*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2011 by Delphix. All rights reserved.
+ * Copyright (c) 2013 by Delphix. All rights reserved.
* Copyright (c) 2011 Nexenta Systems, Inc. All rights reserved.
*/
#include <sys/arc.h>
#include <sys/ddt.h>
-/*
- * ==========================================================================
- * I/O priority table
- * ==========================================================================
- */
-uint8_t zio_priority_table[ZIO_PRIORITY_TABLE_SIZE] = {
- 0, /* ZIO_PRIORITY_NOW */
- 0, /* ZIO_PRIORITY_SYNC_READ */
- 0, /* ZIO_PRIORITY_SYNC_WRITE */
- 0, /* ZIO_PRIORITY_LOG_WRITE */
- 1, /* ZIO_PRIORITY_CACHE_FILL */
- 1, /* ZIO_PRIORITY_AGG */
- 4, /* ZIO_PRIORITY_FREE */
- 4, /* ZIO_PRIORITY_ASYNC_WRITE */
- 6, /* ZIO_PRIORITY_ASYNC_READ */
- 10, /* ZIO_PRIORITY_RESILVER */
- 20, /* ZIO_PRIORITY_SCRUB */
- 2, /* ZIO_PRIORITY_DDT_PREFETCH */
-};
-
/*
* ==========================================================================
* I/O type descriptions
* ==========================================================================
*/
-char *zio_type_name[ZIO_TYPES] = {
+const char *zio_type_name[ZIO_TYPES] = {
"z_null", "z_rd", "z_wr", "z_fr", "z_cl", "z_ioctl"
};
*/
kmem_cache_t *zio_cache;
kmem_cache_t *zio_link_cache;
+kmem_cache_t *zio_vdev_cache;
kmem_cache_t *zio_buf_cache[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT];
kmem_cache_t *zio_data_buf_cache[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT];
int zio_bulk_flags = 0;
int zio_delay_max = ZIO_DELAY_MAX;
-#ifdef _KERNEL
-extern vmem_t *zio_alloc_arena;
-#endif
extern int zfs_mg_alloc_failures;
+/*
+ * The following actions directly effect the spa's sync-to-convergence logic.
+ * The values below define the sync pass when we start performing the action.
+ * Care should be taken when changing these values as they directly impact
+ * spa_sync() performance. Tuning these values may introduce subtle performance
+ * pathologies and should only be done in the context of performance analysis.
+ * These tunables will eventually be removed and replaced with #defines once
+ * enough analysis has been done to determine optimal values.
+ *
+ * The 'zfs_sync_pass_deferred_free' pass must be greater than 1 to ensure that
+ * regular blocks are not deferred.
+ */
+int zfs_sync_pass_deferred_free = 2; /* defer frees starting in this pass */
+int zfs_sync_pass_dont_compress = 5; /* don't compress starting in this pass */
+int zfs_sync_pass_rewrite = 2; /* rewrite new bps starting in this pass */
+
/*
* An allocating zio is one that either currently has the DVA allocate
* stage set or will have it later in its lifetime.
size_t c;
vmem_t *data_alloc_arena = NULL;
-#ifdef _KERNEL
- data_alloc_arena = zio_alloc_arena;
-#endif
zio_cache = kmem_cache_create("zio_cache", sizeof (zio_t), 0,
zio_cons, zio_dest, NULL, NULL, NULL, KMC_KMEM);
zio_link_cache = kmem_cache_create("zio_link_cache",
sizeof (zio_link_t), 0, NULL, NULL, NULL, NULL, NULL, KMC_KMEM);
+ zio_vdev_cache = kmem_cache_create("zio_vdev_cache", sizeof (vdev_io_t),
+ PAGESIZE, NULL, NULL, NULL, NULL, NULL, KMC_VMEM);
/*
* For small buffers, we want a cache for each multiple of
while (p2 & (p2 - 1))
p2 &= p2 - 1;
+#ifndef _KERNEL
+ /*
+ * If we are using watchpoints, put each buffer on its own page,
+ * to eliminate the performance overhead of trapping to the
+ * kernel when modifying a non-watched buffer that shares the
+ * page with a watched buffer.
+ */
+ if (arc_watch && !IS_P2ALIGNED(size, PAGESIZE))
+ continue;
+#endif
if (size <= 4 * SPA_MINBLOCKSIZE) {
align = SPA_MINBLOCKSIZE;
- } else if (P2PHASE(size, PAGESIZE) == 0) {
+ } else if (IS_P2ALIGNED(size, PAGESIZE)) {
align = PAGESIZE;
- } else if (P2PHASE(size, p2 >> 2) == 0) {
+ } else if (IS_P2ALIGNED(size, p2 >> 2)) {
align = p2 >> 2;
}
* The zio write taskqs have 1 thread per cpu, allow 1/2 of the taskqs
* to fail 3 times per txg or 8 failures, whichever is greater.
*/
- zfs_mg_alloc_failures = MAX((3 * max_ncpus / 2), 8);
+ if (zfs_mg_alloc_failures == 0)
+ zfs_mg_alloc_failures = MAX((3 * max_ncpus / 2), 8);
zio_inject_init();
+
+ lz4_init();
}
void
zio_data_buf_cache[c] = NULL;
}
+ kmem_cache_destroy(zio_vdev_cache);
kmem_cache_destroy(zio_link_cache);
kmem_cache_destroy(zio_cache);
zio_inject_fini();
+
+ lz4_fini();
}
/*
ASSERT(c < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT);
- return (kmem_cache_alloc(zio_buf_cache[c], KM_PUSHPAGE));
+ return (kmem_cache_alloc(zio_buf_cache[c], KM_PUSHPAGE | KM_NODEBUG));
}
/*
ASSERT(c < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT);
- return (kmem_cache_alloc(zio_data_buf_cache[c], KM_PUSHPAGE));
+ return (kmem_cache_alloc(zio_data_buf_cache[c],
+ KM_PUSHPAGE | KM_NODEBUG));
}
void
kmem_cache_free(zio_data_buf_cache[c], buf);
}
+/*
+ * Dedicated I/O buffers to ensure that memory fragmentation never prevents
+ * or significantly delays the issuing of a zio. These buffers are used
+ * to aggregate I/O and could be used for raidz stripes.
+ */
+void *
+zio_vdev_alloc(void)
+{
+ return (kmem_cache_alloc(zio_vdev_cache, KM_PUSHPAGE));
+}
+
+void
+zio_vdev_free(void *buf)
+{
+ kmem_cache_free(zio_vdev_cache, buf);
+
+}
+
/*
* ==========================================================================
* Push and pop I/O transform buffers
if (zio->io_error == 0 &&
zio_decompress_data(BP_GET_COMPRESS(zio->io_bp),
zio->io_data, data, zio->io_size, size) != 0)
- zio->io_error = EIO;
+ zio->io_error = SET_ERROR(EIO);
}
/*
*errorp = zio_worst_error(*errorp, zio->io_error);
pio->io_reexecute |= zio->io_reexecute;
ASSERT3U(*countp, >, 0);
- if (--*countp == 0 && pio->io_stall == countp) {
+
+ (*countp)--;
+
+ if (*countp == 0 && pio->io_stall == countp) {
pio->io_stall = NULL;
mutex_exit(&pio->io_lock);
__zio_execute(pio);
static zio_t *
zio_create(zio_t *pio, spa_t *spa, uint64_t txg, const blkptr_t *bp,
void *data, uint64_t size, zio_done_func_t *done, void *private,
- zio_type_t type, int priority, enum zio_flag flags,
+ zio_type_t type, zio_priority_t priority, enum zio_flag flags,
vdev_t *vd, uint64_t offset, const zbookmark_t *zb,
enum zio_stage stage, enum zio_stage pipeline)
{
zio->io_child_type = ZIO_CHILD_LOGICAL;
if (bp != NULL) {
+ zio->io_logical = NULL;
zio->io_bp = (blkptr_t *)bp;
zio->io_bp_copy = *bp;
zio->io_bp_orig = *bp;
zio->io_spa = spa;
zio->io_txg = txg;
zio->io_ready = NULL;
+ zio->io_physdone = NULL;
zio->io_done = done;
zio->io_private = private;
zio->io_prev_space_delta = 0;
zio->io_vsd = NULL;
zio->io_vsd_ops = NULL;
zio->io_offset = offset;
- zio->io_deadline = 0;
+ zio->io_timestamp = 0;
+ zio->io_delta = 0;
+ zio->io_delay = 0;
zio->io_orig_data = zio->io_data = data;
zio->io_orig_size = zio->io_size = size;
zio->io_orig_flags = zio->io_flags = flags;
zio->io_bp_override = NULL;
zio->io_walk_link = NULL;
zio->io_transform_stack = NULL;
- zio->io_delay = 0;
zio->io_error = 0;
zio->io_child_count = 0;
+ zio->io_phys_children = 0;
zio->io_parent_count = 0;
zio->io_stall = NULL;
zio->io_gang_leader = NULL;
zio_t *
zio_read(zio_t *pio, spa_t *spa, const blkptr_t *bp,
void *data, uint64_t size, zio_done_func_t *done, void *private,
- int priority, enum zio_flag flags, const zbookmark_t *zb)
+ zio_priority_t priority, enum zio_flag flags, const zbookmark_t *zb)
{
zio_t *zio;
zio_t *
zio_write(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
void *data, uint64_t size, const zio_prop_t *zp,
- zio_done_func_t *ready, zio_done_func_t *done, void *private,
- int priority, enum zio_flag flags, const zbookmark_t *zb)
+ zio_done_func_t *ready, zio_done_func_t *physdone, zio_done_func_t *done,
+ void *private,
+ zio_priority_t priority, enum zio_flag flags, const zbookmark_t *zb)
{
zio_t *zio;
zp->zp_checksum < ZIO_CHECKSUM_FUNCTIONS &&
zp->zp_compress >= ZIO_COMPRESS_OFF &&
zp->zp_compress < ZIO_COMPRESS_FUNCTIONS &&
- zp->zp_type < DMU_OT_NUMTYPES &&
+ DMU_OT_IS_VALID(zp->zp_type) &&
zp->zp_level < 32 &&
zp->zp_copies > 0 &&
- zp->zp_copies <= spa_max_replication(spa) &&
- zp->zp_dedup <= 1 &&
- zp->zp_dedup_verify <= 1);
+ zp->zp_copies <= spa_max_replication(spa));
zio = zio_create(pio, spa, txg, bp, data, size, done, private,
ZIO_TYPE_WRITE, priority, flags, NULL, 0, zb,
ZIO_DDT_CHILD_WRITE_PIPELINE : ZIO_WRITE_PIPELINE);
zio->io_ready = ready;
+ zio->io_physdone = physdone;
zio->io_prop = *zp;
return (zio);
zio_t *
zio_rewrite(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp, void *data,
- uint64_t size, zio_done_func_t *done, void *private, int priority,
- enum zio_flag flags, zbookmark_t *zb)
+ uint64_t size, zio_done_func_t *done, void *private,
+ zio_priority_t priority, enum zio_flag flags, zbookmark_t *zb)
{
zio_t *zio;
}
void
-zio_write_override(zio_t *zio, blkptr_t *bp, int copies)
+zio_write_override(zio_t *zio, blkptr_t *bp, int copies, boolean_t nopwrite)
{
ASSERT(zio->io_type == ZIO_TYPE_WRITE);
ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);
ASSERT(zio->io_stage == ZIO_STAGE_OPEN);
ASSERT(zio->io_txg == spa_syncing_txg(zio->io_spa));
+ /*
+ * We must reset the io_prop to match the values that existed
+ * when the bp was first written by dmu_sync() keeping in mind
+ * that nopwrite and dedup are mutually exclusive.
+ */
+ zio->io_prop.zp_dedup = nopwrite ? B_FALSE : zio->io_prop.zp_dedup;
+ zio->io_prop.zp_nopwrite = nopwrite;
zio->io_prop.zp_copies = copies;
zio->io_bp_override = bp;
}
void
zio_free(spa_t *spa, uint64_t txg, const blkptr_t *bp)
{
- bplist_append(&spa->spa_free_bplist[txg & TXG_MASK], bp);
+ metaslab_check_free(spa, bp);
+
+ /*
+ * Frees that are for the currently-syncing txg, are not going to be
+ * deferred, and which will not need to do a read (i.e. not GANG or
+ * DEDUP), can be processed immediately. Otherwise, put them on the
+ * in-memory list for later processing.
+ */
+ if (BP_IS_GANG(bp) || BP_GET_DEDUP(bp) ||
+ txg != spa->spa_syncing_txg ||
+ spa_sync_pass(spa) >= zfs_sync_pass_deferred_free) {
+ bplist_append(&spa->spa_free_bplist[txg & TXG_MASK], bp);
+ } else {
+ VERIFY0(zio_wait(zio_free_sync(NULL, spa, txg, bp, 0)));
+ }
}
zio_t *
enum zio_flag flags)
{
zio_t *zio;
+ enum zio_stage stage = ZIO_FREE_PIPELINE;
dprintf_bp(bp, "freeing in txg %llu, pass %u",
(longlong_t)txg, spa->spa_sync_pass);
ASSERT(!BP_IS_HOLE(bp));
ASSERT(spa_syncing_txg(spa) == txg);
- ASSERT(spa_sync_pass(spa) <= SYNC_PASS_DEFERRED_FREE);
+ ASSERT(spa_sync_pass(spa) < zfs_sync_pass_deferred_free);
+
+ metaslab_check_free(spa, bp);
+ arc_freed(spa, bp);
+
+ /*
+ * GANG and DEDUP blocks can induce a read (for the gang block header,
+ * or the DDT), so issue them asynchronously so that this thread is
+ * not tied up.
+ */
+ if (BP_IS_GANG(bp) || BP_GET_DEDUP(bp))
+ stage |= ZIO_STAGE_ISSUE_ASYNC;
zio = zio_create(pio, spa, txg, bp, NULL, BP_GET_PSIZE(bp),
- NULL, NULL, ZIO_TYPE_FREE, ZIO_PRIORITY_FREE, flags,
- NULL, 0, NULL, ZIO_STAGE_OPEN, ZIO_FREE_PIPELINE);
+ NULL, NULL, ZIO_TYPE_FREE, ZIO_PRIORITY_NOW, flags,
+ NULL, 0, NULL, ZIO_STAGE_OPEN, stage);
return (zio);
}
zio_t *
zio_ioctl(zio_t *pio, spa_t *spa, vdev_t *vd, int cmd,
- zio_done_func_t *done, void *private, int priority, enum zio_flag flags)
+ zio_done_func_t *done, void *private, enum zio_flag flags)
{
zio_t *zio;
int c;
if (vd->vdev_children == 0) {
zio = zio_create(pio, spa, 0, NULL, NULL, 0, done, private,
- ZIO_TYPE_IOCTL, priority, flags, vd, 0, NULL,
+ ZIO_TYPE_IOCTL, ZIO_PRIORITY_NOW, flags, vd, 0, NULL,
ZIO_STAGE_OPEN, ZIO_IOCTL_PIPELINE);
zio->io_cmd = cmd;
for (c = 0; c < vd->vdev_children; c++)
zio_nowait(zio_ioctl(zio, spa, vd->vdev_child[c], cmd,
- done, private, priority, flags));
+ done, private, flags));
}
return (zio);
zio_t *
zio_read_phys(zio_t *pio, vdev_t *vd, uint64_t offset, uint64_t size,
void *data, int checksum, zio_done_func_t *done, void *private,
- int priority, enum zio_flag flags, boolean_t labels)
+ zio_priority_t priority, enum zio_flag flags, boolean_t labels)
{
zio_t *zio;
zio_t *
zio_write_phys(zio_t *pio, vdev_t *vd, uint64_t offset, uint64_t size,
void *data, int checksum, zio_done_func_t *done, void *private,
- int priority, enum zio_flag flags, boolean_t labels)
+ zio_priority_t priority, enum zio_flag flags, boolean_t labels)
{
zio_t *zio;
*/
zio_t *
zio_vdev_child_io(zio_t *pio, blkptr_t *bp, vdev_t *vd, uint64_t offset,
- void *data, uint64_t size, int type, int priority, enum zio_flag flags,
- zio_done_func_t *done, void *private)
+ void *data, uint64_t size, int type, zio_priority_t priority,
+ enum zio_flag flags, zio_done_func_t *done, void *private)
{
enum zio_stage pipeline = ZIO_VDEV_CHILD_PIPELINE;
zio_t *zio;
done, private, type, priority, flags, vd, offset, &pio->io_bookmark,
ZIO_STAGE_VDEV_IO_START >> 1, pipeline);
+ zio->io_physdone = pio->io_physdone;
+ if (vd->vdev_ops->vdev_op_leaf && zio->io_logical != NULL)
+ zio->io_logical->io_phys_children++;
+
return (zio);
}
zio_t *
zio_vdev_delegated_io(vdev_t *vd, uint64_t offset, void *data, uint64_t size,
- int type, int priority, enum zio_flag flags,
+ int type, zio_priority_t priority, enum zio_flag flags,
zio_done_func_t *done, void *private)
{
zio_t *zio;
zio = zio_create(NULL, vd->vdev_spa, 0, NULL,
data, size, done, private, type, priority,
- flags | ZIO_FLAG_CANFAIL | ZIO_FLAG_DONT_RETRY,
+ flags | ZIO_FLAG_CANFAIL | ZIO_FLAG_DONT_RETRY | ZIO_FLAG_DELEGATED,
vd, offset, NULL,
ZIO_STAGE_VDEV_IO_START >> 1, ZIO_VDEV_CHILD_PIPELINE);
zio_flush(zio_t *zio, vdev_t *vd)
{
zio_nowait(zio_ioctl(zio, zio->io_spa, vd, DKIOCFLUSHWRITECACHE,
- NULL, NULL, ZIO_PRIORITY_NOW,
+ NULL, NULL,
ZIO_FLAG_CANFAIL | ZIO_FLAG_DONT_PROPAGATE | ZIO_FLAG_DONT_RETRY));
}
zio_push_transform(zio, cbuf, psize, psize, zio_decompress);
}
- if (!dmu_ot[BP_GET_TYPE(bp)].ot_metadata && BP_GET_LEVEL(bp) == 0)
+ if (!DMU_OT_IS_METADATA(BP_GET_TYPE(bp)) && BP_GET_LEVEL(bp) == 0)
zio->io_flags |= ZIO_FLAG_DONT_CACHE;
if (BP_GET_TYPE(bp) == DMU_OT_DDT_ZAP)
*bp = *zio->io_bp_override;
zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;
+ /*
+ * If we've been overridden and nopwrite is set then
+ * set the flag accordingly to indicate that a nopwrite
+ * has already occurred.
+ */
+ if (!BP_IS_HOLE(bp) && zp->zp_nopwrite) {
+ ASSERT(!zp->zp_dedup);
+ zio->io_flags |= ZIO_FLAG_NOPWRITE;
+ return (ZIO_PIPELINE_CONTINUE);
+ }
+
+ ASSERT(!zp->zp_nopwrite);
+
if (BP_IS_HOLE(bp) || !zp->zp_dedup)
return (ZIO_PIPELINE_CONTINUE);
ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);
ASSERT(!BP_GET_DEDUP(bp));
- if (pass > SYNC_PASS_DONT_COMPRESS)
+ if (pass >= zfs_sync_pass_dont_compress)
compress = ZIO_COMPRESS_OFF;
/* Make sure someone doesn't change their mind on overwrites */
* There should only be a handful of blocks after pass 1 in any case.
*/
if (bp->blk_birth == zio->io_txg && BP_GET_PSIZE(bp) == psize &&
- pass > SYNC_PASS_REWRITE) {
+ pass >= zfs_sync_pass_rewrite) {
enum zio_stage gang_stages = zio->io_pipeline & ZIO_GANG_STAGES;
ASSERT(psize != 0);
zio->io_pipeline = ZIO_REWRITE_PIPELINE | gang_stages;
ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REWRITE));
zio->io_pipeline = ZIO_DDT_WRITE_PIPELINE;
}
+ if (zp->zp_nopwrite) {
+ ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);
+ ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REWRITE));
+ zio->io_pipeline |= ZIO_STAGE_NOP_WRITE;
+ }
}
return (ZIO_PIPELINE_CONTINUE);
*/
static void
-zio_taskq_dispatch(zio_t *zio, enum zio_taskq_type q, boolean_t cutinline)
+zio_taskq_dispatch(zio_t *zio, zio_taskq_type_t q, boolean_t cutinline)
{
spa_t *spa = zio->io_spa;
zio_type_t t = zio->io_type;
t = ZIO_TYPE_NULL;
/*
- * If this is a high priority I/O, then use the high priority taskq.
+ * If this is a high priority I/O, then use the high priority taskq if
+ * available.
*/
if (zio->io_priority == ZIO_PRIORITY_NOW &&
- spa->spa_zio_taskq[t][q + 1] != NULL)
+ spa->spa_zio_taskq[t][q + 1].stqs_count != 0)
q++;
ASSERT3U(q, <, ZIO_TASKQ_TYPES);
* to dispatch the zio to another taskq at the same time.
*/
ASSERT(taskq_empty_ent(&zio->io_tqent));
- taskq_dispatch_ent(spa->spa_zio_taskq[t][q],
- (task_func_t *)zio_execute, zio, flags, &zio->io_tqent);
+ spa_taskq_dispatch_ent(spa, t, q, (task_func_t *)zio_execute, zio,
+ flags, &zio->io_tqent);
}
static boolean_t
-zio_taskq_member(zio_t *zio, enum zio_taskq_type q)
+zio_taskq_member(zio_t *zio, zio_taskq_type_t q)
{
kthread_t *executor = zio->io_executor;
spa_t *spa = zio->io_spa;
zio_type_t t;
- for (t = 0; t < ZIO_TYPES; t++)
- if (taskq_member(spa->spa_zio_taskq[t][q], executor))
- return (B_TRUE);
+ for (t = 0; t < ZIO_TYPES; t++) {
+ spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q];
+ uint_t i;
+ for (i = 0; i < tqs->stqs_count; i++) {
+ if (taskq_member(tqs->stqs_taskq[i], executor))
+ return (B_TRUE);
+ }
+ }
return (B_FALSE);
}
* vdev-level caching or aggregation; (5) the I/O is deferred
* due to vdev-level queueing; (6) the I/O is handed off to
* another thread. In all cases, the pipeline stops whenever
- * there's no CPU work; it never burns a thread in cv_wait().
+ * there's no CPU work; it never burns a thread in cv_wait_io().
*
* There's no locking on io_stage because there's no legitimate way
* for multiple threads to be attempting to process the same I/O.
while (zio->io_stage < ZIO_STAGE_DONE) {
enum zio_stage pipeline = zio->io_pipeline;
enum zio_stage stage = zio->io_stage;
- dsl_pool_t *dsl;
+ dsl_pool_t *dp;
boolean_t cut;
int rv;
ASSERT(stage <= ZIO_STAGE_DONE);
- dsl = spa_get_dsl(zio->io_spa);
+ dp = spa_get_dsl(zio->io_spa);
cut = (stage == ZIO_STAGE_VDEV_IO_START) ?
zio_requeue_io_start_cut_in_line : B_FALSE;
* or may wait for an I/O that needs an interrupt thread
* to complete, issue async to avoid deadlock.
*
- * If we are in the txg_sync_thread or being called
- * during pool init issue async to minimize stack depth.
- * Both of these call paths may be recursively called.
- *
* For VDEV_IO_START, we cut in line so that the io will
* be sent to disk promptly.
*/
- if (((stage & ZIO_BLOCKING_STAGES) && zio->io_vd == NULL &&
- zio_taskq_member(zio, ZIO_TASKQ_INTERRUPT)) ||
- (dsl != NULL && dsl_pool_sync_context(dsl))) {
+ if ((stage & ZIO_BLOCKING_STAGES) && zio->io_vd == NULL &&
+ zio_taskq_member(zio, ZIO_TASKQ_INTERRUPT)) {
zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, cut);
return;
}
+#ifdef _KERNEL
+ /*
+ * If we executing in the context of the tx_sync_thread,
+ * or we are performing pool initialization outside of a
+ * zio_taskq[ZIO_TASKQ_ISSUE|ZIO_TASKQ_ISSUE_HIGH] context.
+ * Then issue the zio asynchronously to minimize stack usage
+ * for these deep call paths.
+ */
+ if ((dp && curthread == dp->dp_tx.tx_sync_thread) ||
+ (dp && spa_is_initializing(dp->dp_spa) &&
+ !zio_taskq_member(zio, ZIO_TASKQ_ISSUE) &&
+ !zio_taskq_member(zio, ZIO_TASKQ_ISSUE_HIGH))) {
+ zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, cut);
+ return;
+ }
+#endif
+
zio->io_stage = stage;
rv = zio_pipeline[highbit(stage) - 1](zio);
mutex_enter(&zio->io_lock);
while (zio->io_executor != NULL)
- cv_wait(&zio->io_cv, &zio->io_lock);
+ cv_wait_io(&zio->io_cv, &zio->io_lock);
mutex_exit(&zio->io_lock);
error = zio->io_error;
pio->io_stage = pio->io_orig_stage;
pio->io_pipeline = pio->io_orig_pipeline;
pio->io_reexecute = 0;
+ pio->io_flags |= ZIO_FLAG_REEXECUTED;
pio->io_error = 0;
for (w = 0; w < ZIO_WAIT_TYPES; w++)
pio->io_state[w] = 0;
"failure and the failure mode property for this pool "
"is set to panic.", spa_name(spa));
+ cmn_err(CE_WARN, "Pool '%s' has encountered an uncorrectable I/O "
+ "failure and has been suspended.\n", spa_name(spa));
+
zfs_ereport_post(FM_EREPORT_ZFS_IO_FAILURE, spa, NULL, NULL, 0, 0);
mutex_enter(&spa->spa_suspend_lock);
zio_write_gang_member_ready(zio_t *zio)
{
zio_t *pio = zio_unique_parent(zio);
- ASSERTV(zio_t *gio = zio->io_gang_leader;)
dva_t *cdva = zio->io_bp->blk_dva;
dva_t *pdva = pio->io_bp->blk_dva;
uint64_t asize;
int d;
+ ASSERTV(zio_t *gio = zio->io_gang_leader);
if (BP_IS_HOLE(zio->io_bp))
return;
zp.zp_type = DMU_OT_NONE;
zp.zp_level = 0;
zp.zp_copies = gio->io_prop.zp_copies;
- zp.zp_dedup = 0;
- zp.zp_dedup_verify = 0;
+ zp.zp_dedup = B_FALSE;
+ zp.zp_dedup_verify = B_FALSE;
+ zp.zp_nopwrite = B_FALSE;
zio_nowait(zio_write(zio, spa, txg, &gbh->zg_blkptr[g],
(char *)pio->io_data + (pio->io_size - resid), lsize, &zp,
- zio_write_gang_member_ready, NULL, &gn->gn_child[g],
+ zio_write_gang_member_ready, NULL, NULL, &gn->gn_child[g],
pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio),
&pio->io_bookmark));
}
*/
pio->io_pipeline = ZIO_INTERLOCK_PIPELINE;
+ /*
+ * We didn't allocate this bp, so make sure it doesn't get unmarked.
+ */
+ pio->io_flags &= ~ZIO_FLAG_FASTWRITE;
+
zio_nowait(zio);
return (ZIO_PIPELINE_CONTINUE);
}
+/*
+ * The zio_nop_write stage in the pipeline determines if allocating
+ * a new bp is necessary. By leveraging a cryptographically secure checksum,
+ * such as SHA256, we can compare the checksums of the new data and the old
+ * to determine if allocating a new block is required. The nopwrite
+ * feature can handle writes in either syncing or open context (i.e. zil
+ * writes) and as a result is mutually exclusive with dedup.
+ */
+static int
+zio_nop_write(zio_t *zio)
+{
+ blkptr_t *bp = zio->io_bp;
+ blkptr_t *bp_orig = &zio->io_bp_orig;
+ zio_prop_t *zp = &zio->io_prop;
+
+ ASSERT(BP_GET_LEVEL(bp) == 0);
+ ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REWRITE));
+ ASSERT(zp->zp_nopwrite);
+ ASSERT(!zp->zp_dedup);
+ ASSERT(zio->io_bp_override == NULL);
+ ASSERT(IO_IS_ALLOCATING(zio));
+
+ /*
+ * Check to see if the original bp and the new bp have matching
+ * characteristics (i.e. same checksum, compression algorithms, etc).
+ * If they don't then just continue with the pipeline which will
+ * allocate a new bp.
+ */
+ if (BP_IS_HOLE(bp_orig) ||
+ !zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_dedup ||
+ BP_GET_CHECKSUM(bp) != BP_GET_CHECKSUM(bp_orig) ||
+ BP_GET_COMPRESS(bp) != BP_GET_COMPRESS(bp_orig) ||
+ BP_GET_DEDUP(bp) != BP_GET_DEDUP(bp_orig) ||
+ zp->zp_copies != BP_GET_NDVAS(bp_orig))
+ return (ZIO_PIPELINE_CONTINUE);
+
+ /*
+ * If the checksums match then reset the pipeline so that we
+ * avoid allocating a new bp and issuing any I/O.
+ */
+ if (ZIO_CHECKSUM_EQUAL(bp->blk_cksum, bp_orig->blk_cksum)) {
+ ASSERT(zio_checksum_table[zp->zp_checksum].ci_dedup);
+ ASSERT3U(BP_GET_PSIZE(bp), ==, BP_GET_PSIZE(bp_orig));
+ ASSERT3U(BP_GET_LSIZE(bp), ==, BP_GET_LSIZE(bp_orig));
+ ASSERT(zp->zp_compress != ZIO_COMPRESS_OFF);
+ ASSERT(bcmp(&bp->blk_prop, &bp_orig->blk_prop,
+ sizeof (uint64_t)) == 0);
+
+ *bp = *bp_orig;
+ zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;
+ zio->io_flags |= ZIO_FLAG_NOPWRITE;
+ }
+
+ return (ZIO_PIPELINE_CONTINUE);
+}
+
/*
* ==========================================================================
* Dedup
ddt_exit(ddt);
- error = arc_read_nolock(NULL, spa, &blk,
+ error = arc_read(NULL, spa, &blk,
arc_getbuf_func, &abuf, ZIO_PRIORITY_SYNC_READ,
ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE,
&aflags, &zio->io_bookmark);
if (arc_buf_size(abuf) != zio->io_orig_size ||
bcmp(abuf->b_data, zio->io_orig_data,
zio->io_orig_size) != 0)
- error = EEXIST;
- VERIFY(arc_buf_remove_ref(abuf, &abuf) == 1);
+ error = SET_ERROR(EEXIST);
+ VERIFY(arc_buf_remove_ref(abuf, &abuf));
}
ddt_enter(ddt);
zio->io_stage = ZIO_STAGE_OPEN;
BP_ZERO(bp);
} else {
- zp->zp_dedup = 0;
+ zp->zp_dedup = B_FALSE;
}
zio->io_pipeline = ZIO_WRITE_PIPELINE;
ddt_exit(ddt);
}
dio = zio_write(zio, spa, txg, bp, zio->io_orig_data,
- zio->io_orig_size, &czp, NULL,
+ zio->io_orig_size, &czp, NULL, NULL,
zio_ddt_ditto_write_done, dde, zio->io_priority,
ZIO_DDT_CHILD_FLAGS(zio), &zio->io_bookmark);
ddt_phys_addref(ddp);
} else {
cio = zio_write(zio, spa, txg, bp, zio->io_orig_data,
- zio->io_orig_size, zp, zio_ddt_child_write_ready,
+ zio->io_orig_size, zp, zio_ddt_child_write_ready, NULL,
zio_ddt_child_write_done, dde, zio->io_priority,
ZIO_DDT_CHILD_FLAGS(zio), &zio->io_bookmark);
ddt_enter(ddt);
freedde = dde = ddt_lookup(ddt, bp, B_TRUE);
- ddp = ddt_phys_select(dde, bp);
- ddt_phys_decref(ddp);
+ if (dde) {
+ ddp = ddt_phys_select(dde, bp);
+ if (ddp)
+ ddt_phys_decref(ddp);
+ }
ddt_exit(ddt);
return (ZIO_PIPELINE_CONTINUE);
}
ASSERT(BP_IS_HOLE(bp));
- ASSERT3U(BP_GET_NDVAS(bp), ==, 0);
+ ASSERT0(BP_GET_NDVAS(bp));
ASSERT3U(zio->io_prop.zp_copies, >, 0);
ASSERT3U(zio->io_prop.zp_copies, <=, spa_max_replication(spa));
ASSERT3U(zio->io_size, ==, BP_GET_PSIZE(bp));
flags |= (zio->io_flags & ZIO_FLAG_NODATA) ? METASLAB_GANG_AVOID : 0;
flags |= (zio->io_flags & ZIO_FLAG_GANG_CHILD) ?
METASLAB_GANG_CHILD : 0;
+ flags |= (zio->io_flags & ZIO_FLAG_FASTWRITE) ? METASLAB_FASTWRITE : 0;
error = metaslab_alloc(spa, mc, zio->io_size, bp,
zio->io_prop.zp_copies, zio->io_txg, NULL, flags);
* Try to allocate an intent log block. Return 0 on success, errno on failure.
*/
int
-zio_alloc_zil(spa_t *spa, uint64_t txg, blkptr_t *new_bp, blkptr_t *old_bp,
- uint64_t size, boolean_t use_slog)
+zio_alloc_zil(spa_t *spa, uint64_t txg, blkptr_t *new_bp, uint64_t size,
+ boolean_t use_slog)
{
int error = 1;
ASSERT(txg > spa_syncing_txg(spa));
- if (use_slog)
+ /*
+ * ZIL blocks are always contiguous (i.e. not gang blocks) so we
+ * set the METASLAB_GANG_AVOID flag so that they don't "fast gang"
+ * when allocating them.
+ */
+ if (use_slog) {
error = metaslab_alloc(spa, spa_log_class(spa), size,
- new_bp, 1, txg, old_bp, METASLAB_HINTBP_AVOID);
+ new_bp, 1, txg, NULL,
+ METASLAB_FASTWRITE | METASLAB_GANG_AVOID);
+ }
- if (error)
+ if (error) {
error = metaslab_alloc(spa, spa_normal_class(spa), size,
- new_bp, 1, txg, old_bp, METASLAB_HINTBP_AVOID);
+ new_bp, 1, txg, NULL,
+ METASLAB_FASTWRITE);
+ }
if (error == 0) {
BP_SET_LSIZE(new_bp, size);
return (ZIO_PIPELINE_STOP);
if (!vdev_accessible(vd, zio)) {
- zio->io_error = ENXIO;
+ zio->io_error = SET_ERROR(ENXIO);
zio_interrupt(zio);
return (ZIO_PIPELINE_STOP);
}
if (zio->io_error) {
if (!vdev_accessible(vd, zio)) {
- zio->io_error = ENXIO;
+ zio->io_error = SET_ERROR(ENXIO);
} else {
unexpected_error = B_TRUE;
}
*/
if (zio->io_error && vd != NULL && vd->vdev_ops->vdev_op_leaf &&
!vdev_accessible(vd, zio))
- zio->io_error = ENXIO;
+ zio->io_error = SET_ERROR(ENXIO);
/*
* If we can't write to an interior vdev (mirror or RAID-Z),
* set vdev_cant_write so that we stop trying to allocate from it.
*/
if (zio->io_error == ENXIO && zio->io_type == ZIO_TYPE_WRITE &&
- vd != NULL && !vd->vdev_ops->vdev_op_leaf)
+ vd != NULL && !vd->vdev_ops->vdev_op_leaf) {
vd->vdev_cant_write = B_TRUE;
+ }
if (zio->io_error)
zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;
+ if (vd != NULL && vd->vdev_ops->vdev_op_leaf &&
+ zio->io_physdone != NULL) {
+ ASSERT(!(zio->io_flags & ZIO_FLAG_DELEGATED));
+ ASSERT(zio->io_child_type == ZIO_CHILD_VDEV);
+ zio->io_physdone(zio->io_logical);
+ }
+
return (ZIO_PIPELINE_CONTINUE);
}
if (zio->io_ready) {
ASSERT(IO_IS_ALLOCATING(zio));
- ASSERT(bp->blk_birth == zio->io_txg || BP_IS_HOLE(bp));
+ ASSERT(bp->blk_birth == zio->io_txg || BP_IS_HOLE(bp) ||
+ (zio->io_flags & ZIO_FLAG_NOPWRITE));
ASSERT(zio->io_children[ZIO_CHILD_GANG][ZIO_WAIT_READY] == 0);
zio->io_ready(zio);
if (zio->io_bp != NULL) {
ASSERT(zio->io_bp->blk_pad[0] == 0);
ASSERT(zio->io_bp->blk_pad[1] == 0);
- ASSERT(bcmp(zio->io_bp, &zio->io_bp_copy, sizeof (blkptr_t)) == 0 ||
+ ASSERT(bcmp(zio->io_bp, &zio->io_bp_copy,
+ sizeof (blkptr_t)) == 0 ||
(zio->io_bp == zio_unique_parent(zio)->io_bp));
if (zio->io_type == ZIO_TYPE_WRITE && !BP_IS_HOLE(zio->io_bp) &&
zio->io_bp_override == NULL &&
!(zio->io_flags & ZIO_FLAG_IO_REPAIR)) {
ASSERT(!BP_SHOULD_BYTESWAP(zio->io_bp));
- ASSERT3U(zio->io_prop.zp_copies, <=, BP_GET_NDVAS(zio->io_bp));
+ ASSERT3U(zio->io_prop.zp_copies, <=,
+ BP_GET_NDVAS(zio->io_bp));
ASSERT(BP_COUNT_GANG(zio->io_bp) == 0 ||
- (BP_COUNT_GANG(zio->io_bp) == BP_GET_NDVAS(zio->io_bp)));
+ (BP_COUNT_GANG(zio->io_bp) ==
+ BP_GET_NDVAS(zio->io_bp)));
}
+ if (zio->io_flags & ZIO_FLAG_NOPWRITE)
+ VERIFY(BP_EQUAL(zio->io_bp, &zio->io_bp_orig));
}
/*
if (asize != zio->io_size) {
abuf = zio_buf_alloc(asize);
bcopy(zio->io_data, abuf, zio->io_size);
- bzero(abuf + zio->io_size, asize - zio->io_size);
+ bzero(abuf+zio->io_size, asize-zio->io_size);
}
zio->io_cksum_report = zcr->zcr_next;
vdev_stat_update(zio, zio->io_size);
/*
- * If this I/O is attached to a particular vdev is slow, exeeding
+ * If this I/O is attached to a particular vdev is slow, exceeding
* 30 seconds to complete, post an error described the I/O delay.
* We ignore these errors if the device is currently unavailable.
*/
- if (zio->io_delay >= zio_delay_max) {
+ if (zio->io_delay >= MSEC_TO_TICK(zio_delay_max)) {
if (zio->io_vd != NULL && !vdev_is_dead(zio->io_vd))
zfs_ereport_post(FM_EREPORT_ZFS_DELAY, zio->io_spa,
- zio->io_vd, zio, 0, 0);
+ zio->io_vd, zio, 0, 0);
}
if (zio->io_error) {
* error and generate a logical data ereport.
*/
spa_log_error(zio->io_spa, zio);
- zfs_ereport_post(FM_EREPORT_ZFS_DATA, zio->io_spa, NULL, zio,
- 0, 0);
+ zfs_ereport_post(FM_EREPORT_ZFS_DATA, zio->io_spa,
+ NULL, zio, 0, 0);
}
}
if ((zio->io_error || zio->io_reexecute) &&
IO_IS_ALLOCATING(zio) && zio->io_gang_leader == zio &&
- !(zio->io_flags & ZIO_FLAG_IO_REWRITE))
+ !(zio->io_flags & (ZIO_FLAG_IO_REWRITE | ZIO_FLAG_NOPWRITE)))
zio_dva_unallocate(zio, zio->io_gang_tree, zio->io_bp);
zio_gang_tree_free(&zio->io_gang_tree);
* Hand it off to the otherwise-unused claim taskq.
*/
ASSERT(taskq_empty_ent(&zio->io_tqent));
- (void) taskq_dispatch_ent(
- zio->io_spa->spa_zio_taskq[ZIO_TYPE_CLAIM][ZIO_TASKQ_ISSUE],
+ spa_taskq_dispatch_ent(zio->io_spa,
+ ZIO_TYPE_CLAIM, ZIO_TASKQ_ISSUE,
(task_func_t *)zio_reexecute, zio, 0,
&zio->io_tqent);
}
zfs_ereport_free_checksum(zcr);
}
+ if (zio->io_flags & ZIO_FLAG_FASTWRITE && zio->io_bp &&
+ !BP_IS_HOLE(zio->io_bp) && !(zio->io_flags & ZIO_FLAG_NOPWRITE)) {
+ metaslab_fastwrite_unmark(zio->io_spa, zio->io_bp);
+ }
+
/*
* It is the responsibility of the done callback to ensure that this
* particular zio is no longer discoverable for adoption, and as
zio_issue_async,
zio_write_bp_init,
zio_checksum_generate,
+ zio_nop_write,
zio_ddt_read_start,
zio_ddt_read_done,
zio_ddt_write,
zio_done
};
+/* dnp is the dnode for zb1->zb_object */
+boolean_t
+zbookmark_is_before(const dnode_phys_t *dnp, const zbookmark_t *zb1,
+ const zbookmark_t *zb2)
+{
+ uint64_t zb1nextL0, zb2thisobj;
+
+ ASSERT(zb1->zb_objset == zb2->zb_objset);
+ ASSERT(zb2->zb_level == 0);
+
+ /*
+ * A bookmark in the deadlist is considered to be after
+ * everything else.
+ */
+ if (zb2->zb_object == DMU_DEADLIST_OBJECT)
+ return (B_TRUE);
+
+ /* The objset_phys_t isn't before anything. */
+ if (dnp == NULL)
+ return (B_FALSE);
+
+ zb1nextL0 = (zb1->zb_blkid + 1) <<
+ ((zb1->zb_level) * (dnp->dn_indblkshift - SPA_BLKPTRSHIFT));
+
+ zb2thisobj = zb2->zb_object ? zb2->zb_object :
+ zb2->zb_blkid << (DNODE_BLOCK_SHIFT - DNODE_SHIFT);
+
+ if (zb1->zb_object == DMU_META_DNODE_OBJECT) {
+ uint64_t nextobj = zb1nextL0 *
+ (dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT) >> DNODE_SHIFT;
+ return (nextobj <= zb2thisobj);
+ }
+
+ if (zb1->zb_object < zb2thisobj)
+ return (B_TRUE);
+ if (zb1->zb_object > zb2thisobj)
+ return (B_FALSE);
+ if (zb2->zb_object == DMU_META_DNODE_OBJECT)
+ return (B_FALSE);
+ return (zb1nextL0 <= zb2->zb_blkid);
+}
+
#if defined(_KERNEL) && defined(HAVE_SPL)
/* Fault injection */
EXPORT_SYMBOL(zio_injection_enabled);
EXPORT_SYMBOL(zio_handle_fault_injection);
EXPORT_SYMBOL(zio_handle_device_injection);
EXPORT_SYMBOL(zio_handle_label_injection);
-EXPORT_SYMBOL(zio_priority_table);
EXPORT_SYMBOL(zio_type_name);
module_param(zio_bulk_flags, int, 0644);
module_param(zio_requeue_io_start_cut_in_line, int, 0644);
MODULE_PARM_DESC(zio_requeue_io_start_cut_in_line, "Prioritize requeued I/O");
+
+module_param(zfs_sync_pass_deferred_free, int, 0644);
+MODULE_PARM_DESC(zfs_sync_pass_deferred_free,
+ "Defer frees starting in this pass");
+
+module_param(zfs_sync_pass_dont_compress, int, 0644);
+MODULE_PARM_DESC(zfs_sync_pass_dont_compress,
+ "Don't compress starting in this pass");
+
+module_param(zfs_sync_pass_rewrite, int, 0644);
+MODULE_PARM_DESC(zfs_sync_pass_rewrite,
+ "Rewrite new bps starting in this pass");
#endif