *os); int dmu_objset_is_snapshot(objset_t *os); extern struct spa *dmu_objset_spa(objset_t *os); extern struct zilog *dmu_objset_zil(objset_t *os); extern struct dsl_pool *dmu_objset_pool(objset_t *os); extern struct dsl_dataset *dmu_objset_ds(objset_t *os); extern void dmu_objset_name(objset_t *os, char *buf); extern dmu_objset_type_t dmu_objset_type(objset_t *os); extern uint64_t dmu_objset_id(objset_t *os); extern zfs_sync_type_t dmu_objset_syncprop(objset_t *os); extern zfs_logbias_op_t dmu_objset_logbias(objset_t *os); extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name, uint64_t *id, uint64_t *offp, boolean_t *case_conflict); extern int dmu_snapshot_realname(objset_t *os, char *name, char *real, int maxlen, boolean_t *conflict); extern int dmu_dir_list_next(objset_t *os, int namelen, char *name, uint64_t *idp, uint64_t *offp); typedef int objset_used_cb_t(dmu_object_type_t bonustype, void *bonus, uint64_t *userp, uint64_t *groupp); extern void dmu_objset_register_type(dmu_objset_type_t ost, objset_used_cb_t *cb); extern void dmu_objset_set_user(objset_t *os, void *user_ptr); extern void *dmu_objset_get_user(objset_t *os); /* * Return the txg number for the given assigned transaction. */ uint64_t dmu_tx_get_txg(dmu_tx_t *tx); /* * Synchronous write. * If a parent zio is provided this function initiates a write on the * provided buffer as a child of the parent zio. * In the absence of a parent zio, the write is completed synchronously. * At write completion, blk is filled with the bp of the written block. * Note that while the data covered by this function will be on stable * storage when the write completes this new data does not become a * permanent part of the file until the associated transaction commits. */ /* * {zfs,zvol,ztest}_get_done() args */ typedef struct zgd { struct lwb *zgd_lwb; struct blkptr *zgd_bp; dmu_buf_t *zgd_db; struct rl *zgd_rl; void *zgd_private; } zgd_t; typedef void dmu_sync_cb_t(zgd_t *arg, int error); int dmu_sync(struct zio *zio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd); /* * Find the next hole or data block in file starting at *off * Return found offset in *off. Return ESRCH for end of file. */ int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole, uint64_t *off); /* * Check if a DMU object has any dirty blocks. If so, sync out * all pending transaction groups. Otherwise, this function * does not alter DMU state. This could be improved to only sync * out the necessary transaction groups for this particular * object. */ int dmu_object_wait_synced(objset_t *os, uint64_t object); /* * Initial setup and final teardown. */ extern void dmu_init(void); extern void dmu_fini(void); typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp, uint64_t object, uint64_t offset, int len); void dmu_traverse_objset(objset_t *os, uint64_t txg_start, dmu_traverse_cb_t cb, void *arg); int dmu_diff(const char *tosnap_name, const char *fromsnap_name, struct vnode *vp, offset_t *offp); /* CRC64 table */ #define ZFS_CRC64_POLY 0xC96C5795D7870F42ULL /* ECMA-182, reflected form */ extern uint64_t zfs_crc64_table[256]; extern int zfs_mdcomp_disable; #ifdef __cplusplus } #endif #endif /* _SYS_DMU_H */ Index: vendor-sys/illumos/dist/uts/common/fs/zfs/sys/dmu_tx.h =================================================================== --- vendor-sys/illumos/dist/uts/common/fs/zfs/sys/dmu_tx.h (revision 329714) +++ vendor-sys/illumos/dist/uts/common/fs/zfs/sys/dmu_tx.h (revision 329715) @@ -1,153 +1,153 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright 2010 Sun Microsystems, Inc. All rights reserved. * Use is subject to license terms. */ /* * Copyright (c) 2012, 2016 by Delphix. All rights reserved. */ #ifndef _SYS_DMU_TX_H #define _SYS_DMU_TX_H #include #include #include #include #ifdef __cplusplus extern "C" { #endif struct dmu_buf_impl; struct dmu_tx_hold; struct dnode_link; struct dsl_pool; struct dnode; struct dsl_dir; struct dmu_tx { /* * No synchronization is needed because a tx can only be handled * by one thread. */ list_t tx_holds; /* list of dmu_tx_hold_t */ objset_t *tx_objset; struct dsl_dir *tx_dir; struct dsl_pool *tx_pool; uint64_t tx_txg; uint64_t tx_lastsnap_txg; uint64_t tx_lasttried_txg; txg_handle_t tx_txgh; void *tx_tempreserve_cookie; struct dmu_tx_hold *tx_needassign_txh; /* list of dmu_tx_callback_t on this dmu_tx */ list_t tx_callbacks; /* placeholder for syncing context, doesn't need specific holds */ boolean_t tx_anyobj; - /* has this transaction already been delayed? */ - boolean_t tx_waited; - /* transaction is marked as being a "net free" of space */ boolean_t tx_netfree; /* time this transaction was created */ hrtime_t tx_start; /* need to wait for sufficient dirty space */ boolean_t tx_wait_dirty; + /* has this transaction already been delayed? */ + boolean_t tx_dirty_delayed; + int tx_err; }; enum dmu_tx_hold_type { THT_NEWOBJECT, THT_WRITE, THT_BONUS, THT_FREE, THT_ZAP, THT_SPACE, THT_SPILL, THT_NUMTYPES }; typedef struct dmu_tx_hold { dmu_tx_t *txh_tx; list_node_t txh_node; struct dnode *txh_dnode; refcount_t txh_space_towrite; refcount_t txh_memory_tohold; enum dmu_tx_hold_type txh_type; uint64_t txh_arg1; uint64_t txh_arg2; } dmu_tx_hold_t; typedef struct dmu_tx_callback { list_node_t dcb_node; /* linked to tx_callbacks list */ dmu_tx_callback_func_t *dcb_func; /* caller function pointer */ void *dcb_data; /* caller private data */ } dmu_tx_callback_t; /* * These routines are defined in dmu.h, and are called by the user. */ dmu_tx_t *dmu_tx_create(objset_t *dd); -int dmu_tx_assign(dmu_tx_t *tx, txg_how_t txg_how); +int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how); void dmu_tx_commit(dmu_tx_t *tx); void dmu_tx_abort(dmu_tx_t *tx); uint64_t dmu_tx_get_txg(dmu_tx_t *tx); struct dsl_pool *dmu_tx_pool(dmu_tx_t *tx); void dmu_tx_wait(dmu_tx_t *tx); void dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *dcb_func, void *dcb_data); void dmu_tx_do_callbacks(list_t *cb_list, int error); /* * These routines are defined in dmu_spa.h, and are called by the SPA. */ extern dmu_tx_t *dmu_tx_create_assigned(struct dsl_pool *dp, uint64_t txg); /* * These routines are only called by the DMU. */ dmu_tx_t *dmu_tx_create_dd(dsl_dir_t *dd); int dmu_tx_is_syncing(dmu_tx_t *tx); int dmu_tx_private_ok(dmu_tx_t *tx); void dmu_tx_add_new_object(dmu_tx_t *tx, dnode_t *dn); void dmu_tx_dirty_buf(dmu_tx_t *tx, struct dmu_buf_impl *db); void dmu_tx_hold_space(dmu_tx_t *tx, uint64_t space); #ifdef ZFS_DEBUG #define DMU_TX_DIRTY_BUF(tx, db) dmu_tx_dirty_buf(tx, db) #else #define DMU_TX_DIRTY_BUF(tx, db) #endif #ifdef __cplusplus } #endif #endif /* _SYS_DMU_TX_H */ Index: vendor-sys/illumos/dist/uts/common/fs/zfs/zfs_vnops.c =================================================================== --- vendor-sys/illumos/dist/uts/common/fs/zfs/zfs_vnops.c (revision 329714) +++ vendor-sys/illumos/dist/uts/common/fs/zfs/zfs_vnops.c (revision 329715) @@ -1,5397 +1,5398 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2012, 2017 by Delphix. All rights reserved. * Copyright (c) 2014 Integros [integros.com] * Copyright 2015 Joyent, Inc. * Copyright 2017 Nexenta Systems, Inc. */ /* Portions Copyright 2007 Jeremy Teo */ /* Portions Copyright 2010 Robert Milkowski */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "fs/fs_subr.h" #include #include #include #include #include #include #include #include #include #include /* * Programming rules. * * Each vnode op performs some logical unit of work. To do this, the ZPL must * properly lock its in-core state, create a DMU transaction, do the work, * record this work in the intent log (ZIL), commit the DMU transaction, * and wait for the intent log to commit if it is a synchronous operation. * Moreover, the vnode ops must work in both normal and log replay context. * The ordering of events is important to avoid deadlocks and references * to freed memory. The example below illustrates the following Big Rules: * * (1) A check must be made in each zfs thread for a mounted file system. * This is done avoiding races using ZFS_ENTER(zfsvfs). * A ZFS_EXIT(zfsvfs) is needed before all returns. Any znodes * must be checked with ZFS_VERIFY_ZP(zp). Both of these macros * can return EIO from the calling function. * * (2) VN_RELE() should always be the last thing except for zil_commit() * (if necessary) and ZFS_EXIT(). This is for 3 reasons: * First, if it's the last reference, the vnode/znode * can be freed, so the zp may point to freed memory. Second, the last * reference will call zfs_zinactive(), which may induce a lot of work -- * pushing cached pages (which acquires range locks) and syncing out * cached atime changes. Third, zfs_zinactive() may require a new tx, * which could deadlock the system if you were already holding one. * If you must call VN_RELE() within a tx then use VN_RELE_ASYNC(). * * (3) All range locks must be grabbed before calling dmu_tx_assign(), * as they can span dmu_tx_assign() calls. * * (4) If ZPL locks are held, pass TXG_NOWAIT as the second argument to * dmu_tx_assign(). This is critical because we don't want to block * while holding locks. * * If no ZPL locks are held (aside from ZFS_ENTER()), use TXG_WAIT. This * reduces lock contention and CPU usage when we must wait (note that if * throughput is constrained by the storage, nearly every transaction * must wait). * * Note, in particular, that if a lock is sometimes acquired before * the tx assigns, and sometimes after (e.g. z_lock), then failing * to use a non-blocking assign can deadlock the system. The scenario: * * Thread A has grabbed a lock before calling dmu_tx_assign(). * Thread B is in an already-assigned tx, and blocks for this lock. * Thread A calls dmu_tx_assign(TXG_WAIT) and blocks in txg_wait_open() * forever, because the previous txg can't quiesce until B's tx commits. * * If dmu_tx_assign() returns ERESTART and zfsvfs->z_assign is TXG_NOWAIT, * then drop all locks, call dmu_tx_wait(), and try again. On subsequent - * calls to dmu_tx_assign(), pass TXG_WAITED rather than TXG_NOWAIT, + * calls to dmu_tx_assign(), pass TXG_NOTHROTTLE in addition to TXG_NOWAIT, * to indicate that this operation has already called dmu_tx_wait(). * This will ensure that we don't retry forever, waiting a short bit * each time. * * (5) If the operation succeeded, generate the intent log entry for it * before dropping locks. This ensures that the ordering of events * in the intent log matches the order in which they actually occurred. * During ZIL replay the zfs_log_* functions will update the sequence * number to indicate the zil transaction has replayed. * * (6) At the end of each vnode op, the DMU tx must always commit, * regardless of whether there were any errors. * * (7) After dropping all locks, invoke zil_commit(zilog, foid) * to ensure that synchronous semantics are provided when necessary. * * In general, this is how things should be ordered in each vnode op: * * ZFS_ENTER(zfsvfs); // exit if unmounted * top: * zfs_dirent_lock(&dl, ...) // lock directory entry (may VN_HOLD()) * rw_enter(...); // grab any other locks you need * tx = dmu_tx_create(...); // get DMU tx * dmu_tx_hold_*(); // hold each object you might modify - * error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); + * error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); * if (error) { * rw_exit(...); // drop locks * zfs_dirent_unlock(dl); // unlock directory entry * VN_RELE(...); // release held vnodes * if (error == ERESTART) { * waited = B_TRUE; * dmu_tx_wait(tx); * dmu_tx_abort(tx); * goto top; * } * dmu_tx_abort(tx); // abort DMU tx * ZFS_EXIT(zfsvfs); // finished in zfs * return (error); // really out of space * } * error = do_real_work(); // do whatever this VOP does * if (error == 0) * zfs_log_*(...); // on success, make ZIL entry * dmu_tx_commit(tx); // commit DMU tx -- error or not * rw_exit(...); // drop locks * zfs_dirent_unlock(dl); // unlock directory entry * VN_RELE(...); // release held vnodes * zil_commit(zilog, foid); // synchronous when necessary * ZFS_EXIT(zfsvfs); // finished in zfs * return (error); // done, report error */ /* ARGSUSED */ static int zfs_open(vnode_t **vpp, int flag, cred_t *cr, caller_context_t *ct) { znode_t *zp = VTOZ(*vpp); zfsvfs_t *zfsvfs = zp->z_zfsvfs; ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(zp); if ((flag & FWRITE) && (zp->z_pflags & ZFS_APPENDONLY) && ((flag & FAPPEND) == 0)) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EPERM)); } if (!zfs_has_ctldir(zp) && zp->z_zfsvfs->z_vscan && ZTOV(zp)->v_type == VREG && !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0) { if (fs_vscan(*vpp, cr, 0) != 0) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EACCES)); } } /* Keep a count of the synchronous opens in the znode */ if (flag & (FSYNC | FDSYNC)) atomic_inc_32(&zp->z_sync_cnt); ZFS_EXIT(zfsvfs); return (0); } /* ARGSUSED */ static int zfs_close(vnode_t *vp, int flag, int count, offset_t offset, cred_t *cr, caller_context_t *ct) { znode_t *zp = VTOZ(vp); zfsvfs_t *zfsvfs = zp->z_zfsvfs; /* * Clean up any locks held by this process on the vp. */ cleanlocks(vp, ddi_get_pid(), 0); cleanshares(vp, ddi_get_pid()); ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(zp); /* Decrement the synchronous opens in the znode */ if ((flag & (FSYNC | FDSYNC)) && (count == 1)) atomic_dec_32(&zp->z_sync_cnt); if (!zfs_has_ctldir(zp) && zp->z_zfsvfs->z_vscan && ZTOV(zp)->v_type == VREG && !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0) VERIFY(fs_vscan(vp, cr, 1) == 0); ZFS_EXIT(zfsvfs); return (0); } /* * Lseek support for finding holes (cmd == _FIO_SEEK_HOLE) and * data (cmd == _FIO_SEEK_DATA). "off" is an in/out parameter. */ static int zfs_holey(vnode_t *vp, int cmd, offset_t *off) { znode_t *zp = VTOZ(vp); uint64_t noff = (uint64_t)*off; /* new offset */ uint64_t file_sz; int error; boolean_t hole; file_sz = zp->z_size; if (noff >= file_sz) { return (SET_ERROR(ENXIO)); } if (cmd == _FIO_SEEK_HOLE) hole = B_TRUE; else hole = B_FALSE; error = dmu_offset_next(zp->z_zfsvfs->z_os, zp->z_id, hole, &noff); if (error == ESRCH) return (SET_ERROR(ENXIO)); /* * We could find a hole that begins after the logical end-of-file, * because dmu_offset_next() only works on whole blocks. If the * EOF falls mid-block, then indicate that the "virtual hole" * at the end of the file begins at the logical EOF, rather than * at the end of the last block. */ if (noff > file_sz) { ASSERT(hole); noff = file_sz; } if (noff < *off) return (error); *off = noff; return (error); } /* ARGSUSED */ static int zfs_ioctl(vnode_t *vp, int com, intptr_t data, int flag, cred_t *cred, int *rvalp, caller_context_t *ct) { offset_t off; offset_t ndata; dmu_object_info_t doi; int error; zfsvfs_t *zfsvfs; znode_t *zp; switch (com) { case _FIOFFS: { return (zfs_sync(vp->v_vfsp, 0, cred)); /* * The following two ioctls are used by bfu. Faking out, * necessary to avoid bfu errors. */ } case _FIOGDIO: case _FIOSDIO: { return (0); } case _FIO_SEEK_DATA: case _FIO_SEEK_HOLE: { if (ddi_copyin((void *)data, &off, sizeof (off), flag)) return (SET_ERROR(EFAULT)); zp = VTOZ(vp); zfsvfs = zp->z_zfsvfs; ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(zp); /* offset parameter is in/out */ error = zfs_holey(vp, com, &off); ZFS_EXIT(zfsvfs); if (error) return (error); if (ddi_copyout(&off, (void *)data, sizeof (off), flag)) return (SET_ERROR(EFAULT)); return (0); } case _FIO_COUNT_FILLED: { /* * _FIO_COUNT_FILLED adds a new ioctl command which * exposes the number of filled blocks in a * ZFS object. */ zp = VTOZ(vp); zfsvfs = zp->z_zfsvfs; ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(zp); /* * Wait for all dirty blocks for this object * to get synced out to disk, and the DMU info * updated. */ error = dmu_object_wait_synced(zfsvfs->z_os, zp->z_id); if (error) { ZFS_EXIT(zfsvfs); return (error); } /* * Retrieve fill count from DMU object. */ error = dmu_object_info(zfsvfs->z_os, zp->z_id, &doi); if (error) { ZFS_EXIT(zfsvfs); return (error); } ndata = doi.doi_fill_count; ZFS_EXIT(zfsvfs); if (ddi_copyout(&ndata, (void *)data, sizeof (ndata), flag)) return (SET_ERROR(EFAULT)); return (0); } } return (SET_ERROR(ENOTTY)); } /* * Utility functions to map and unmap a single physical page. These * are used to manage the mappable copies of ZFS file data, and therefore * do not update ref/mod bits. */ caddr_t zfs_map_page(page_t *pp, enum seg_rw rw) { if (kpm_enable) return (hat_kpm_mapin(pp, 0)); ASSERT(rw == S_READ || rw == S_WRITE); return (ppmapin(pp, PROT_READ | ((rw == S_WRITE) ? PROT_WRITE : 0), (caddr_t)-1)); } void zfs_unmap_page(page_t *pp, caddr_t addr) { if (kpm_enable) { hat_kpm_mapout(pp, 0, addr); } else { ppmapout(addr); } } /* * When a file is memory mapped, we must keep the IO data synchronized * between the DMU cache and the memory mapped pages. What this means: * * On Write: If we find a memory mapped page, we write to *both* * the page and the dmu buffer. */ static void update_pages(vnode_t *vp, int64_t start, int len, objset_t *os, uint64_t oid) { int64_t off; off = start & PAGEOFFSET; for (start &= PAGEMASK; len > 0; start += PAGESIZE) { page_t *pp; uint64_t nbytes = MIN(PAGESIZE - off, len); if (pp = page_lookup(vp, start, SE_SHARED)) { caddr_t va; va = zfs_map_page(pp, S_WRITE); (void) dmu_read(os, oid, start+off, nbytes, va+off, DMU_READ_PREFETCH); zfs_unmap_page(pp, va); page_unlock(pp); } len -= nbytes; off = 0; } } /* * When a file is memory mapped, we must keep the IO data synchronized * between the DMU cache and the memory mapped pages. What this means: * * On Read: We "read" preferentially from memory mapped pages, * else we default from the dmu buffer. * * NOTE: We will always "break up" the IO into PAGESIZE uiomoves when * the file is memory mapped. */ static int mappedread(vnode_t *vp, int nbytes, uio_t *uio) { znode_t *zp = VTOZ(vp); int64_t start, off; int len = nbytes; int error = 0; start = uio->uio_loffset; off = start & PAGEOFFSET; for (start &= PAGEMASK; len > 0; start += PAGESIZE) { page_t *pp; uint64_t bytes = MIN(PAGESIZE - off, len); if (pp = page_lookup(vp, start, SE_SHARED)) { caddr_t va; va = zfs_map_page(pp, S_READ); error = uiomove(va + off, bytes, UIO_READ, uio); zfs_unmap_page(pp, va); page_unlock(pp); } else { error = dmu_read_uio_dbuf(sa_get_db(zp->z_sa_hdl), uio, bytes); } len -= bytes; off = 0; if (error) break; } return (error); } offset_t zfs_read_chunk_size = 1024 * 1024; /* Tunable */ /* * Read bytes from specified file into supplied buffer. * * IN: vp - vnode of file to be read from. * uio - structure supplying read location, range info, * and return buffer. * ioflag - SYNC flags; used to provide FRSYNC semantics. * cr - credentials of caller. * ct - caller context * * OUT: uio - updated offset and range, buffer filled. * * RETURN: 0 on success, error code on failure. * * Side Effects: * vp - atime updated if byte count > 0 */ /* ARGSUSED */ static int zfs_read(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct) { znode_t *zp = VTOZ(vp); zfsvfs_t *zfsvfs = zp->z_zfsvfs; ssize_t n, nbytes; int error = 0; rl_t *rl; xuio_t *xuio = NULL; ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(zp); if (zp->z_pflags & ZFS_AV_QUARANTINED) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EACCES)); } /* * Validate file offset */ if (uio->uio_loffset < (offset_t)0) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EINVAL)); } /* * Fasttrack empty reads */ if (uio->uio_resid == 0) { ZFS_EXIT(zfsvfs); return (0); } /* * Check for mandatory locks */ if (MANDMODE(zp->z_mode)) { if (error = chklock(vp, FREAD, uio->uio_loffset, uio->uio_resid, uio->uio_fmode, ct)) { ZFS_EXIT(zfsvfs); return (error); } } /* * If we're in FRSYNC mode, sync out this znode before reading it. */ if (ioflag & FRSYNC || zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) zil_commit(zfsvfs->z_log, zp->z_id); /* * Lock the range against changes. */ rl = zfs_range_lock(zp, uio->uio_loffset, uio->uio_resid, RL_READER); /* * If we are reading past end-of-file we can skip * to the end; but we might still need to set atime. */ if (uio->uio_loffset >= zp->z_size) { error = 0; goto out; } ASSERT(uio->uio_loffset < zp->z_size); n = MIN(uio->uio_resid, zp->z_size - uio->uio_loffset); if ((uio->uio_extflg == UIO_XUIO) && (((xuio_t *)uio)->xu_type == UIOTYPE_ZEROCOPY)) { int nblk; int blksz = zp->z_blksz; uint64_t offset = uio->uio_loffset; xuio = (xuio_t *)uio; if ((ISP2(blksz))) { nblk = (P2ROUNDUP(offset + n, blksz) - P2ALIGN(offset, blksz)) / blksz; } else { ASSERT(offset + n <= blksz); nblk = 1; } (void) dmu_xuio_init(xuio, nblk); if (vn_has_cached_data(vp)) { /* * For simplicity, we always allocate a full buffer * even if we only expect to read a portion of a block. */ while (--nblk >= 0) { (void) dmu_xuio_add(xuio, dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), blksz), 0, blksz); } } } while (n > 0) { nbytes = MIN(n, zfs_read_chunk_size - P2PHASE(uio->uio_loffset, zfs_read_chunk_size)); if (vn_has_cached_data(vp)) { error = mappedread(vp, nbytes, uio); } else { error = dmu_read_uio_dbuf(sa_get_db(zp->z_sa_hdl), uio, nbytes); } if (error) { /* convert checksum errors into IO errors */ if (error == ECKSUM) error = SET_ERROR(EIO); break; } n -= nbytes; } out: zfs_range_unlock(rl); ZFS_ACCESSTIME_STAMP(zfsvfs, zp); ZFS_EXIT(zfsvfs); return (error); } /* * Write the bytes to a file. * * IN: vp - vnode of file to be written to. * uio - structure supplying write location, range info, * and data buffer. * ioflag - FAPPEND, FSYNC, and/or FDSYNC. FAPPEND is * set if in append mode. * cr - credentials of caller. * ct - caller context (NFS/CIFS fem monitor only) * * OUT: uio - updated offset and range. * * RETURN: 0 on success, error code on failure. * * Timestamps: * vp - ctime|mtime updated if byte count > 0 */ /* ARGSUSED */ static int zfs_write(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct) { znode_t *zp = VTOZ(vp); rlim64_t limit = uio->uio_llimit; ssize_t start_resid = uio->uio_resid; ssize_t tx_bytes; uint64_t end_size; dmu_tx_t *tx; zfsvfs_t *zfsvfs = zp->z_zfsvfs; zilog_t *zilog; offset_t woff; ssize_t n, nbytes; rl_t *rl; int max_blksz = zfsvfs->z_max_blksz; int error = 0; arc_buf_t *abuf; iovec_t *aiov = NULL; xuio_t *xuio = NULL; int i_iov = 0; int iovcnt = uio->uio_iovcnt; iovec_t *iovp = uio->uio_iov; int write_eof; int count = 0; sa_bulk_attr_t bulk[4]; uint64_t mtime[2], ctime[2]; /* * Fasttrack empty write */ n = start_resid; if (n == 0) return (0); if (limit == RLIM64_INFINITY || limit > MAXOFFSET_T) limit = MAXOFFSET_T; ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(zp); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, &zp->z_size, 8); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, &zp->z_pflags, 8); /* * In a case vp->v_vfsp != zp->z_zfsvfs->z_vfs (e.g. snapshots) our * callers might not be able to detect properly that we are read-only, * so check it explicitly here. */ if (zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EROFS)); } /* * If immutable or not appending then return EPERM. * Intentionally allow ZFS_READONLY through here. * See zfs_zaccess_common() */ if ((zp->z_pflags & ZFS_IMMUTABLE) || ((zp->z_pflags & ZFS_APPENDONLY) && !(ioflag & FAPPEND) && (uio->uio_loffset < zp->z_size))) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EPERM)); } zilog = zfsvfs->z_log; /* * Validate file offset */ woff = ioflag & FAPPEND ? zp->z_size : uio->uio_loffset; if (woff < 0) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EINVAL)); } /* * Check for mandatory locks before calling zfs_range_lock() * in order to prevent a deadlock with locks set via fcntl(). */ if (MANDMODE((mode_t)zp->z_mode) && (error = chklock(vp, FWRITE, woff, n, uio->uio_fmode, ct)) != 0) { ZFS_EXIT(zfsvfs); return (error); } /* * Pre-fault the pages to ensure slow (eg NFS) pages * don't hold up txg. * Skip this if uio contains loaned arc_buf. */ if ((uio->uio_extflg == UIO_XUIO) && (((xuio_t *)uio)->xu_type == UIOTYPE_ZEROCOPY)) xuio = (xuio_t *)uio; else uio_prefaultpages(MIN(n, max_blksz), uio); /* * If in append mode, set the io offset pointer to eof. */ if (ioflag & FAPPEND) { /* * Obtain an appending range lock to guarantee file append * semantics. We reset the write offset once we have the lock. */ rl = zfs_range_lock(zp, 0, n, RL_APPEND); woff = rl->r_off; if (rl->r_len == UINT64_MAX) { /* * We overlocked the file because this write will cause * the file block size to increase. * Note that zp_size cannot change with this lock held. */ woff = zp->z_size; } uio->uio_loffset = woff; } else { /* * Note that if the file block size will change as a result of * this write, then this range lock will lock the entire file * so that we can re-write the block safely. */ rl = zfs_range_lock(zp, woff, n, RL_WRITER); } if (woff >= limit) { zfs_range_unlock(rl); ZFS_EXIT(zfsvfs); return (SET_ERROR(EFBIG)); } if ((woff + n) > limit || woff > (limit - n)) n = limit - woff; /* Will this write extend the file length? */ write_eof = (woff + n > zp->z_size); end_size = MAX(zp->z_size, woff + n); /* * Write the file in reasonable size chunks. Each chunk is written * in a separate transaction; this keeps the intent log records small * and allows us to do more fine-grained space accounting. */ while (n > 0) { abuf = NULL; woff = uio->uio_loffset; if (zfs_owner_overquota(zfsvfs, zp, B_FALSE) || zfs_owner_overquota(zfsvfs, zp, B_TRUE)) { if (abuf != NULL) dmu_return_arcbuf(abuf); error = SET_ERROR(EDQUOT); break; } if (xuio && abuf == NULL) { ASSERT(i_iov < iovcnt); aiov = &iovp[i_iov]; abuf = dmu_xuio_arcbuf(xuio, i_iov); dmu_xuio_clear(xuio, i_iov); DTRACE_PROBE3(zfs_cp_write, int, i_iov, iovec_t *, aiov, arc_buf_t *, abuf); ASSERT((aiov->iov_base == abuf->b_data) || ((char *)aiov->iov_base - (char *)abuf->b_data + aiov->iov_len == arc_buf_size(abuf))); i_iov++; } else if (abuf == NULL && n >= max_blksz && woff >= zp->z_size && P2PHASE(woff, max_blksz) == 0 && zp->z_blksz == max_blksz) { /* * This write covers a full block. "Borrow" a buffer * from the dmu so that we can fill it before we enter * a transaction. This avoids the possibility of * holding up the transaction if the data copy hangs * up on a pagefault (e.g., from an NFS server mapping). */ size_t cbytes; abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), max_blksz); ASSERT(abuf != NULL); ASSERT(arc_buf_size(abuf) == max_blksz); if (error = uiocopy(abuf->b_data, max_blksz, UIO_WRITE, uio, &cbytes)) { dmu_return_arcbuf(abuf); break; } ASSERT(cbytes == max_blksz); } /* * Start a transaction. */ tx = dmu_tx_create(zfsvfs->z_os); dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); dmu_tx_hold_write(tx, zp->z_id, woff, MIN(n, max_blksz)); zfs_sa_upgrade_txholds(tx, zp); error = dmu_tx_assign(tx, TXG_WAIT); if (error) { dmu_tx_abort(tx); if (abuf != NULL) dmu_return_arcbuf(abuf); break; } /* * If zfs_range_lock() over-locked we grow the blocksize * and then reduce the lock range. This will only happen * on the first iteration since zfs_range_reduce() will * shrink down r_len to the appropriate size. */ if (rl->r_len == UINT64_MAX) { uint64_t new_blksz; if (zp->z_blksz > max_blksz) { /* * File's blocksize is already larger than the * "recordsize" property. Only let it grow to * the next power of 2. */ ASSERT(!ISP2(zp->z_blksz)); new_blksz = MIN(end_size, 1 << highbit64(zp->z_blksz)); } else { new_blksz = MIN(end_size, max_blksz); } zfs_grow_blocksize(zp, new_blksz, tx); zfs_range_reduce(rl, woff, n); } /* * XXX - should we really limit each write to z_max_blksz? * Perhaps we should use SPA_MAXBLOCKSIZE chunks? */ nbytes = MIN(n, max_blksz - P2PHASE(woff, max_blksz)); if (abuf == NULL) { tx_bytes = uio->uio_resid; error = dmu_write_uio_dbuf(sa_get_db(zp->z_sa_hdl), uio, nbytes, tx); tx_bytes -= uio->uio_resid; } else { tx_bytes = nbytes; ASSERT(xuio == NULL || tx_bytes == aiov->iov_len); /* * If this is not a full block write, but we are * extending the file past EOF and this data starts * block-aligned, use assign_arcbuf(). Otherwise, * write via dmu_write(). */ if (tx_bytes < max_blksz && (!write_eof || aiov->iov_base != abuf->b_data)) { ASSERT(xuio); dmu_write(zfsvfs->z_os, zp->z_id, woff, aiov->iov_len, aiov->iov_base, tx); dmu_return_arcbuf(abuf); xuio_stat_wbuf_copied(); } else { ASSERT(xuio || tx_bytes == max_blksz); dmu_assign_arcbuf(sa_get_db(zp->z_sa_hdl), woff, abuf, tx); } ASSERT(tx_bytes <= uio->uio_resid); uioskip(uio, tx_bytes); } if (tx_bytes && vn_has_cached_data(vp)) { update_pages(vp, woff, tx_bytes, zfsvfs->z_os, zp->z_id); } /* * If we made no progress, we're done. If we made even * partial progress, update the znode and ZIL accordingly. */ if (tx_bytes == 0) { (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs), (void *)&zp->z_size, sizeof (uint64_t), tx); dmu_tx_commit(tx); ASSERT(error != 0); break; } /* * Clear Set-UID/Set-GID bits on successful write if not * privileged and at least one of the excute bits is set. * * It would be nice to to this after all writes have * been done, but that would still expose the ISUID/ISGID * to another app after the partial write is committed. * * Note: we don't call zfs_fuid_map_id() here because * user 0 is not an ephemeral uid. */ mutex_enter(&zp->z_acl_lock); if ((zp->z_mode & (S_IXUSR | (S_IXUSR >> 3) | (S_IXUSR >> 6))) != 0 && (zp->z_mode & (S_ISUID | S_ISGID)) != 0 && secpolicy_vnode_setid_retain(cr, (zp->z_mode & S_ISUID) != 0 && zp->z_uid == 0) != 0) { uint64_t newmode; zp->z_mode &= ~(S_ISUID | S_ISGID); newmode = zp->z_mode; (void) sa_update(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), (void *)&newmode, sizeof (uint64_t), tx); } mutex_exit(&zp->z_acl_lock); zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, B_TRUE); /* * Update the file size (zp_size) if it has changed; * account for possible concurrent updates. */ while ((end_size = zp->z_size) < uio->uio_loffset) { (void) atomic_cas_64(&zp->z_size, end_size, uio->uio_loffset); ASSERT(error == 0); } /* * If we are replaying and eof is non zero then force * the file size to the specified eof. Note, there's no * concurrency during replay. */ if (zfsvfs->z_replay && zfsvfs->z_replay_eof != 0) zp->z_size = zfsvfs->z_replay_eof; error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); zfs_log_write(zilog, tx, TX_WRITE, zp, woff, tx_bytes, ioflag); dmu_tx_commit(tx); if (error != 0) break; ASSERT(tx_bytes == nbytes); n -= nbytes; if (!xuio && n > 0) uio_prefaultpages(MIN(n, max_blksz), uio); } zfs_range_unlock(rl); /* * If we're in replay mode, or we made no progress, return error. * Otherwise, it's at least a partial write, so it's successful. */ if (zfsvfs->z_replay || uio->uio_resid == start_resid) { ZFS_EXIT(zfsvfs); return (error); } if (ioflag & (FSYNC | FDSYNC) || zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) zil_commit(zilog, zp->z_id); ZFS_EXIT(zfsvfs); return (0); } void zfs_get_done(zgd_t *zgd, int error) { znode_t *zp = zgd->zgd_private; objset_t *os = zp->z_zfsvfs->z_os; if (zgd->zgd_db) dmu_buf_rele(zgd->zgd_db, zgd); zfs_range_unlock(zgd->zgd_rl); /* * Release the vnode asynchronously as we currently have the * txg stopped from syncing. */ VN_RELE_ASYNC(ZTOV(zp), dsl_pool_vnrele_taskq(dmu_objset_pool(os))); if (error == 0 && zgd->zgd_bp) zil_lwb_add_block(zgd->zgd_lwb, zgd->zgd_bp); kmem_free(zgd, sizeof (zgd_t)); } #ifdef DEBUG static int zil_fault_io = 0; #endif /* * Get data to generate a TX_WRITE intent log record. */ int zfs_get_data(void *arg, lr_write_t *lr, char *buf, struct lwb *lwb, zio_t *zio) { zfsvfs_t *zfsvfs = arg; objset_t *os = zfsvfs->z_os; znode_t *zp; uint64_t object = lr->lr_foid; uint64_t offset = lr->lr_offset; uint64_t size = lr->lr_length; dmu_buf_t *db; zgd_t *zgd; int error = 0; ASSERT3P(lwb, !=, NULL); ASSERT3P(zio, !=, NULL); ASSERT3U(size, !=, 0); /* * Nothing to do if the file has been removed */ if (zfs_zget(zfsvfs, object, &zp) != 0) return (SET_ERROR(ENOENT)); if (zp->z_unlinked) { /* * Release the vnode asynchronously as we currently have the * txg stopped from syncing. */ VN_RELE_ASYNC(ZTOV(zp), dsl_pool_vnrele_taskq(dmu_objset_pool(os))); return (SET_ERROR(ENOENT)); } zgd = (zgd_t *)kmem_zalloc(sizeof (zgd_t), KM_SLEEP); zgd->zgd_lwb = lwb; zgd->zgd_private = zp; /* * Write records come in two flavors: immediate and indirect. * For small writes it's cheaper to store the data with the * log record (immediate); for large writes it's cheaper to * sync the data and get a pointer to it (indirect) so that * we don't have to write the data twice. */ if (buf != NULL) { /* immediate write */ zgd->zgd_rl = zfs_range_lock(zp, offset, size, RL_READER); /* test for truncation needs to be done while range locked */ if (offset >= zp->z_size) { error = SET_ERROR(ENOENT); } else { error = dmu_read(os, object, offset, size, buf, DMU_READ_NO_PREFETCH); } ASSERT(error == 0 || error == ENOENT); } else { /* indirect write */ /* * Have to lock the whole block to ensure when it's * written out and its checksum is being calculated * that no one can change the data. We need to re-check * blocksize after we get the lock in case it's changed! */ for (;;) { uint64_t blkoff; size = zp->z_blksz; blkoff = ISP2(size) ? P2PHASE(offset, size) : offset; offset -= blkoff; zgd->zgd_rl = zfs_range_lock(zp, offset, size, RL_READER); if (zp->z_blksz == size) break; offset += blkoff; zfs_range_unlock(zgd->zgd_rl); } /* test for truncation needs to be done while range locked */ if (lr->lr_offset >= zp->z_size) error = SET_ERROR(ENOENT); #ifdef DEBUG if (zil_fault_io) { error = SET_ERROR(EIO); zil_fault_io = 0; } #endif if (error == 0) error = dmu_buf_hold(os, object, offset, zgd, &db, DMU_READ_NO_PREFETCH); if (error == 0) { blkptr_t *bp = &lr->lr_blkptr; zgd->zgd_db = db; zgd->zgd_bp = bp; ASSERT(db->db_offset == offset); ASSERT(db->db_size == size); error = dmu_sync(zio, lr->lr_common.lrc_txg, zfs_get_done, zgd); ASSERT(error || lr->lr_length <= size); /* * On success, we need to wait for the write I/O * initiated by dmu_sync() to complete before we can * release this dbuf. We will finish everything up * in the zfs_get_done() callback. */ if (error == 0) return (0); if (error == EALREADY) { lr->lr_common.lrc_txtype = TX_WRITE2; /* * TX_WRITE2 relies on the data previously * written by the TX_WRITE that caused * EALREADY. We zero out the BP because * it is the old, currently-on-disk BP, * so there's no need to zio_flush() its * vdevs (flushing would needlesly hurt * performance, and doesn't work on * indirect vdevs). */ zgd->zgd_bp = NULL; BP_ZERO(bp); error = 0; } } } zfs_get_done(zgd, error); return (error); } /*ARGSUSED*/ static int zfs_access(vnode_t *vp, int mode, int flag, cred_t *cr, caller_context_t *ct) { znode_t *zp = VTOZ(vp); zfsvfs_t *zfsvfs = zp->z_zfsvfs; int error; ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(zp); if (flag & V_ACE_MASK) error = zfs_zaccess(zp, mode, flag, B_FALSE, cr); else error = zfs_zaccess_rwx(zp, mode, flag, cr); ZFS_EXIT(zfsvfs); return (error); } /* * If vnode is for a device return a specfs vnode instead. */ static int specvp_check(vnode_t **vpp, cred_t *cr) { int error = 0; if (IS_DEVVP(*vpp)) { struct vnode *svp; svp = specvp(*vpp, (*vpp)->v_rdev, (*vpp)->v_type, cr); VN_RELE(*vpp); if (svp == NULL) error = SET_ERROR(ENOSYS); *vpp = svp; } return (error); } /* * Lookup an entry in a directory, or an extended attribute directory. * If it exists, return a held vnode reference for it. * * IN: dvp - vnode of directory to search. * nm - name of entry to lookup. * pnp - full pathname to lookup [UNUSED]. * flags - LOOKUP_XATTR set if looking for an attribute. * rdir - root directory vnode [UNUSED]. * cr - credentials of caller. * ct - caller context * direntflags - directory lookup flags * realpnp - returned pathname. * * OUT: vpp - vnode of located entry, NULL if not found. * * RETURN: 0 on success, error code on failure. * * Timestamps: * NA */ /* ARGSUSED */ static int zfs_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, struct pathname *pnp, int flags, vnode_t *rdir, cred_t *cr, caller_context_t *ct, int *direntflags, pathname_t *realpnp) { znode_t *zdp = VTOZ(dvp); zfsvfs_t *zfsvfs = zdp->z_zfsvfs; int error = 0; /* * Fast path lookup, however we must skip DNLC lookup * for case folding or normalizing lookups because the * DNLC code only stores the passed in name. This means * creating 'a' and removing 'A' on a case insensitive * file system would work, but DNLC still thinks 'a' * exists and won't let you create it again on the next * pass through fast path. */ if (!(flags & (LOOKUP_XATTR | FIGNORECASE))) { if (dvp->v_type != VDIR) { return (SET_ERROR(ENOTDIR)); } else if (zdp->z_sa_hdl == NULL) { return (SET_ERROR(EIO)); } if (nm[0] == 0 || (nm[0] == '.' && nm[1] == '\0')) { error = zfs_fastaccesschk_execute(zdp, cr); if (!error) { *vpp = dvp; VN_HOLD(*vpp); return (0); } return (error); } else if (!zdp->z_zfsvfs->z_norm && (zdp->z_zfsvfs->z_case == ZFS_CASE_SENSITIVE)) { vnode_t *tvp = dnlc_lookup(dvp, nm); if (tvp) { error = zfs_fastaccesschk_execute(zdp, cr); if (error) { VN_RELE(tvp); return (error); } if (tvp == DNLC_NO_VNODE) { VN_RELE(tvp); return (SET_ERROR(ENOENT)); } else { *vpp = tvp; return (specvp_check(vpp, cr)); } } } } DTRACE_PROBE2(zfs__fastpath__lookup__miss, vnode_t *, dvp, char *, nm); ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(zdp); *vpp = NULL; if (flags & LOOKUP_XATTR) { /* * If the xattr property is off, refuse the lookup request. */ if (!(zfsvfs->z_vfs->vfs_flag & VFS_XATTR)) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EINVAL)); } /* * We don't allow recursive attributes.. * Maybe someday we will. */ if (zdp->z_pflags & ZFS_XATTR) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EINVAL)); } if (error = zfs_get_xattrdir(VTOZ(dvp), vpp, cr, flags)) { ZFS_EXIT(zfsvfs); return (error); } /* * Do we have permission to get into attribute directory? */ if (error = zfs_zaccess(VTOZ(*vpp), ACE_EXECUTE, 0, B_FALSE, cr)) { VN_RELE(*vpp); *vpp = NULL; } ZFS_EXIT(zfsvfs); return (error); } if (dvp->v_type != VDIR) { ZFS_EXIT(zfsvfs); return (SET_ERROR(ENOTDIR)); } /* * Check accessibility of directory. */ if (error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr)) { ZFS_EXIT(zfsvfs); return (error); } if (zfsvfs->z_utf8 && u8_validate(nm, strlen(nm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EILSEQ)); } error = zfs_dirlook(zdp, nm, vpp, flags, direntflags, realpnp); if (error == 0) error = specvp_check(vpp, cr); ZFS_EXIT(zfsvfs); return (error); } /* * Attempt to create a new entry in a directory. If the entry * already exists, truncate the file if permissible, else return * an error. Return the vp of the created or trunc'd file. * * IN: dvp - vnode of directory to put new file entry in. * name - name of new file entry. * vap - attributes of new file. * excl - flag indicating exclusive or non-exclusive mode. * mode - mode to open file with. * cr - credentials of caller. * flag - large file flag [UNUSED]. * ct - caller context * vsecp - ACL to be set * * OUT: vpp - vnode of created or trunc'd entry. * * RETURN: 0 on success, error code on failure. * * Timestamps: * dvp - ctime|mtime updated if new entry created * vp - ctime|mtime always, atime if new */ /* ARGSUSED */ static int zfs_create(vnode_t *dvp, char *name, vattr_t *vap, vcexcl_t excl, int mode, vnode_t **vpp, cred_t *cr, int flag, caller_context_t *ct, vsecattr_t *vsecp) { znode_t *zp, *dzp = VTOZ(dvp); zfsvfs_t *zfsvfs = dzp->z_zfsvfs; zilog_t *zilog; objset_t *os; zfs_dirlock_t *dl; dmu_tx_t *tx; int error; ksid_t *ksid; uid_t uid; gid_t gid = crgetgid(cr); zfs_acl_ids_t acl_ids; boolean_t fuid_dirtied; boolean_t have_acl = B_FALSE; boolean_t waited = B_FALSE; /* * If we have an ephemeral id, ACL, or XVATTR then * make sure file system is at proper version */ ksid = crgetsid(cr, KSID_OWNER); if (ksid) uid = ksid_getid(ksid); else uid = crgetuid(cr); if (zfsvfs->z_use_fuids == B_FALSE && (vsecp || (vap->va_mask & AT_XVATTR) || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) return (SET_ERROR(EINVAL)); ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(dzp); os = zfsvfs->z_os; zilog = zfsvfs->z_log; if (zfsvfs->z_utf8 && u8_validate(name, strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EILSEQ)); } if (vap->va_mask & AT_XVATTR) { if ((error = secpolicy_xvattr((xvattr_t *)vap, crgetuid(cr), cr, vap->va_type)) != 0) { ZFS_EXIT(zfsvfs); return (error); } } top: *vpp = NULL; if ((vap->va_mode & VSVTX) && secpolicy_vnode_stky_modify(cr)) vap->va_mode &= ~VSVTX; if (*name == '\0') { /* * Null component name refers to the directory itself. */ VN_HOLD(dvp); zp = dzp; dl = NULL; error = 0; } else { /* possible VN_HOLD(zp) */ int zflg = 0; if (flag & FIGNORECASE) zflg |= ZCILOOK; error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, NULL); if (error) { if (have_acl) zfs_acl_ids_free(&acl_ids); if (strcmp(name, "..") == 0) error = SET_ERROR(EISDIR); ZFS_EXIT(zfsvfs); return (error); } } if (zp == NULL) { uint64_t txtype; /* * Create a new file object and update the directory * to reference it. */ if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) { if (have_acl) zfs_acl_ids_free(&acl_ids); goto out; } /* * We only support the creation of regular files in * extended attribute directories. */ if ((dzp->z_pflags & ZFS_XATTR) && (vap->va_type != VREG)) { if (have_acl) zfs_acl_ids_free(&acl_ids); error = SET_ERROR(EINVAL); goto out; } if (!have_acl && (error = zfs_acl_ids_create(dzp, 0, vap, cr, vsecp, &acl_ids)) != 0) goto out; have_acl = B_TRUE; if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) { zfs_acl_ids_free(&acl_ids); error = SET_ERROR(EDQUOT); goto out; } tx = dmu_tx_create(os); dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + ZFS_SA_BASE_ATTR_SIZE); fuid_dirtied = zfsvfs->z_fuid_dirty; if (fuid_dirtied) zfs_fuid_txhold(zfsvfs, tx); dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, acl_ids.z_aclp->z_acl_bytes); } - error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); + error = dmu_tx_assign(tx, + (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); if (error) { zfs_dirent_unlock(dl); if (error == ERESTART) { waited = B_TRUE; dmu_tx_wait(tx); dmu_tx_abort(tx); goto top; } zfs_acl_ids_free(&acl_ids); dmu_tx_abort(tx); ZFS_EXIT(zfsvfs); return (error); } zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); if (fuid_dirtied) zfs_fuid_sync(zfsvfs, tx); (void) zfs_link_create(dl, zp, tx, ZNEW); txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap); if (flag & FIGNORECASE) txtype |= TX_CI; zfs_log_create(zilog, tx, txtype, dzp, zp, name, vsecp, acl_ids.z_fuidp, vap); zfs_acl_ids_free(&acl_ids); dmu_tx_commit(tx); } else { int aflags = (flag & FAPPEND) ? V_APPEND : 0; if (have_acl) zfs_acl_ids_free(&acl_ids); have_acl = B_FALSE; /* * A directory entry already exists for this name. */ /* * Can't truncate an existing file if in exclusive mode. */ if (excl == EXCL) { error = SET_ERROR(EEXIST); goto out; } /* * Can't open a directory for writing. */ if ((ZTOV(zp)->v_type == VDIR) && (mode & S_IWRITE)) { error = SET_ERROR(EISDIR); goto out; } /* * Verify requested access to file. */ if (mode && (error = zfs_zaccess_rwx(zp, mode, aflags, cr))) { goto out; } mutex_enter(&dzp->z_lock); dzp->z_seq++; mutex_exit(&dzp->z_lock); /* * Truncate regular files if requested. */ if ((ZTOV(zp)->v_type == VREG) && (vap->va_mask & AT_SIZE) && (vap->va_size == 0)) { /* we can't hold any locks when calling zfs_freesp() */ zfs_dirent_unlock(dl); dl = NULL; error = zfs_freesp(zp, 0, 0, mode, TRUE); if (error == 0) { vnevent_create(ZTOV(zp), ct); } } } out: if (dl) zfs_dirent_unlock(dl); if (error) { if (zp) VN_RELE(ZTOV(zp)); } else { *vpp = ZTOV(zp); error = specvp_check(vpp, cr); } if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) zil_commit(zilog, 0); ZFS_EXIT(zfsvfs); return (error); } /* * Remove an entry from a directory. * * IN: dvp - vnode of directory to remove entry from. * name - name of entry to remove. * cr - credentials of caller. * ct - caller context * flags - case flags * * RETURN: 0 on success, error code on failure. * * Timestamps: * dvp - ctime|mtime * vp - ctime (if nlink > 0) */ uint64_t null_xattr = 0; /*ARGSUSED*/ static int zfs_remove(vnode_t *dvp, char *name, cred_t *cr, caller_context_t *ct, int flags) { znode_t *zp, *dzp = VTOZ(dvp); znode_t *xzp; vnode_t *vp; zfsvfs_t *zfsvfs = dzp->z_zfsvfs; zilog_t *zilog; uint64_t acl_obj, xattr_obj; uint64_t xattr_obj_unlinked = 0; uint64_t obj = 0; zfs_dirlock_t *dl; dmu_tx_t *tx; boolean_t may_delete_now, delete_now = FALSE; boolean_t unlinked, toobig = FALSE; uint64_t txtype; pathname_t *realnmp = NULL; pathname_t realnm; int error; int zflg = ZEXISTS; boolean_t waited = B_FALSE; ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(dzp); zilog = zfsvfs->z_log; if (flags & FIGNORECASE) { zflg |= ZCILOOK; pn_alloc(&realnm); realnmp = &realnm; } top: xattr_obj = 0; xzp = NULL; /* * Attempt to lock directory; fail if entry doesn't exist. */ if (error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, realnmp)) { if (realnmp) pn_free(realnmp); ZFS_EXIT(zfsvfs); return (error); } vp = ZTOV(zp); if (error = zfs_zaccess_delete(dzp, zp, cr)) { goto out; } /* * Need to use rmdir for removing directories. */ if (vp->v_type == VDIR) { error = SET_ERROR(EPERM); goto out; } vnevent_remove(vp, dvp, name, ct); if (realnmp) dnlc_remove(dvp, realnmp->pn_buf); else dnlc_remove(dvp, name); mutex_enter(&vp->v_lock); may_delete_now = vp->v_count == 1 && !vn_has_cached_data(vp); mutex_exit(&vp->v_lock); /* * We may delete the znode now, or we may put it in the unlinked set; * it depends on whether we're the last link, and on whether there are * other holds on the vnode. So we dmu_tx_hold() the right things to * allow for either case. */ obj = zp->z_id; tx = dmu_tx_create(zfsvfs->z_os); dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); zfs_sa_upgrade_txholds(tx, zp); zfs_sa_upgrade_txholds(tx, dzp); if (may_delete_now) { toobig = zp->z_size > zp->z_blksz * DMU_MAX_DELETEBLKCNT; /* if the file is too big, only hold_free a token amount */ dmu_tx_hold_free(tx, zp->z_id, 0, (toobig ? DMU_MAX_ACCESS : DMU_OBJECT_END)); } /* are there any extended attributes? */ error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &xattr_obj, sizeof (xattr_obj)); if (error == 0 && xattr_obj) { error = zfs_zget(zfsvfs, xattr_obj, &xzp); ASSERT0(error); dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE); } mutex_enter(&zp->z_lock); if ((acl_obj = zfs_external_acl(zp)) != 0 && may_delete_now) dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END); mutex_exit(&zp->z_lock); /* charge as an update -- would be nice not to charge at all */ dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); /* * Mark this transaction as typically resulting in a net free of space */ dmu_tx_mark_netfree(tx); - error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); + error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); if (error) { zfs_dirent_unlock(dl); VN_RELE(vp); if (xzp) VN_RELE(ZTOV(xzp)); if (error == ERESTART) { waited = B_TRUE; dmu_tx_wait(tx); dmu_tx_abort(tx); goto top; } if (realnmp) pn_free(realnmp); dmu_tx_abort(tx); ZFS_EXIT(zfsvfs); return (error); } /* * Remove the directory entry. */ error = zfs_link_destroy(dl, zp, tx, zflg, &unlinked); if (error) { dmu_tx_commit(tx); goto out; } if (unlinked) { /* * Hold z_lock so that we can make sure that the ACL obj * hasn't changed. Could have been deleted due to * zfs_sa_upgrade(). */ mutex_enter(&zp->z_lock); mutex_enter(&vp->v_lock); (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &xattr_obj_unlinked, sizeof (xattr_obj_unlinked)); delete_now = may_delete_now && !toobig && vp->v_count == 1 && !vn_has_cached_data(vp) && xattr_obj == xattr_obj_unlinked && zfs_external_acl(zp) == acl_obj; mutex_exit(&vp->v_lock); } if (delete_now) { if (xattr_obj_unlinked) { ASSERT3U(xzp->z_links, ==, 2); mutex_enter(&xzp->z_lock); xzp->z_unlinked = 1; xzp->z_links = 0; error = sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs), &xzp->z_links, sizeof (xzp->z_links), tx); ASSERT3U(error, ==, 0); mutex_exit(&xzp->z_lock); zfs_unlinked_add(xzp, tx); if (zp->z_is_sa) error = sa_remove(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), tx); else error = sa_update(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &null_xattr, sizeof (uint64_t), tx); ASSERT0(error); } mutex_enter(&vp->v_lock); VN_RELE_LOCKED(vp); ASSERT0(vp->v_count); mutex_exit(&vp->v_lock); mutex_exit(&zp->z_lock); zfs_znode_delete(zp, tx); } else if (unlinked) { mutex_exit(&zp->z_lock); zfs_unlinked_add(zp, tx); } txtype = TX_REMOVE; if (flags & FIGNORECASE) txtype |= TX_CI; zfs_log_remove(zilog, tx, txtype, dzp, name, obj); dmu_tx_commit(tx); out: if (realnmp) pn_free(realnmp); zfs_dirent_unlock(dl); if (!delete_now) VN_RELE(vp); if (xzp) VN_RELE(ZTOV(xzp)); if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) zil_commit(zilog, 0); ZFS_EXIT(zfsvfs); return (error); } /* * Create a new directory and insert it into dvp using the name * provided. Return a pointer to the inserted directory. * * IN: dvp - vnode of directory to add subdir to. * dirname - name of new directory. * vap - attributes of new directory. * cr - credentials of caller. * ct - caller context * flags - case flags * vsecp - ACL to be set * * OUT: vpp - vnode of created directory. * * RETURN: 0 on success, error code on failure. * * Timestamps: * dvp - ctime|mtime updated * vp - ctime|mtime|atime updated */ /*ARGSUSED*/ static int zfs_mkdir(vnode_t *dvp, char *dirname, vattr_t *vap, vnode_t **vpp, cred_t *cr, caller_context_t *ct, int flags, vsecattr_t *vsecp) { znode_t *zp, *dzp = VTOZ(dvp); zfsvfs_t *zfsvfs = dzp->z_zfsvfs; zilog_t *zilog; zfs_dirlock_t *dl; uint64_t txtype; dmu_tx_t *tx; int error; int zf = ZNEW; ksid_t *ksid; uid_t uid; gid_t gid = crgetgid(cr); zfs_acl_ids_t acl_ids; boolean_t fuid_dirtied; boolean_t waited = B_FALSE; ASSERT(vap->va_type == VDIR); /* * If we have an ephemeral id, ACL, or XVATTR then * make sure file system is at proper version */ ksid = crgetsid(cr, KSID_OWNER); if (ksid) uid = ksid_getid(ksid); else uid = crgetuid(cr); if (zfsvfs->z_use_fuids == B_FALSE && (vsecp || (vap->va_mask & AT_XVATTR) || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) return (SET_ERROR(EINVAL)); ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(dzp); zilog = zfsvfs->z_log; if (dzp->z_pflags & ZFS_XATTR) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EINVAL)); } if (zfsvfs->z_utf8 && u8_validate(dirname, strlen(dirname), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EILSEQ)); } if (flags & FIGNORECASE) zf |= ZCILOOK; if (vap->va_mask & AT_XVATTR) { if ((error = secpolicy_xvattr((xvattr_t *)vap, crgetuid(cr), cr, vap->va_type)) != 0) { ZFS_EXIT(zfsvfs); return (error); } } if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, vsecp, &acl_ids)) != 0) { ZFS_EXIT(zfsvfs); return (error); } /* * First make sure the new directory doesn't exist. * * Existence is checked first to make sure we don't return * EACCES instead of EEXIST which can cause some applications * to fail. */ top: *vpp = NULL; if (error = zfs_dirent_lock(&dl, dzp, dirname, &zp, zf, NULL, NULL)) { zfs_acl_ids_free(&acl_ids); ZFS_EXIT(zfsvfs); return (error); } if (error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr)) { zfs_acl_ids_free(&acl_ids); zfs_dirent_unlock(dl); ZFS_EXIT(zfsvfs); return (error); } if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) { zfs_acl_ids_free(&acl_ids); zfs_dirent_unlock(dl); ZFS_EXIT(zfsvfs); return (SET_ERROR(EDQUOT)); } /* * Add a new entry to the directory. */ tx = dmu_tx_create(zfsvfs->z_os); dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname); dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); fuid_dirtied = zfsvfs->z_fuid_dirty; if (fuid_dirtied) zfs_fuid_txhold(zfsvfs, tx); if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, acl_ids.z_aclp->z_acl_bytes); } dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + ZFS_SA_BASE_ATTR_SIZE); - error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); + error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); if (error) { zfs_dirent_unlock(dl); if (error == ERESTART) { waited = B_TRUE; dmu_tx_wait(tx); dmu_tx_abort(tx); goto top; } zfs_acl_ids_free(&acl_ids); dmu_tx_abort(tx); ZFS_EXIT(zfsvfs); return (error); } /* * Create new node. */ zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); if (fuid_dirtied) zfs_fuid_sync(zfsvfs, tx); /* * Now put new name in parent dir. */ (void) zfs_link_create(dl, zp, tx, ZNEW); *vpp = ZTOV(zp); txtype = zfs_log_create_txtype(Z_DIR, vsecp, vap); if (flags & FIGNORECASE) txtype |= TX_CI; zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, vsecp, acl_ids.z_fuidp, vap); zfs_acl_ids_free(&acl_ids); dmu_tx_commit(tx); zfs_dirent_unlock(dl); if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) zil_commit(zilog, 0); ZFS_EXIT(zfsvfs); return (0); } /* * Remove a directory subdir entry. If the current working * directory is the same as the subdir to be removed, the * remove will fail. * * IN: dvp - vnode of directory to remove from. * name - name of directory to be removed. * cwd - vnode of current working directory. * cr - credentials of caller. * ct - caller context * flags - case flags * * RETURN: 0 on success, error code on failure. * * Timestamps: * dvp - ctime|mtime updated */ /*ARGSUSED*/ static int zfs_rmdir(vnode_t *dvp, char *name, vnode_t *cwd, cred_t *cr, caller_context_t *ct, int flags) { znode_t *dzp = VTOZ(dvp); znode_t *zp; vnode_t *vp; zfsvfs_t *zfsvfs = dzp->z_zfsvfs; zilog_t *zilog; zfs_dirlock_t *dl; dmu_tx_t *tx; int error; int zflg = ZEXISTS; boolean_t waited = B_FALSE; ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(dzp); zilog = zfsvfs->z_log; if (flags & FIGNORECASE) zflg |= ZCILOOK; top: zp = NULL; /* * Attempt to lock directory; fail if entry doesn't exist. */ if (error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, NULL)) { ZFS_EXIT(zfsvfs); return (error); } vp = ZTOV(zp); if (error = zfs_zaccess_delete(dzp, zp, cr)) { goto out; } if (vp->v_type != VDIR) { error = SET_ERROR(ENOTDIR); goto out; } if (vp == cwd) { error = SET_ERROR(EINVAL); goto out; } vnevent_rmdir(vp, dvp, name, ct); /* * Grab a lock on the directory to make sure that noone is * trying to add (or lookup) entries while we are removing it. */ rw_enter(&zp->z_name_lock, RW_WRITER); /* * Grab a lock on the parent pointer to make sure we play well * with the treewalk and directory rename code. */ rw_enter(&zp->z_parent_lock, RW_WRITER); tx = dmu_tx_create(zfsvfs->z_os); dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); zfs_sa_upgrade_txholds(tx, zp); zfs_sa_upgrade_txholds(tx, dzp); dmu_tx_mark_netfree(tx); - error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); + error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); if (error) { rw_exit(&zp->z_parent_lock); rw_exit(&zp->z_name_lock); zfs_dirent_unlock(dl); VN_RELE(vp); if (error == ERESTART) { waited = B_TRUE; dmu_tx_wait(tx); dmu_tx_abort(tx); goto top; } dmu_tx_abort(tx); ZFS_EXIT(zfsvfs); return (error); } error = zfs_link_destroy(dl, zp, tx, zflg, NULL); if (error == 0) { uint64_t txtype = TX_RMDIR; if (flags & FIGNORECASE) txtype |= TX_CI; zfs_log_remove(zilog, tx, txtype, dzp, name, ZFS_NO_OBJECT); } dmu_tx_commit(tx); rw_exit(&zp->z_parent_lock); rw_exit(&zp->z_name_lock); out: zfs_dirent_unlock(dl); VN_RELE(vp); if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) zil_commit(zilog, 0); ZFS_EXIT(zfsvfs); return (error); } /* * Read as many directory entries as will fit into the provided * buffer from the given directory cursor position (specified in * the uio structure). * * IN: vp - vnode of directory to read. * uio - structure supplying read location, range info, * and return buffer. * cr - credentials of caller. * ct - caller context * flags - case flags * * OUT: uio - updated offset and range, buffer filled. * eofp - set to true if end-of-file detected. * * RETURN: 0 on success, error code on failure. * * Timestamps: * vp - atime updated * * Note that the low 4 bits of the cookie returned by zap is always zero. * This allows us to use the low range for "special" directory entries: * We use 0 for '.', and 1 for '..'. If this is the root of the filesystem, * we use the offset 2 for the '.zfs' directory. */ /* ARGSUSED */ static int zfs_readdir(vnode_t *vp, uio_t *uio, cred_t *cr, int *eofp, caller_context_t *ct, int flags) { znode_t *zp = VTOZ(vp); iovec_t *iovp; edirent_t *eodp; dirent64_t *odp; zfsvfs_t *zfsvfs = zp->z_zfsvfs; objset_t *os; caddr_t outbuf; size_t bufsize; zap_cursor_t zc; zap_attribute_t zap; uint_t bytes_wanted; uint64_t offset; /* must be unsigned; checks for < 1 */ uint64_t parent; int local_eof; int outcount; int error; uint8_t prefetch; boolean_t check_sysattrs; ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(zp); if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs), &parent, sizeof (parent))) != 0) { ZFS_EXIT(zfsvfs); return (error); } /* * If we are not given an eof variable, * use a local one. */ if (eofp == NULL) eofp = &local_eof; /* * Check for valid iov_len. */ if (uio->uio_iov->iov_len <= 0) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EINVAL)); } /* * Quit if directory has been removed (posix) */ if ((*eofp = zp->z_unlinked) != 0) { ZFS_EXIT(zfsvfs); return (0); } error = 0; os = zfsvfs->z_os; offset = uio->uio_loffset; prefetch = zp->z_zn_prefetch; /* * Initialize the iterator cursor. */ if (offset <= 3) { /* * Start iteration from the beginning of the directory. */ zap_cursor_init(&zc, os, zp->z_id); } else { /* * The offset is a serialized cursor. */ zap_cursor_init_serialized(&zc, os, zp->z_id, offset); } /* * Get space to change directory entries into fs independent format. */ iovp = uio->uio_iov; bytes_wanted = iovp->iov_len; if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1) { bufsize = bytes_wanted; outbuf = kmem_alloc(bufsize, KM_SLEEP); odp = (struct dirent64 *)outbuf; } else { bufsize = bytes_wanted; outbuf = NULL; odp = (struct dirent64 *)iovp->iov_base; } eodp = (struct edirent *)odp; /* * If this VFS supports the system attribute view interface; and * we're looking at an extended attribute directory; and we care * about normalization conflicts on this vfs; then we must check * for normalization conflicts with the sysattr name space. */ check_sysattrs = vfs_has_feature(vp->v_vfsp, VFSFT_SYSATTR_VIEWS) && (vp->v_flag & V_XATTRDIR) && zfsvfs->z_norm && (flags & V_RDDIR_ENTFLAGS); /* * Transform to file-system independent format */ outcount = 0; while (outcount < bytes_wanted) { ino64_t objnum; ushort_t reclen; off64_t *next = NULL; /* * Special case `.', `..', and `.zfs'. */ if (offset == 0) { (void) strcpy(zap.za_name, "."); zap.za_normalization_conflict = 0; objnum = zp->z_id; } else if (offset == 1) { (void) strcpy(zap.za_name, ".."); zap.za_normalization_conflict = 0; objnum = parent; } else if (offset == 2 && zfs_show_ctldir(zp)) { (void) strcpy(zap.za_name, ZFS_CTLDIR_NAME); zap.za_normalization_conflict = 0; objnum = ZFSCTL_INO_ROOT; } else { /* * Grab next entry. */ if (error = zap_cursor_retrieve(&zc, &zap)) { if ((*eofp = (error == ENOENT)) != 0) break; else goto update; } if (zap.za_integer_length != 8 || zap.za_num_integers != 1) { cmn_err(CE_WARN, "zap_readdir: bad directory " "entry, obj = %lld, offset = %lld\n", (u_longlong_t)zp->z_id, (u_longlong_t)offset); error = SET_ERROR(ENXIO); goto update; } objnum = ZFS_DIRENT_OBJ(zap.za_first_integer); /* * MacOS X can extract the object type here such as: * uint8_t type = ZFS_DIRENT_TYPE(zap.za_first_integer); */ if (check_sysattrs && !zap.za_normalization_conflict) { zap.za_normalization_conflict = xattr_sysattr_casechk(zap.za_name); } } if (flags & V_RDDIR_ACCFILTER) { /* * If we have no access at all, don't include * this entry in the returned information */ znode_t *ezp; if (zfs_zget(zp->z_zfsvfs, objnum, &ezp) != 0) goto skip_entry; if (!zfs_has_access(ezp, cr)) { VN_RELE(ZTOV(ezp)); goto skip_entry; } VN_RELE(ZTOV(ezp)); } if (flags & V_RDDIR_ENTFLAGS) reclen = EDIRENT_RECLEN(strlen(zap.za_name)); else reclen = DIRENT64_RECLEN(strlen(zap.za_name)); /* * Will this entry fit in the buffer? */ if (outcount + reclen > bufsize) { /* * Did we manage to fit anything in the buffer? */ if (!outcount) { error = SET_ERROR(EINVAL); goto update; } break; } if (flags & V_RDDIR_ENTFLAGS) { /* * Add extended flag entry: */ eodp->ed_ino = objnum; eodp->ed_reclen = reclen; /* NOTE: ed_off is the offset for the *next* entry */ next = &(eodp->ed_off); eodp->ed_eflags = zap.za_normalization_conflict ? ED_CASE_CONFLICT : 0; (void) strncpy(eodp->ed_name, zap.za_name, EDIRENT_NAMELEN(reclen)); eodp = (edirent_t *)((intptr_t)eodp + reclen); } else { /* * Add normal entry: */ odp->d_ino = objnum; odp->d_reclen = reclen; /* NOTE: d_off is the offset for the *next* entry */ next = &(odp->d_off); (void) strncpy(odp->d_name, zap.za_name, DIRENT64_NAMELEN(reclen)); odp = (dirent64_t *)((intptr_t)odp + reclen); } outcount += reclen; ASSERT(outcount <= bufsize); /* Prefetch znode */ if (prefetch) dmu_prefetch(os, objnum, 0, 0, 0, ZIO_PRIORITY_SYNC_READ); skip_entry: /* * Move to the next entry, fill in the previous offset. */ if (offset > 2 || (offset == 2 && !zfs_show_ctldir(zp))) { zap_cursor_advance(&zc); offset = zap_cursor_serialize(&zc); } else { offset += 1; } if (next) *next = offset; } zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */ if (uio->uio_segflg == UIO_SYSSPACE && uio->uio_iovcnt == 1) { iovp->iov_base += outcount; iovp->iov_len -= outcount; uio->uio_resid -= outcount; } else if (error = uiomove(outbuf, (long)outcount, UIO_READ, uio)) { /* * Reset the pointer. */ offset = uio->uio_loffset; } update: zap_cursor_fini(&zc); if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1) kmem_free(outbuf, bufsize); if (error == ENOENT) error = 0; ZFS_ACCESSTIME_STAMP(zfsvfs, zp); uio->uio_loffset = offset; ZFS_EXIT(zfsvfs); return (error); } ulong_t zfs_fsync_sync_cnt = 4; static int zfs_fsync(vnode_t *vp, int syncflag, cred_t *cr, caller_context_t *ct) { znode_t *zp = VTOZ(vp); zfsvfs_t *zfsvfs = zp->z_zfsvfs; /* * Regardless of whether this is required for standards conformance, * this is the logical behavior when fsync() is called on a file with * dirty pages. We use B_ASYNC since the ZIL transactions are already * going to be pushed out as part of the zil_commit(). */ if (vn_has_cached_data(vp) && !(syncflag & FNODSYNC) && (vp->v_type == VREG) && !(IS_SWAPVP(vp))) (void) VOP_PUTPAGE(vp, (offset_t)0, (size_t)0, B_ASYNC, cr, ct); (void) tsd_set(zfs_fsyncer_key, (void *)zfs_fsync_sync_cnt); if (zfsvfs->z_os->os_sync != ZFS_SYNC_DISABLED) { ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(zp); zil_commit(zfsvfs->z_log, zp->z_id); ZFS_EXIT(zfsvfs); } return (0); } /* * Get the requested file attributes and place them in the provided * vattr structure. * * IN: vp - vnode of file. * vap - va_mask identifies requested attributes. * If AT_XVATTR set, then optional attrs are requested * flags - ATTR_NOACLCHECK (CIFS server context) * cr - credentials of caller. * ct - caller context * * OUT: vap - attribute values. * * RETURN: 0 (always succeeds). */ /* ARGSUSED */ static int zfs_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr, caller_context_t *ct) { znode_t *zp = VTOZ(vp); zfsvfs_t *zfsvfs = zp->z_zfsvfs; int error = 0; uint64_t links; uint64_t mtime[2], ctime[2]; xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */ xoptattr_t *xoap = NULL; boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; sa_bulk_attr_t bulk[2]; int count = 0; ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(zp); zfs_fuid_map_ids(zp, cr, &vap->va_uid, &vap->va_gid); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); if ((error = sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) != 0) { ZFS_EXIT(zfsvfs); return (error); } /* * If ACL is trivial don't bother looking for ACE_READ_ATTRIBUTES. * Also, if we are the owner don't bother, since owner should * always be allowed to read basic attributes of file. */ if (!(zp->z_pflags & ZFS_ACL_TRIVIAL) && (vap->va_uid != crgetuid(cr))) { if (error = zfs_zaccess(zp, ACE_READ_ATTRIBUTES, 0, skipaclchk, cr)) { ZFS_EXIT(zfsvfs); return (error); } } /* * Return all attributes. It's cheaper to provide the answer * than to determine whether we were asked the question. */ mutex_enter(&zp->z_lock); vap->va_type = vp->v_type; vap->va_mode = zp->z_mode & MODEMASK; vap->va_fsid = zp->z_zfsvfs->z_vfs->vfs_dev; vap->va_nodeid = zp->z_id; if ((vp->v_flag & VROOT) && zfs_show_ctldir(zp)) links = zp->z_links + 1; else links = zp->z_links; vap->va_nlink = MIN(links, UINT32_MAX); /* nlink_t limit! */ vap->va_size = zp->z_size; vap->va_rdev = vp->v_rdev; vap->va_seq = zp->z_seq; /* * Add in any requested optional attributes and the create time. * Also set the corresponding bits in the returned attribute bitmap. */ if ((xoap = xva_getxoptattr(xvap)) != NULL && zfsvfs->z_use_fuids) { if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) { xoap->xoa_archive = ((zp->z_pflags & ZFS_ARCHIVE) != 0); XVA_SET_RTN(xvap, XAT_ARCHIVE); } if (XVA_ISSET_REQ(xvap, XAT_READONLY)) { xoap->xoa_readonly = ((zp->z_pflags & ZFS_READONLY) != 0); XVA_SET_RTN(xvap, XAT_READONLY); } if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) { xoap->xoa_system = ((zp->z_pflags & ZFS_SYSTEM) != 0); XVA_SET_RTN(xvap, XAT_SYSTEM); } if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) { xoap->xoa_hidden = ((zp->z_pflags & ZFS_HIDDEN) != 0); XVA_SET_RTN(xvap, XAT_HIDDEN); } if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { xoap->xoa_nounlink = ((zp->z_pflags & ZFS_NOUNLINK) != 0); XVA_SET_RTN(xvap, XAT_NOUNLINK); } if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { xoap->xoa_immutable = ((zp->z_pflags & ZFS_IMMUTABLE) != 0); XVA_SET_RTN(xvap, XAT_IMMUTABLE); } if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { xoap->xoa_appendonly = ((zp->z_pflags & ZFS_APPENDONLY) != 0); XVA_SET_RTN(xvap, XAT_APPENDONLY); } if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { xoap->xoa_nodump = ((zp->z_pflags & ZFS_NODUMP) != 0); XVA_SET_RTN(xvap, XAT_NODUMP); } if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) { xoap->xoa_opaque = ((zp->z_pflags & ZFS_OPAQUE) != 0); XVA_SET_RTN(xvap, XAT_OPAQUE); } if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { xoap->xoa_av_quarantined = ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0); XVA_SET_RTN(xvap, XAT_AV_QUARANTINED); } if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { xoap->xoa_av_modified = ((zp->z_pflags & ZFS_AV_MODIFIED) != 0); XVA_SET_RTN(xvap, XAT_AV_MODIFIED); } if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) && vp->v_type == VREG) { zfs_sa_get_scanstamp(zp, xvap); } if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) { uint64_t times[2]; (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(zfsvfs), times, sizeof (times)); ZFS_TIME_DECODE(&xoap->xoa_createtime, times); XVA_SET_RTN(xvap, XAT_CREATETIME); } if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { xoap->xoa_reparse = ((zp->z_pflags & ZFS_REPARSE) != 0); XVA_SET_RTN(xvap, XAT_REPARSE); } if (XVA_ISSET_REQ(xvap, XAT_GEN)) { xoap->xoa_generation = zp->z_gen; XVA_SET_RTN(xvap, XAT_GEN); } if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) { xoap->xoa_offline = ((zp->z_pflags & ZFS_OFFLINE) != 0); XVA_SET_RTN(xvap, XAT_OFFLINE); } if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) { xoap->xoa_sparse = ((zp->z_pflags & ZFS_SPARSE) != 0); XVA_SET_RTN(xvap, XAT_SPARSE); } } ZFS_TIME_DECODE(&vap->va_atime, zp->z_atime); ZFS_TIME_DECODE(&vap->va_mtime, mtime); ZFS_TIME_DECODE(&vap->va_ctime, ctime); mutex_exit(&zp->z_lock); sa_object_size(zp->z_sa_hdl, &vap->va_blksize, &vap->va_nblocks); if (zp->z_blksz == 0) { /* * Block size hasn't been set; suggest maximal I/O transfers. */ vap->va_blksize = zfsvfs->z_max_blksz; } ZFS_EXIT(zfsvfs); return (0); } /* * Set the file attributes to the values contained in the * vattr structure. * * IN: vp - vnode of file to be modified. * vap - new attribute values. * If AT_XVATTR set, then optional attrs are being set * flags - ATTR_UTIME set if non-default time values provided. * - ATTR_NOACLCHECK (CIFS context only). * cr - credentials of caller. * ct - caller context * * RETURN: 0 on success, error code on failure. * * Timestamps: * vp - ctime updated, mtime updated if size changed. */ /* ARGSUSED */ static int zfs_setattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr, caller_context_t *ct) { znode_t *zp = VTOZ(vp); zfsvfs_t *zfsvfs = zp->z_zfsvfs; zilog_t *zilog; dmu_tx_t *tx; vattr_t oldva; xvattr_t tmpxvattr; uint_t mask = vap->va_mask; uint_t saved_mask = 0; int trim_mask = 0; uint64_t new_mode; uint64_t new_uid, new_gid; uint64_t xattr_obj; uint64_t mtime[2], ctime[2]; znode_t *attrzp; int need_policy = FALSE; int err, err2; zfs_fuid_info_t *fuidp = NULL; xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */ xoptattr_t *xoap; zfs_acl_t *aclp; boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; boolean_t fuid_dirtied = B_FALSE; sa_bulk_attr_t bulk[7], xattr_bulk[7]; int count = 0, xattr_count = 0; if (mask == 0) return (0); if (mask & AT_NOSET) return (SET_ERROR(EINVAL)); ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(zp); zilog = zfsvfs->z_log; /* * Make sure that if we have ephemeral uid/gid or xvattr specified * that file system is at proper version level */ if (zfsvfs->z_use_fuids == B_FALSE && (((mask & AT_UID) && IS_EPHEMERAL(vap->va_uid)) || ((mask & AT_GID) && IS_EPHEMERAL(vap->va_gid)) || (mask & AT_XVATTR))) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EINVAL)); } if (mask & AT_SIZE && vp->v_type == VDIR) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EISDIR)); } if (mask & AT_SIZE && vp->v_type != VREG && vp->v_type != VFIFO) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EINVAL)); } /* * If this is an xvattr_t, then get a pointer to the structure of * optional attributes. If this is NULL, then we have a vattr_t. */ xoap = xva_getxoptattr(xvap); xva_init(&tmpxvattr); /* * Immutable files can only alter immutable bit and atime */ if ((zp->z_pflags & ZFS_IMMUTABLE) && ((mask & (AT_SIZE|AT_UID|AT_GID|AT_MTIME|AT_MODE)) || ((mask & AT_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EPERM)); } /* * Note: ZFS_READONLY is handled in zfs_zaccess_common. */ /* * Verify timestamps doesn't overflow 32 bits. * ZFS can handle large timestamps, but 32bit syscalls can't * handle times greater than 2039. This check should be removed * once large timestamps are fully supported. */ if (mask & (AT_ATIME | AT_MTIME)) { if (((mask & AT_ATIME) && TIMESPEC_OVERFLOW(&vap->va_atime)) || ((mask & AT_MTIME) && TIMESPEC_OVERFLOW(&vap->va_mtime))) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EOVERFLOW)); } } top: attrzp = NULL; aclp = NULL; /* Can this be moved to before the top label? */ if (zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EROFS)); } /* * First validate permissions */ if (mask & AT_SIZE) { err = zfs_zaccess(zp, ACE_WRITE_DATA, 0, skipaclchk, cr); if (err) { ZFS_EXIT(zfsvfs); return (err); } /* * XXX - Note, we are not providing any open * mode flags here (like FNDELAY), so we may * block if there are locks present... this * should be addressed in openat(). */ /* XXX - would it be OK to generate a log record here? */ err = zfs_freesp(zp, vap->va_size, 0, 0, FALSE); if (err) { ZFS_EXIT(zfsvfs); return (err); } if (vap->va_size == 0) vnevent_truncate(ZTOV(zp), ct); } if (mask & (AT_ATIME|AT_MTIME) || ((mask & AT_XVATTR) && (XVA_ISSET_REQ(xvap, XAT_HIDDEN) || XVA_ISSET_REQ(xvap, XAT_READONLY) || XVA_ISSET_REQ(xvap, XAT_ARCHIVE) || XVA_ISSET_REQ(xvap, XAT_OFFLINE) || XVA_ISSET_REQ(xvap, XAT_SPARSE) || XVA_ISSET_REQ(xvap, XAT_CREATETIME) || XVA_ISSET_REQ(xvap, XAT_SYSTEM)))) { need_policy = zfs_zaccess(zp, ACE_WRITE_ATTRIBUTES, 0, skipaclchk, cr); } if (mask & (AT_UID|AT_GID)) { int idmask = (mask & (AT_UID|AT_GID)); int take_owner; int take_group; /* * NOTE: even if a new mode is being set, * we may clear S_ISUID/S_ISGID bits. */ if (!(mask & AT_MODE)) vap->va_mode = zp->z_mode; /* * Take ownership or chgrp to group we are a member of */ take_owner = (mask & AT_UID) && (vap->va_uid == crgetuid(cr)); take_group = (mask & AT_GID) && zfs_groupmember(zfsvfs, vap->va_gid, cr); /* * If both AT_UID and AT_GID are set then take_owner and * take_group must both be set in order to allow taking * ownership. * * Otherwise, send the check through secpolicy_vnode_setattr() * */ if (((idmask == (AT_UID|AT_GID)) && take_owner && take_group) || ((idmask == AT_UID) && take_owner) || ((idmask == AT_GID) && take_group)) { if (zfs_zaccess(zp, ACE_WRITE_OWNER, 0, skipaclchk, cr) == 0) { /* * Remove setuid/setgid for non-privileged users */ secpolicy_setid_clear(vap, cr); trim_mask = (mask & (AT_UID|AT_GID)); } else { need_policy = TRUE; } } else { need_policy = TRUE; } } mutex_enter(&zp->z_lock); oldva.va_mode = zp->z_mode; zfs_fuid_map_ids(zp, cr, &oldva.va_uid, &oldva.va_gid); if (mask & AT_XVATTR) { /* * Update xvattr mask to include only those attributes * that are actually changing. * * the bits will be restored prior to actually setting * the attributes so the caller thinks they were set. */ if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { if (xoap->xoa_appendonly != ((zp->z_pflags & ZFS_APPENDONLY) != 0)) { need_policy = TRUE; } else { XVA_CLR_REQ(xvap, XAT_APPENDONLY); XVA_SET_REQ(&tmpxvattr, XAT_APPENDONLY); } } if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { if (xoap->xoa_nounlink != ((zp->z_pflags & ZFS_NOUNLINK) != 0)) { need_policy = TRUE; } else { XVA_CLR_REQ(xvap, XAT_NOUNLINK); XVA_SET_REQ(&tmpxvattr, XAT_NOUNLINK); } } if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { if (xoap->xoa_immutable != ((zp->z_pflags & ZFS_IMMUTABLE) != 0)) { need_policy = TRUE; } else { XVA_CLR_REQ(xvap, XAT_IMMUTABLE); XVA_SET_REQ(&tmpxvattr, XAT_IMMUTABLE); } } if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { if (xoap->xoa_nodump != ((zp->z_pflags & ZFS_NODUMP) != 0)) { need_policy = TRUE; } else { XVA_CLR_REQ(xvap, XAT_NODUMP); XVA_SET_REQ(&tmpxvattr, XAT_NODUMP); } } if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { if (xoap->xoa_av_modified != ((zp->z_pflags & ZFS_AV_MODIFIED) != 0)) { need_policy = TRUE; } else { XVA_CLR_REQ(xvap, XAT_AV_MODIFIED); XVA_SET_REQ(&tmpxvattr, XAT_AV_MODIFIED); } } if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { if ((vp->v_type != VREG && xoap->xoa_av_quarantined) || xoap->xoa_av_quarantined != ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0)) { need_policy = TRUE; } else { XVA_CLR_REQ(xvap, XAT_AV_QUARANTINED); XVA_SET_REQ(&tmpxvattr, XAT_AV_QUARANTINED); } } if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { mutex_exit(&zp->z_lock); ZFS_EXIT(zfsvfs); return (SET_ERROR(EPERM)); } if (need_policy == FALSE && (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) || XVA_ISSET_REQ(xvap, XAT_OPAQUE))) { need_policy = TRUE; } } mutex_exit(&zp->z_lock); if (mask & AT_MODE) { if (zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr) == 0) { err = secpolicy_setid_setsticky_clear(vp, vap, &oldva, cr); if (err) { ZFS_EXIT(zfsvfs); return (err); } trim_mask |= AT_MODE; } else { need_policy = TRUE; } } if (need_policy) { /* * If trim_mask is set then take ownership * has been granted or write_acl is present and user * has the ability to modify mode. In that case remove * UID|GID and or MODE from mask so that * secpolicy_vnode_setattr() doesn't revoke it. */ if (trim_mask) { saved_mask = vap->va_mask; vap->va_mask &= ~trim_mask; } err = secpolicy_vnode_setattr(cr, vp, vap, &oldva, flags, (int (*)(void *, int, cred_t *))zfs_zaccess_unix, zp); if (err) { ZFS_EXIT(zfsvfs); return (err); } if (trim_mask) vap->va_mask |= saved_mask; } /* * secpolicy_vnode_setattr, or take ownership may have * changed va_mask */ mask = vap->va_mask; if ((mask & (AT_UID | AT_GID))) { err = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &xattr_obj, sizeof (xattr_obj)); if (err == 0 && xattr_obj) { err = zfs_zget(zp->z_zfsvfs, xattr_obj, &attrzp); if (err) goto out2; } if (mask & AT_UID) { new_uid = zfs_fuid_create(zfsvfs, (uint64_t)vap->va_uid, cr, ZFS_OWNER, &fuidp); if (new_uid != zp->z_uid && zfs_fuid_overquota(zfsvfs, B_FALSE, new_uid)) { if (attrzp) VN_RELE(ZTOV(attrzp)); err = SET_ERROR(EDQUOT); goto out2; } } if (mask & AT_GID) { new_gid = zfs_fuid_create(zfsvfs, (uint64_t)vap->va_gid, cr, ZFS_GROUP, &fuidp); if (new_gid != zp->z_gid && zfs_fuid_overquota(zfsvfs, B_TRUE, new_gid)) { if (attrzp) VN_RELE(ZTOV(attrzp)); err = SET_ERROR(EDQUOT); goto out2; } } } tx = dmu_tx_create(zfsvfs->z_os); if (mask & AT_MODE) { uint64_t pmode = zp->z_mode; uint64_t acl_obj; new_mode = (pmode & S_IFMT) | (vap->va_mode & ~S_IFMT); if (zp->z_zfsvfs->z_acl_mode == ZFS_ACL_RESTRICTED && !(zp->z_pflags & ZFS_ACL_TRIVIAL)) { err = SET_ERROR(EPERM); goto out; } if (err = zfs_acl_chmod_setattr(zp, &aclp, new_mode)) goto out; mutex_enter(&zp->z_lock); if (!zp->z_is_sa && ((acl_obj = zfs_external_acl(zp)) != 0)) { /* * Are we upgrading ACL from old V0 format * to V1 format? */ if (zfsvfs->z_version >= ZPL_VERSION_FUID && zfs_znode_acl_version(zp) == ZFS_ACL_VERSION_INITIAL) { dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END); dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, aclp->z_acl_bytes); } else { dmu_tx_hold_write(tx, acl_obj, 0, aclp->z_acl_bytes); } } else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) { dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, aclp->z_acl_bytes); } mutex_exit(&zp->z_lock); dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); } else { if ((mask & AT_XVATTR) && XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); else dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); } if (attrzp) { dmu_tx_hold_sa(tx, attrzp->z_sa_hdl, B_FALSE); } fuid_dirtied = zfsvfs->z_fuid_dirty; if (fuid_dirtied) zfs_fuid_txhold(zfsvfs, tx); zfs_sa_upgrade_txholds(tx, zp); err = dmu_tx_assign(tx, TXG_WAIT); if (err) goto out; count = 0; /* * Set each attribute requested. * We group settings according to the locks they need to acquire. * * Note: you cannot set ctime directly, although it will be * updated as a side-effect of calling this function. */ if (mask & (AT_UID|AT_GID|AT_MODE)) mutex_enter(&zp->z_acl_lock); mutex_enter(&zp->z_lock); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, &zp->z_pflags, sizeof (zp->z_pflags)); if (attrzp) { if (mask & (AT_UID|AT_GID|AT_MODE)) mutex_enter(&attrzp->z_acl_lock); mutex_enter(&attrzp->z_lock); SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, SA_ZPL_FLAGS(zfsvfs), NULL, &attrzp->z_pflags, sizeof (attrzp->z_pflags)); } if (mask & (AT_UID|AT_GID)) { if (mask & AT_UID) { SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, &new_uid, sizeof (new_uid)); zp->z_uid = new_uid; if (attrzp) { SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, SA_ZPL_UID(zfsvfs), NULL, &new_uid, sizeof (new_uid)); attrzp->z_uid = new_uid; } } if (mask & AT_GID) { SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, &new_gid, sizeof (new_gid)); zp->z_gid = new_gid; if (attrzp) { SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, SA_ZPL_GID(zfsvfs), NULL, &new_gid, sizeof (new_gid)); attrzp->z_gid = new_gid; } } if (!(mask & AT_MODE)) { SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &new_mode, sizeof (new_mode)); new_mode = zp->z_mode; } err = zfs_acl_chown_setattr(zp); ASSERT(err == 0); if (attrzp) { err = zfs_acl_chown_setattr(attrzp); ASSERT(err == 0); } } if (mask & AT_MODE) { SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &new_mode, sizeof (new_mode)); zp->z_mode = new_mode; ASSERT3U((uintptr_t)aclp, !=, NULL); err = zfs_aclset_common(zp, aclp, cr, tx); ASSERT0(err); if (zp->z_acl_cached) zfs_acl_free(zp->z_acl_cached); zp->z_acl_cached = aclp; aclp = NULL; } if (mask & AT_ATIME) { ZFS_TIME_ENCODE(&vap->va_atime, zp->z_atime); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, &zp->z_atime, sizeof (zp->z_atime)); } if (mask & AT_MTIME) { ZFS_TIME_ENCODE(&vap->va_mtime, mtime); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, mtime, sizeof (mtime)); } /* XXX - shouldn't this be done *before* the ATIME/MTIME checks? */ if (mask & AT_SIZE && !(mask & AT_MTIME)) { SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, mtime, sizeof (mtime)); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, sizeof (ctime)); zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, B_TRUE); } else if (mask != 0) { SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, sizeof (ctime)); zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime, ctime, B_TRUE); if (attrzp) { SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, sizeof (ctime)); zfs_tstamp_update_setup(attrzp, STATE_CHANGED, mtime, ctime, B_TRUE); } } /* * Do this after setting timestamps to prevent timestamp * update from toggling bit */ if (xoap && (mask & AT_XVATTR)) { /* * restore trimmed off masks * so that return masks can be set for caller. */ if (XVA_ISSET_REQ(&tmpxvattr, XAT_APPENDONLY)) { XVA_SET_REQ(xvap, XAT_APPENDONLY); } if (XVA_ISSET_REQ(&tmpxvattr, XAT_NOUNLINK)) { XVA_SET_REQ(xvap, XAT_NOUNLINK); } if (XVA_ISSET_REQ(&tmpxvattr, XAT_IMMUTABLE)) { XVA_SET_REQ(xvap, XAT_IMMUTABLE); } if (XVA_ISSET_REQ(&tmpxvattr, XAT_NODUMP)) { XVA_SET_REQ(xvap, XAT_NODUMP); } if (XVA_ISSET_REQ(&tmpxvattr, XAT_AV_MODIFIED)) { XVA_SET_REQ(xvap, XAT_AV_MODIFIED); } if (XVA_ISSET_REQ(&tmpxvattr, XAT_AV_QUARANTINED)) { XVA_SET_REQ(xvap, XAT_AV_QUARANTINED); } if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) ASSERT(vp->v_type == VREG); zfs_xvattr_set(zp, xvap, tx); } if (fuid_dirtied) zfs_fuid_sync(zfsvfs, tx); if (mask != 0) zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask, fuidp); mutex_exit(&zp->z_lock); if (mask & (AT_UID|AT_GID|AT_MODE)) mutex_exit(&zp->z_acl_lock); if (attrzp) { if (mask & (AT_UID|AT_GID|AT_MODE)) mutex_exit(&attrzp->z_acl_lock); mutex_exit(&attrzp->z_lock); } out: if (err == 0 && attrzp) { err2 = sa_bulk_update(attrzp->z_sa_hdl, xattr_bulk, xattr_count, tx); ASSERT(err2 == 0); } if (attrzp) VN_RELE(ZTOV(attrzp)); if (aclp) zfs_acl_free(aclp); if (fuidp) { zfs_fuid_info_free(fuidp); fuidp = NULL; } if (err) { dmu_tx_abort(tx); if (err == ERESTART) goto top; } else { err2 = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); dmu_tx_commit(tx); } out2: if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) zil_commit(zilog, 0); ZFS_EXIT(zfsvfs); return (err); } typedef struct zfs_zlock { krwlock_t *zl_rwlock; /* lock we acquired */ znode_t *zl_znode; /* znode we held */ struct zfs_zlock *zl_next; /* next in list */ } zfs_zlock_t; /* * Drop locks and release vnodes that were held by zfs_rename_lock(). */ static void zfs_rename_unlock(zfs_zlock_t **zlpp) { zfs_zlock_t *zl; while ((zl = *zlpp) != NULL) { if (zl->zl_znode != NULL) VN_RELE(ZTOV(zl->zl_znode)); rw_exit(zl->zl_rwlock); *zlpp = zl->zl_next; kmem_free(zl, sizeof (*zl)); } } /* * Search back through the directory tree, using the ".." entries. * Lock each directory in the chain to prevent concurrent renames. * Fail any attempt to move a directory into one of its own descendants. * XXX - z_parent_lock can overlap with map or grow locks */ static int zfs_rename_lock(znode_t *szp, znode_t *tdzp, znode_t *sdzp, zfs_zlock_t **zlpp) { zfs_zlock_t *zl; znode_t *zp = tdzp; uint64_t rootid = zp->z_zfsvfs->z_root; uint64_t oidp = zp->z_id; krwlock_t *rwlp = &szp->z_parent_lock; krw_t rw = RW_WRITER; /* * First pass write-locks szp and compares to zp->z_id. * Later passes read-lock zp and compare to zp->z_parent. */ do { if (!rw_tryenter(rwlp, rw)) { /* * Another thread is renaming in this path. * Note that if we are a WRITER, we don't have any * parent_locks held yet. */ if (rw == RW_READER && zp->z_id > szp->z_id) { /* * Drop our locks and restart */ zfs_rename_unlock(&zl); *zlpp = NULL; zp = tdzp; oidp = zp->z_id; rwlp = &szp->z_parent_lock; rw = RW_WRITER; continue; } else { /* * Wait for other thread to drop its locks */ rw_enter(rwlp, rw); } } zl = kmem_alloc(sizeof (*zl), KM_SLEEP); zl->zl_rwlock = rwlp; zl->zl_znode = NULL; zl->zl_next = *zlpp; *zlpp = zl; if (oidp == szp->z_id) /* We're a descendant of szp */ return (SET_ERROR(EINVAL)); if (oidp == rootid) /* We've hit the top */ return (0); if (rw == RW_READER) { /* i.e. not the first pass */ int error = zfs_zget(zp->z_zfsvfs, oidp, &zp); if (error) return (error); zl->zl_znode = zp; } (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zp->z_zfsvfs), &oidp, sizeof (oidp)); rwlp = &zp->z_parent_lock; rw = RW_READER; } while (zp->z_id != sdzp->z_id); return (0); } /* * Move an entry from the provided source directory to the target * directory. Change the entry name as indicated. * * IN: sdvp - Source directory containing the "old entry". * snm - Old entry name. * tdvp - Target directory to contain the "new entry". * tnm - New entry name. * cr - credentials of caller. * ct - caller context * flags - case flags * * RETURN: 0 on success, error code on failure. * * Timestamps: * sdvp,tdvp - ctime|mtime updated */ /*ARGSUSED*/ static int zfs_rename(vnode_t *sdvp, char *snm, vnode_t *tdvp, char *tnm, cred_t *cr, caller_context_t *ct, int flags) { znode_t *tdzp, *szp, *tzp; znode_t *sdzp = VTOZ(sdvp); zfsvfs_t *zfsvfs = sdzp->z_zfsvfs; zilog_t *zilog; vnode_t *realvp; zfs_dirlock_t *sdl, *tdl; dmu_tx_t *tx; zfs_zlock_t *zl; int cmp, serr, terr; int error = 0, rm_err = 0; int zflg = 0; boolean_t waited = B_FALSE; ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(sdzp); zilog = zfsvfs->z_log; /* * Make sure we have the real vp for the target directory. */ if (VOP_REALVP(tdvp, &realvp, ct) == 0) tdvp = realvp; tdzp = VTOZ(tdvp); ZFS_VERIFY_ZP(tdzp); /* * We check z_zfsvfs rather than v_vfsp here, because snapshots and the * ctldir appear to have the same v_vfsp. */ if (tdzp->z_zfsvfs != zfsvfs || zfsctl_is_node(tdvp)) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EXDEV)); } if (zfsvfs->z_utf8 && u8_validate(tnm, strlen(tnm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EILSEQ)); } if (flags & FIGNORECASE) zflg |= ZCILOOK; top: szp = NULL; tzp = NULL; zl = NULL; /* * This is to prevent the creation of links into attribute space * by renaming a linked file into/outof an attribute directory. * See the comment in zfs_link() for why this is considered bad. */ if ((tdzp->z_pflags & ZFS_XATTR) != (sdzp->z_pflags & ZFS_XATTR)) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EINVAL)); } /* * Lock source and target directory entries. To prevent deadlock, * a lock ordering must be defined. We lock the directory with * the smallest object id first, or if it's a tie, the one with * the lexically first name. */ if (sdzp->z_id < tdzp->z_id) { cmp = -1; } else if (sdzp->z_id > tdzp->z_id) { cmp = 1; } else { /* * First compare the two name arguments without * considering any case folding. */ int nofold = (zfsvfs->z_norm & ~U8_TEXTPREP_TOUPPER); cmp = u8_strcmp(snm, tnm, 0, nofold, U8_UNICODE_LATEST, &error); ASSERT(error == 0 || !zfsvfs->z_utf8); if (cmp == 0) { /* * POSIX: "If the old argument and the new argument * both refer to links to the same existing file, * the rename() function shall return successfully * and perform no other action." */ ZFS_EXIT(zfsvfs); return (0); } /* * If the file system is case-folding, then we may * have some more checking to do. A case-folding file * system is either supporting mixed case sensitivity * access or is completely case-insensitive. Note * that the file system is always case preserving. * * In mixed sensitivity mode case sensitive behavior * is the default. FIGNORECASE must be used to * explicitly request case insensitive behavior. * * If the source and target names provided differ only * by case (e.g., a request to rename 'tim' to 'Tim'), * we will treat this as a special case in the * case-insensitive mode: as long as the source name * is an exact match, we will allow this to proceed as * a name-change request. */ if ((zfsvfs->z_case == ZFS_CASE_INSENSITIVE || (zfsvfs->z_case == ZFS_CASE_MIXED && flags & FIGNORECASE)) && u8_strcmp(snm, tnm, 0, zfsvfs->z_norm, U8_UNICODE_LATEST, &error) == 0) { /* * case preserving rename request, require exact * name matches */ zflg |= ZCIEXACT; zflg &= ~ZCILOOK; } } /* * If the source and destination directories are the same, we should * grab the z_name_lock of that directory only once. */ if (sdzp == tdzp) { zflg |= ZHAVELOCK; rw_enter(&sdzp->z_name_lock, RW_READER); } if (cmp < 0) { serr = zfs_dirent_lock(&sdl, sdzp, snm, &szp, ZEXISTS | zflg, NULL, NULL); terr = zfs_dirent_lock(&tdl, tdzp, tnm, &tzp, ZRENAMING | zflg, NULL, NULL); } else { terr = zfs_dirent_lock(&tdl, tdzp, tnm, &tzp, zflg, NULL, NULL); serr = zfs_dirent_lock(&sdl, sdzp, snm, &szp, ZEXISTS | ZRENAMING | zflg, NULL, NULL); } if (serr) { /* * Source entry invalid or not there. */ if (!terr) { zfs_dirent_unlock(tdl); if (tzp) VN_RELE(ZTOV(tzp)); } if (sdzp == tdzp) rw_exit(&sdzp->z_name_lock); if (strcmp(snm, "..") == 0) serr = SET_ERROR(EINVAL); ZFS_EXIT(zfsvfs); return (serr); } if (terr) { zfs_dirent_unlock(sdl); VN_RELE(ZTOV(szp)); if (sdzp == tdzp) rw_exit(&sdzp->z_name_lock); if (strcmp(tnm, "..") == 0) terr = SET_ERROR(EINVAL); ZFS_EXIT(zfsvfs); return (terr); } /* * Must have write access at the source to remove the old entry * and write access at the target to create the new entry. * Note that if target and source are the same, this can be * done in a single check. */ if (error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr)) goto out; if (ZTOV(szp)->v_type == VDIR) { /* * Check to make sure rename is valid. * Can't do a move like this: /usr/a/b to /usr/a/b/c/d */ if (error = zfs_rename_lock(szp, tdzp, sdzp, &zl)) goto out; } /* * Does target exist? */ if (tzp) { /* * Source and target must be the same type. */ if (ZTOV(szp)->v_type == VDIR) { if (ZTOV(tzp)->v_type != VDIR) { error = SET_ERROR(ENOTDIR); goto out; } } else { if (ZTOV(tzp)->v_type == VDIR) { error = SET_ERROR(EISDIR); goto out; } } /* * POSIX dictates that when the source and target * entries refer to the same file object, rename * must do nothing and exit without error. */ if (szp->z_id == tzp->z_id) { error = 0; goto out; } } vnevent_pre_rename_src(ZTOV(szp), sdvp, snm, ct); if (tzp) vnevent_pre_rename_dest(ZTOV(tzp), tdvp, tnm, ct); /* * notify the target directory if it is not the same * as source directory. */ if (tdvp != sdvp) { vnevent_pre_rename_dest_dir(tdvp, ZTOV(szp), tnm, ct); } tx = dmu_tx_create(zfsvfs->z_os); dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); dmu_tx_hold_sa(tx, sdzp->z_sa_hdl, B_FALSE); dmu_tx_hold_zap(tx, sdzp->z_id, FALSE, snm); dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, tnm); if (sdzp != tdzp) { dmu_tx_hold_sa(tx, tdzp->z_sa_hdl, B_FALSE); zfs_sa_upgrade_txholds(tx, tdzp); } if (tzp) { dmu_tx_hold_sa(tx, tzp->z_sa_hdl, B_FALSE); zfs_sa_upgrade_txholds(tx, tzp); } zfs_sa_upgrade_txholds(tx, szp); dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); - error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); + error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); if (error) { if (zl != NULL) zfs_rename_unlock(&zl); zfs_dirent_unlock(sdl); zfs_dirent_unlock(tdl); if (sdzp == tdzp) rw_exit(&sdzp->z_name_lock); VN_RELE(ZTOV(szp)); if (tzp) VN_RELE(ZTOV(tzp)); if (error == ERESTART) { waited = B_TRUE; dmu_tx_wait(tx); dmu_tx_abort(tx); goto top; } dmu_tx_abort(tx); ZFS_EXIT(zfsvfs); return (error); } if (tzp) /* Attempt to remove the existing target */ error = rm_err = zfs_link_destroy(tdl, tzp, tx, zflg, NULL); if (error == 0) { error = zfs_link_create(tdl, szp, tx, ZRENAMING); if (error == 0) { szp->z_pflags |= ZFS_AV_MODIFIED; error = sa_update(szp->z_sa_hdl, SA_ZPL_FLAGS(zfsvfs), (void *)&szp->z_pflags, sizeof (uint64_t), tx); ASSERT0(error); error = zfs_link_destroy(sdl, szp, tx, ZRENAMING, NULL); if (error == 0) { zfs_log_rename(zilog, tx, TX_RENAME | (flags & FIGNORECASE ? TX_CI : 0), sdzp, sdl->dl_name, tdzp, tdl->dl_name, szp); /* * Update path information for the target vnode */ vn_renamepath(tdvp, ZTOV(szp), tnm, strlen(tnm)); } else { /* * At this point, we have successfully created * the target name, but have failed to remove * the source name. Since the create was done * with the ZRENAMING flag, there are * complications; for one, the link count is * wrong. The easiest way to deal with this * is to remove the newly created target, and * return the original error. This must * succeed; fortunately, it is very unlikely to * fail, since we just created it. */ VERIFY3U(zfs_link_destroy(tdl, szp, tx, ZRENAMING, NULL), ==, 0); } } } dmu_tx_commit(tx); if (tzp && rm_err == 0) vnevent_rename_dest(ZTOV(tzp), tdvp, tnm, ct); if (error == 0) { vnevent_rename_src(ZTOV(szp), sdvp, snm, ct); /* notify the target dir if it is not the same as source dir */ if (tdvp != sdvp) vnevent_rename_dest_dir(tdvp, ct); } out: if (zl != NULL) zfs_rename_unlock(&zl); zfs_dirent_unlock(sdl); zfs_dirent_unlock(tdl); if (sdzp == tdzp) rw_exit(&sdzp->z_name_lock); VN_RELE(ZTOV(szp)); if (tzp) VN_RELE(ZTOV(tzp)); if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) zil_commit(zilog, 0); ZFS_EXIT(zfsvfs); return (error); } /* * Insert the indicated symbolic reference entry into the directory. * * IN: dvp - Directory to contain new symbolic link. * link - Name for new symlink entry. * vap - Attributes of new entry. * cr - credentials of caller. * ct - caller context * flags - case flags * * RETURN: 0 on success, error code on failure. * * Timestamps: * dvp - ctime|mtime updated */ /*ARGSUSED*/ static int zfs_symlink(vnode_t *dvp, char *name, vattr_t *vap, char *link, cred_t *cr, caller_context_t *ct, int flags) { znode_t *zp, *dzp = VTOZ(dvp); zfs_dirlock_t *dl; dmu_tx_t *tx; zfsvfs_t *zfsvfs = dzp->z_zfsvfs; zilog_t *zilog; uint64_t len = strlen(link); int error; int zflg = ZNEW; zfs_acl_ids_t acl_ids; boolean_t fuid_dirtied; uint64_t txtype = TX_SYMLINK; boolean_t waited = B_FALSE; ASSERT(vap->va_type == VLNK); ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(dzp); zilog = zfsvfs->z_log; if (zfsvfs->z_utf8 && u8_validate(name, strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EILSEQ)); } if (flags & FIGNORECASE) zflg |= ZCILOOK; if (len > MAXPATHLEN) { ZFS_EXIT(zfsvfs); return (SET_ERROR(ENAMETOOLONG)); } if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, NULL, &acl_ids)) != 0) { ZFS_EXIT(zfsvfs); return (error); } top: /* * Attempt to lock directory; fail if entry already exists. */ error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, NULL); if (error) { zfs_acl_ids_free(&acl_ids); ZFS_EXIT(zfsvfs); return (error); } if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) { zfs_acl_ids_free(&acl_ids); zfs_dirent_unlock(dl); ZFS_EXIT(zfsvfs); return (error); } if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) { zfs_acl_ids_free(&acl_ids); zfs_dirent_unlock(dl); ZFS_EXIT(zfsvfs); return (SET_ERROR(EDQUOT)); } tx = dmu_tx_create(zfsvfs->z_os); fuid_dirtied = zfsvfs->z_fuid_dirty; dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len)); dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + ZFS_SA_BASE_ATTR_SIZE + len); dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, acl_ids.z_aclp->z_acl_bytes); } if (fuid_dirtied) zfs_fuid_txhold(zfsvfs, tx); - error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); + error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); if (error) { zfs_dirent_unlock(dl); if (error == ERESTART) { waited = B_TRUE; dmu_tx_wait(tx); dmu_tx_abort(tx); goto top; } zfs_acl_ids_free(&acl_ids); dmu_tx_abort(tx); ZFS_EXIT(zfsvfs); return (error); } /* * Create a new object for the symlink. * for version 4 ZPL datsets the symlink will be an SA attribute */ zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); if (fuid_dirtied) zfs_fuid_sync(zfsvfs, tx); mutex_enter(&zp->z_lock); if (zp->z_is_sa) error = sa_update(zp->z_sa_hdl, SA_ZPL_SYMLINK(zfsvfs), link, len, tx); else zfs_sa_symlink(zp, link, len, tx); mutex_exit(&zp->z_lock); zp->z_size = len; (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs), &zp->z_size, sizeof (zp->z_size), tx); /* * Insert the new object into the directory. */ (void) zfs_link_create(dl, zp, tx, ZNEW); if (flags & FIGNORECASE) txtype |= TX_CI; zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link); zfs_acl_ids_free(&acl_ids); dmu_tx_commit(tx); zfs_dirent_unlock(dl); VN_RELE(ZTOV(zp)); if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) zil_commit(zilog, 0); ZFS_EXIT(zfsvfs); return (error); } /* * Return, in the buffer contained in the provided uio structure, * the symbolic path referred to by vp. * * IN: vp - vnode of symbolic link. * uio - structure to contain the link path. * cr - credentials of caller. * ct - caller context * * OUT: uio - structure containing the link path. * * RETURN: 0 on success, error code on failure. * * Timestamps: * vp - atime updated */ /* ARGSUSED */ static int zfs_readlink(vnode_t *vp, uio_t *uio, cred_t *cr, caller_context_t *ct) { znode_t *zp = VTOZ(vp); zfsvfs_t *zfsvfs = zp->z_zfsvfs; int error; ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(zp); mutex_enter(&zp->z_lock); if (zp->z_is_sa) error = sa_lookup_uio(zp->z_sa_hdl, SA_ZPL_SYMLINK(zfsvfs), uio); else error = zfs_sa_readlink(zp, uio); mutex_exit(&zp->z_lock); ZFS_ACCESSTIME_STAMP(zfsvfs, zp); ZFS_EXIT(zfsvfs); return (error); } /* * Insert a new entry into directory tdvp referencing svp. * * IN: tdvp - Directory to contain new entry. * svp - vnode of new entry. * name - name of new entry. * cr - credentials of caller. * ct - caller context * * RETURN: 0 on success, error code on failure. * * Timestamps: * tdvp - ctime|mtime updated * svp - ctime updated */ /* ARGSUSED */ static int zfs_link(vnode_t *tdvp, vnode_t *svp, char *name, cred_t *cr, caller_context_t *ct, int flags) { znode_t *dzp = VTOZ(tdvp); znode_t *tzp, *szp; zfsvfs_t *zfsvfs = dzp->z_zfsvfs; zilog_t *zilog; zfs_dirlock_t *dl; dmu_tx_t *tx; vnode_t *realvp; int error; int zf = ZNEW; uint64_t parent; uid_t owner; boolean_t waited = B_FALSE; ASSERT(tdvp->v_type == VDIR); ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(dzp); zilog = zfsvfs->z_log; if (VOP_REALVP(svp, &realvp, ct) == 0) svp = realvp; /* * POSIX dictates that we return EPERM here. * Better choices include ENOTSUP or EISDIR. */ if (svp->v_type == VDIR) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EPERM)); } szp = VTOZ(svp); ZFS_VERIFY_ZP(szp); /* * We check z_zfsvfs rather than v_vfsp here, because snapshots and the * ctldir appear to have the same v_vfsp. */ if (szp->z_zfsvfs != zfsvfs || zfsctl_is_node(svp)) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EXDEV)); } /* Prevent links to .zfs/shares files */ if ((error = sa_lookup(szp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs), &parent, sizeof (uint64_t))) != 0) { ZFS_EXIT(zfsvfs); return (error); } if (parent == zfsvfs->z_shares_dir) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EPERM)); } if (zfsvfs->z_utf8 && u8_validate(name, strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EILSEQ)); } if (flags & FIGNORECASE) zf |= ZCILOOK; /* * We do not support links between attributes and non-attributes * because of the potential security risk of creating links * into "normal" file space in order to circumvent restrictions * imposed in attribute space. */ if ((szp->z_pflags & ZFS_XATTR) != (dzp->z_pflags & ZFS_XATTR)) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EINVAL)); } owner = zfs_fuid_map_id(zfsvfs, szp->z_uid, cr, ZFS_OWNER); if (owner != crgetuid(cr) && secpolicy_basic_link(cr) != 0) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EPERM)); } if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) { ZFS_EXIT(zfsvfs); return (error); } top: /* * Attempt to lock directory; fail if entry already exists. */ error = zfs_dirent_lock(&dl, dzp, name, &tzp, zf, NULL, NULL); if (error) { ZFS_EXIT(zfsvfs); return (error); } tx = dmu_tx_create(zfsvfs->z_os); dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); zfs_sa_upgrade_txholds(tx, szp); zfs_sa_upgrade_txholds(tx, dzp); - error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); + error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); if (error) { zfs_dirent_unlock(dl); if (error == ERESTART) { waited = B_TRUE; dmu_tx_wait(tx); dmu_tx_abort(tx); goto top; } dmu_tx_abort(tx); ZFS_EXIT(zfsvfs); return (error); } error = zfs_link_create(dl, szp, tx, 0); if (error == 0) { uint64_t txtype = TX_LINK; if (flags & FIGNORECASE) txtype |= TX_CI; zfs_log_link(zilog, tx, txtype, dzp, szp, name); } dmu_tx_commit(tx); zfs_dirent_unlock(dl); if (error == 0) { vnevent_link(svp, ct); } if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) zil_commit(zilog, 0); ZFS_EXIT(zfsvfs); return (error); } /* * zfs_null_putapage() is used when the file system has been force * unmounted. It just drops the pages. */ /* ARGSUSED */ static int zfs_null_putapage(vnode_t *vp, page_t *pp, u_offset_t *offp, size_t *lenp, int flags, cred_t *cr) { pvn_write_done(pp, B_INVAL|B_FORCE|B_ERROR); return (0); } /* * Push a page out to disk, klustering if possible. * * IN: vp - file to push page to. * pp - page to push. * flags - additional flags. * cr - credentials of caller. * * OUT: offp - start of range pushed. * lenp - len of range pushed. * * RETURN: 0 on success, error code on failure. * * NOTE: callers must have locked the page to be pushed. On * exit, the page (and all other pages in the kluster) must be * unlocked. */ /* ARGSUSED */ static int zfs_putapage(vnode_t *vp, page_t *pp, u_offset_t *offp, size_t *lenp, int flags, cred_t *cr) { znode_t *zp = VTOZ(vp); zfsvfs_t *zfsvfs = zp->z_zfsvfs; dmu_tx_t *tx; u_offset_t off, koff; size_t len, klen; int err; off = pp->p_offset; len = PAGESIZE; /* * If our blocksize is bigger than the page size, try to kluster * multiple pages so that we write a full block (thus avoiding * a read-modify-write). */ if (off < zp->z_size && zp->z_blksz > PAGESIZE) { klen = P2ROUNDUP((ulong_t)zp->z_blksz, PAGESIZE); koff = ISP2(klen) ? P2ALIGN(off, (u_offset_t)klen) : 0; ASSERT(koff <= zp->z_size); if (koff + klen > zp->z_size) klen = P2ROUNDUP(zp->z_size - koff, (uint64_t)PAGESIZE); pp = pvn_write_kluster(vp, pp, &off, &len, koff, klen, flags); } ASSERT3U(btop(len), ==, btopr(len)); /* * Can't push pages past end-of-file. */ if (off >= zp->z_size) { /* ignore all pages */ err = 0; goto out; } else if (off + len > zp->z_size) { int npages = btopr(zp->z_size - off); page_t *trunc; page_list_break(&pp, &trunc, npages); /* ignore pages past end of file */ if (trunc) pvn_write_done(trunc, flags); len = zp->z_size - off; } if (zfs_owner_overquota(zfsvfs, zp, B_FALSE) || zfs_owner_overquota(zfsvfs, zp, B_TRUE)) { err = SET_ERROR(EDQUOT); goto out; } tx = dmu_tx_create(zfsvfs->z_os); dmu_tx_hold_write(tx, zp->z_id, off, len); dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); zfs_sa_upgrade_txholds(tx, zp); err = dmu_tx_assign(tx, TXG_WAIT); if (err != 0) { dmu_tx_abort(tx); goto out; } if (zp->z_blksz <= PAGESIZE) { caddr_t va = zfs_map_page(pp, S_READ); ASSERT3U(len, <=, PAGESIZE); dmu_write(zfsvfs->z_os, zp->z_id, off, len, va, tx); zfs_unmap_page(pp, va); } else { err = dmu_write_pages(zfsvfs->z_os, zp->z_id, off, len, pp, tx); } if (err == 0) { uint64_t mtime[2], ctime[2]; sa_bulk_attr_t bulk[3]; int count = 0; SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, &zp->z_pflags, 8); zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, B_TRUE); err = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); ASSERT0(err); zfs_log_write(zfsvfs->z_log, tx, TX_WRITE, zp, off, len, 0); } dmu_tx_commit(tx); out: pvn_write_done(pp, (err ? B_ERROR : 0) | flags); if (offp) *offp = off; if (lenp) *lenp = len; return (err); } /* * Copy the portion of the file indicated from pages into the file. * The pages are stored in a page list attached to the files vnode. * * IN: vp - vnode of file to push page data to. * off - position in file to put data. * len - amount of data to write. * flags - flags to control the operation. * cr - credentials of caller. * ct - caller context. * * RETURN: 0 on success, error code on failure. * * Timestamps: * vp - ctime|mtime updated */ /*ARGSUSED*/ static int zfs_putpage(vnode_t *vp, offset_t off, size_t len, int flags, cred_t *cr, caller_context_t *ct) { znode_t *zp = VTOZ(vp); zfsvfs_t *zfsvfs = zp->z_zfsvfs; page_t *pp; size_t io_len; u_offset_t io_off; uint_t blksz; rl_t *rl; int error = 0; ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(zp); /* * There's nothing to do if no data is cached. */ if (!vn_has_cached_data(vp)) { ZFS_EXIT(zfsvfs); return (0); } /* * Align this request to the file block size in case we kluster. * XXX - this can result in pretty aggresive locking, which can * impact simultanious read/write access. One option might be * to break up long requests (len == 0) into block-by-block * operations to get narrower locking. */ blksz = zp->z_blksz; if (ISP2(blksz)) io_off = P2ALIGN_TYPED(off, blksz, u_offset_t); else io_off = 0; if (len > 0 && ISP2(blksz)) io_len = P2ROUNDUP_TYPED(len + (off - io_off), blksz, size_t); else io_len = 0; if (io_len == 0) { /* * Search the entire vp list for pages >= io_off. */ rl = zfs_range_lock(zp, io_off, UINT64_MAX, RL_WRITER); error = pvn_vplist_dirty(vp, io_off, zfs_putapage, flags, cr); goto out; } rl = zfs_range_lock(zp, io_off, io_len, RL_WRITER); if (off > zp->z_size) { /* past end of file */ zfs_range_unlock(rl); ZFS_EXIT(zfsvfs); return (0); } len = MIN(io_len, P2ROUNDUP(zp->z_size, PAGESIZE) - io_off); for (off = io_off; io_off < off + len; io_off += io_len) { if ((flags & B_INVAL) || ((flags & B_ASYNC) == 0)) { pp = page_lookup(vp, io_off, (flags & (B_INVAL | B_FREE)) ? SE_EXCL : SE_SHARED); } else { pp = page_lookup_nowait(vp, io_off, (flags & B_FREE) ? SE_EXCL : SE_SHARED); } if (pp != NULL && pvn_getdirty(pp, flags)) { int err; /* * Found a dirty page to push */ err = zfs_putapage(vp, pp, &io_off, &io_len, flags, cr); if (err) error = err; } else { io_len = PAGESIZE; } } out: zfs_range_unlock(rl); if ((flags & B_ASYNC) == 0 || zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) zil_commit(zfsvfs->z_log, zp->z_id); ZFS_EXIT(zfsvfs); return (error); } /*ARGSUSED*/ void zfs_inactive(vnode_t *vp, cred_t *cr, caller_context_t *ct) { znode_t *zp = VTOZ(vp); zfsvfs_t *zfsvfs = zp->z_zfsvfs; int error; rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER); if (zp->z_sa_hdl == NULL) { /* * The fs has been unmounted, or we did a * suspend/resume and this file no longer exists. */ if (vn_has_cached_data(vp)) { (void) pvn_vplist_dirty(vp, 0, zfs_null_putapage, B_INVAL, cr); } mutex_enter(&zp->z_lock); mutex_enter(&vp->v_lock); ASSERT(vp->v_count == 1); VN_RELE_LOCKED(vp); mutex_exit(&vp->v_lock); mutex_exit(&zp->z_lock); rw_exit(&zfsvfs->z_teardown_inactive_lock); zfs_znode_free(zp); return; } /* * Attempt to push any data in the page cache. If this fails * we will get kicked out later in zfs_zinactive(). */ if (vn_has_cached_data(vp)) { (void) pvn_vplist_dirty(vp, 0, zfs_putapage, B_INVAL|B_ASYNC, cr); } if (zp->z_atime_dirty && zp->z_unlinked == 0) { dmu_tx_t *tx = dmu_tx_create(zfsvfs->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_WAIT); if (error) { dmu_tx_abort(tx); } else { mutex_enter(&zp->z_lock); (void) sa_update(zp->z_sa_hdl, SA_ZPL_ATIME(zfsvfs), (void *)&zp->z_atime, sizeof (zp->z_atime), tx); zp->z_atime_dirty = 0; mutex_exit(&zp->z_lock); dmu_tx_commit(tx); } } zfs_zinactive(zp); rw_exit(&zfsvfs->z_teardown_inactive_lock); } /* * Bounds-check the seek operation. * * IN: vp - vnode seeking within * ooff - old file offset * noffp - pointer to new file offset * ct - caller context * * RETURN: 0 on success, EINVAL if new offset invalid. */ /* ARGSUSED */ static int zfs_seek(vnode_t *vp, offset_t ooff, offset_t *noffp, caller_context_t *ct) { if (vp->v_type == VDIR) return (0); return ((*noffp < 0 || *noffp > MAXOFFSET_T) ? EINVAL : 0); } /* * Pre-filter the generic locking function to trap attempts to place * a mandatory lock on a memory mapped file. */ static int zfs_frlock(vnode_t *vp, int cmd, flock64_t *bfp, int flag, offset_t offset, flk_callback_t *flk_cbp, cred_t *cr, caller_context_t *ct) { znode_t *zp = VTOZ(vp); zfsvfs_t *zfsvfs = zp->z_zfsvfs; ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(zp); /* * We are following the UFS semantics with respect to mapcnt * here: If we see that the file is mapped already, then we will * return an error, but we don't worry about races between this * function and zfs_map(). */ if (zp->z_mapcnt > 0 && MANDMODE(zp->z_mode)) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EAGAIN)); } ZFS_EXIT(zfsvfs); return (fs_frlock(vp, cmd, bfp, flag, offset, flk_cbp, cr, ct)); } /* * If we can't find a page in the cache, we will create a new page * and fill it with file data. For efficiency, we may try to fill * multiple pages at once (klustering) to fill up the supplied page * list. Note that the pages to be filled are held with an exclusive * lock to prevent access by other threads while they are being filled. */ static int zfs_fillpage(vnode_t *vp, u_offset_t off, struct seg *seg, caddr_t addr, page_t *pl[], size_t plsz, enum seg_rw rw) { znode_t *zp = VTOZ(vp); page_t *pp, *cur_pp; objset_t *os = zp->z_zfsvfs->z_os; u_offset_t io_off, total; size_t io_len; int err; if (plsz == PAGESIZE || zp->z_blksz <= PAGESIZE) { /* * We only have a single page, don't bother klustering */ io_off = off; io_len = PAGESIZE; pp = page_create_va(vp, io_off, io_len, PG_EXCL | PG_WAIT, seg, addr); } else { /* * Try to find enough pages to fill the page list */ pp = pvn_read_kluster(vp, off, seg, addr, &io_off, &io_len, off, plsz, 0); } if (pp == NULL) { /* * The page already exists, nothing to do here. */ *pl = NULL; return (0); } /* * Fill the pages in the kluster. */ cur_pp = pp; for (total = io_off + io_len; io_off < total; io_off += PAGESIZE) { caddr_t va; ASSERT3U(io_off, ==, cur_pp->p_offset); va = zfs_map_page(cur_pp, S_WRITE); err = dmu_read(os, zp->z_id, io_off, PAGESIZE, va, DMU_READ_PREFETCH); zfs_unmap_page(cur_pp, va); if (err) { /* On error, toss the entire kluster */ pvn_read_done(pp, B_ERROR); /* convert checksum errors into IO errors */ if (err == ECKSUM) err = SET_ERROR(EIO); return (err); } cur_pp = cur_pp->p_next; } /* * Fill in the page list array from the kluster starting * from the desired offset `off'. * NOTE: the page list will always be null terminated. */ pvn_plist_init(pp, pl, plsz, off, io_len, rw); ASSERT(pl == NULL || (*pl)->p_offset == off); return (0); } /* * Return pointers to the pages for the file region [off, off + len] * in the pl array. If plsz is greater than len, this function may * also return page pointers from after the specified region * (i.e. the region [off, off + plsz]). These additional pages are * only returned if they are already in the cache, or were created as * part of a klustered read. * * IN: vp - vnode of file to get data from. * off - position in file to get data from. * len - amount of data to retrieve. * plsz - length of provided page list. * seg - segment to obtain pages for. * addr - virtual address of fault. * rw - mode of created pages. * cr - credentials of caller. * ct - caller context. * * OUT: protp - protection mode of created pages. * pl - list of pages created. * * RETURN: 0 on success, error code on failure. * * Timestamps: * vp - atime updated */ /* ARGSUSED */ static int zfs_getpage(vnode_t *vp, offset_t off, size_t len, uint_t *protp, page_t *pl[], size_t plsz, struct seg *seg, caddr_t addr, enum seg_rw rw, cred_t *cr, caller_context_t *ct) { znode_t *zp = VTOZ(vp); zfsvfs_t *zfsvfs = zp->z_zfsvfs; page_t **pl0 = pl; int err = 0; /* we do our own caching, faultahead is unnecessary */ if (pl == NULL) return (0); else if (len > plsz) len = plsz; else len = P2ROUNDUP(len, PAGESIZE); ASSERT(plsz >= len); ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(zp); if (protp) *protp = PROT_ALL; /* * Loop through the requested range [off, off + len) looking * for pages. If we don't find a page, we will need to create * a new page and fill it with data from the file. */ while (len > 0) { if (*pl = page_lookup(vp, off, SE_SHARED)) *(pl+1) = NULL; else if (err = zfs_fillpage(vp, off, seg, addr, pl, plsz, rw)) goto out; while (*pl) { ASSERT3U((*pl)->p_offset, ==, off); off += PAGESIZE; addr += PAGESIZE; if (len > 0) { ASSERT3U(len, >=, PAGESIZE); len -= PAGESIZE; } ASSERT3U(plsz, >=, PAGESIZE); plsz -= PAGESIZE; pl++; } } /* * Fill out the page array with any pages already in the cache. */ while (plsz > 0 && (*pl++ = page_lookup_nowait(vp, off, SE_SHARED))) { off += PAGESIZE; plsz -= PAGESIZE; } out: if (err) { /* * Release any pages we have previously locked. */ while (pl > pl0) page_unlock(*--pl); } else { ZFS_ACCESSTIME_STAMP(zfsvfs, zp); } *pl = NULL; ZFS_EXIT(zfsvfs); return (err); } /* * Request a memory map for a section of a file. This code interacts * with common code and the VM system as follows: * * - common code calls mmap(), which ends up in smmap_common() * - this calls VOP_MAP(), which takes you into (say) zfs * - zfs_map() calls as_map(), passing segvn_create() as the callback * - segvn_create() creates the new segment and calls VOP_ADDMAP() * - zfs_addmap() updates z_mapcnt */ /*ARGSUSED*/ static int zfs_map(vnode_t *vp, offset_t off, struct as *as, caddr_t *addrp, size_t len, uchar_t prot, uchar_t maxprot, uint_t flags, cred_t *cr, caller_context_t *ct) { znode_t *zp = VTOZ(vp); zfsvfs_t *zfsvfs = zp->z_zfsvfs; segvn_crargs_t vn_a; int error; ZFS_ENTER(zfsvfs); ZFS_VERIFY_ZP(zp); /* * Note: ZFS_READONLY is handled in zfs_zaccess_common. */ if ((prot & PROT_WRITE) && (zp->z_pflags & (ZFS_IMMUTABLE | ZFS_APPENDONLY))) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EPERM)); } if ((prot & (PROT_READ | PROT_EXEC)) && (zp->z_pflags & ZFS_AV_QUARANTINED)) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EACCES)); } if (vp->v_flag & VNOMAP) { ZFS_EXIT(zfsvfs); return (SET_ERROR(ENOSYS)); } if (off < 0 || len > MAXOFFSET_T - off) { ZFS_EXIT(zfsvfs); return (SET_ERROR(ENXIO)); } if (vp->v_type != VREG) { ZFS_EXIT(zfsvfs); return (SET_ERROR(ENODEV)); } /* * If file is locked, disallow mapping. */ if (MANDMODE(zp->z_mode) && vn_has_flocks(vp)) { ZFS_EXIT(zfsvfs); return (SET_ERROR(EAGAIN)); } as_rangelock(as); error = choose_addr(as, addrp, len, off, ADDR_VACALIGN, flags); if (error != 0) { as_rangeunlock(as); ZFS_EXIT(zfsvfs); return (error); } vn_a.vp = vp; vn_a.offset = (u_offset_t)off; vn_a.type = flags & MAP_TYPE; vn_a.prot = prot; vn_a.maxprot = maxprot; vn_a.cred = cr; vn_a.amp = NULL; vn_a.flags = flags & ~MAP_TYPE; vn_a.szc = 0; vn_a.lgrp_mem_policy_flags = 0; error = as_map(as, *addrp, len, segvn_create, &vn_a); as_rangeunlock(as); ZFS_EXIT(zfsvfs); return (error); } /* ARGSUSED */ static int zfs_addmap(vnode_t *vp, offset_t off, struct as *as, caddr_t addr, size_t len, uchar_t prot, uchar_t maxprot, uint_t flags, cred_t *cr, caller_context_t *ct) { uint64_t pages = btopr(len); atomic_add_64(&VTOZ(vp)->z_mapcnt, pages); return (0); } /* * The reason we push dirty pages as part of zfs_delmap() is so that we get a * more accurate mtime for the associated file. Since we don't have a way of * detecting when the data was actually modified, we have to resort to * heuristics. If an explicit msync() is done, then we mark the mtime when the * last page is pushed. The problem occurs when the msync() call is omitted, * which by far the most common case: * * open() * mmap() * * munmap() * close() *