summaryrefslogtreecommitdiffstats
path: root/module/zfs
diff options
context:
space:
mode:
Diffstat (limited to 'module/zfs')
-rw-r--r--module/zfs/dmu.c4
-rw-r--r--module/zfs/metaslab.c72
-rw-r--r--module/zfs/vdev.c2
-rw-r--r--module/zfs/zil.c52
-rw-r--r--module/zfs/zio.c23
5 files changed, 135 insertions, 18 deletions
diff --git a/module/zfs/dmu.c b/module/zfs/dmu.c
index 1d4d1257d..e2abf8cf2 100644
--- a/module/zfs/dmu.c
+++ b/module/zfs/dmu.c
@@ -1440,7 +1440,7 @@ dmu_sync_late_arrival(zio_t *pio, objset_t *os, dmu_sync_cb_t *done, zgd_t *zgd,
zio_nowait(zio_write(pio, os->os_spa, dmu_tx_get_txg(tx), zgd->zgd_bp,
zgd->zgd_db->db_data, zgd->zgd_db->db_size, zp,
dmu_sync_late_arrival_ready, dmu_sync_late_arrival_done, dsa,
- ZIO_PRIORITY_SYNC_WRITE, ZIO_FLAG_CANFAIL, zb));
+ ZIO_PRIORITY_SYNC_WRITE, ZIO_FLAG_CANFAIL | ZIO_FLAG_FASTWRITE, zb));
return (0);
}
@@ -1564,7 +1564,7 @@ dmu_sync(zio_t *pio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd)
zio_nowait(arc_write(pio, os->os_spa, txg,
bp, dr->dt.dl.dr_data, DBUF_IS_L2CACHEABLE(db), &zp,
dmu_sync_ready, dmu_sync_done, dsa,
- ZIO_PRIORITY_SYNC_WRITE, ZIO_FLAG_CANFAIL, &zb));
+ ZIO_PRIORITY_SYNC_WRITE, ZIO_FLAG_CANFAIL | ZIO_FLAG_FASTWRITE, &zb));
return (0);
}
diff --git a/module/zfs/metaslab.c b/module/zfs/metaslab.c
index d06012ffb..d199921b7 100644
--- a/module/zfs/metaslab.c
+++ b/module/zfs/metaslab.c
@@ -107,6 +107,7 @@ metaslab_class_create(spa_t *spa, space_map_ops_t *ops)
mc->mc_spa = spa;
mc->mc_rotor = NULL;
mc->mc_ops = ops;
+ mutex_init(&mc->mc_fastwrite_lock, NULL, MUTEX_DEFAULT, NULL);
return (mc);
}
@@ -120,6 +121,7 @@ metaslab_class_destroy(metaslab_class_t *mc)
ASSERT(mc->mc_space == 0);
ASSERT(mc->mc_dspace == 0);
+ mutex_destroy(&mc->mc_fastwrite_lock);
kmem_free(mc, sizeof (metaslab_class_t));
}
@@ -1307,7 +1309,7 @@ static int
metaslab_alloc_dva(spa_t *spa, metaslab_class_t *mc, uint64_t psize,
dva_t *dva, int d, dva_t *hintdva, uint64_t txg, int flags)
{
- metaslab_group_t *mg, *rotor;
+ metaslab_group_t *mg, *fast_mg, *rotor;
vdev_t *vd;
int dshift = 3;
int all_zero;
@@ -1325,6 +1327,9 @@ metaslab_alloc_dva(spa_t *spa, metaslab_class_t *mc, uint64_t psize,
if (psize >= metaslab_gang_bang && (ddi_get_lbolt() & 3) == 0)
return (ENOSPC);
+ if (flags & METASLAB_FASTWRITE)
+ mutex_enter(&mc->mc_fastwrite_lock);
+
/*
* Start at the rotor and loop through all mgs until we find something.
* Note that there's no locking on mc_rotor or mc_aliquot because
@@ -1367,6 +1372,15 @@ metaslab_alloc_dva(spa_t *spa, metaslab_class_t *mc, uint64_t psize,
} else if (d != 0) {
vd = vdev_lookup_top(spa, DVA_GET_VDEV(&dva[d - 1]));
mg = vd->vdev_mg->mg_next;
+ } else if (flags & METASLAB_FASTWRITE) {
+ mg = fast_mg = mc->mc_rotor;
+
+ do {
+ if (fast_mg->mg_vd->vdev_pending_fastwrite <
+ mg->mg_vd->vdev_pending_fastwrite)
+ mg = fast_mg;
+ } while ((fast_mg = fast_mg->mg_next) != mc->mc_rotor);
+
} else {
mg = mc->mc_rotor;
}
@@ -1453,7 +1467,8 @@ top:
(int64_t)mg->mg_aliquot) / 100;
}
- if (atomic_add_64_nv(&mc->mc_aliquot, asize) >=
+ if ((flags & METASLAB_FASTWRITE) ||
+ atomic_add_64_nv(&mc->mc_aliquot, asize) >=
mg->mg_aliquot + mg->mg_bias) {
mc->mc_rotor = mg->mg_next;
mc->mc_aliquot = 0;
@@ -1464,6 +1479,12 @@ top:
DVA_SET_GANG(&dva[d], !!(flags & METASLAB_GANG_HEADER));
DVA_SET_ASIZE(&dva[d], asize);
+ if (flags & METASLAB_FASTWRITE) {
+ atomic_add_64(&vd->vdev_pending_fastwrite,
+ psize);
+ mutex_exit(&mc->mc_fastwrite_lock);
+ }
+
return (0);
}
next:
@@ -1485,6 +1506,8 @@ next:
bzero(&dva[d], sizeof (dva_t));
+ if (flags & METASLAB_FASTWRITE)
+ mutex_exit(&mc->mc_fastwrite_lock);
return (ENOSPC);
}
@@ -1678,3 +1701,48 @@ metaslab_claim(spa_t *spa, const blkptr_t *bp, uint64_t txg)
return (error);
}
+
+void metaslab_fastwrite_mark(spa_t *spa, const blkptr_t *bp)
+{
+ const dva_t *dva = bp->blk_dva;
+ int ndvas = BP_GET_NDVAS(bp);
+ uint64_t psize = BP_GET_PSIZE(bp);
+ int d;
+ vdev_t *vd;
+
+ ASSERT(!BP_IS_HOLE(bp));
+ ASSERT(psize > 0);
+
+ spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
+
+ for (d = 0; d < ndvas; d++) {
+ if ((vd = vdev_lookup_top(spa, DVA_GET_VDEV(&dva[d]))) == NULL)
+ continue;
+ atomic_add_64(&vd->vdev_pending_fastwrite, psize);
+ }
+
+ spa_config_exit(spa, SCL_VDEV, FTAG);
+}
+
+void metaslab_fastwrite_unmark(spa_t *spa, const blkptr_t *bp)
+{
+ const dva_t *dva = bp->blk_dva;
+ int ndvas = BP_GET_NDVAS(bp);
+ uint64_t psize = BP_GET_PSIZE(bp);
+ int d;
+ vdev_t *vd;
+
+ ASSERT(!BP_IS_HOLE(bp));
+ ASSERT(psize > 0);
+
+ spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
+
+ for (d = 0; d < ndvas; d++) {
+ if ((vd = vdev_lookup_top(spa, DVA_GET_VDEV(&dva[d]))) == NULL)
+ continue;
+ ASSERT3U(vd->vdev_pending_fastwrite, >=, psize);
+ atomic_sub_64(&vd->vdev_pending_fastwrite, psize);
+ }
+
+ spa_config_exit(spa, SCL_VDEV, FTAG);
+}
diff --git a/module/zfs/vdev.c b/module/zfs/vdev.c
index 205a1d1aa..7d6d5278a 100644
--- a/module/zfs/vdev.c
+++ b/module/zfs/vdev.c
@@ -904,6 +904,8 @@ vdev_metaslab_fini(vdev_t *vd)
kmem_free(vd->vdev_ms, count * sizeof (metaslab_t *));
vd->vdev_ms = NULL;
}
+
+ ASSERT3U(vd->vdev_pending_fastwrite, ==, 0);
}
typedef struct vdev_probe_stats {
diff --git a/module/zfs/zil.c b/module/zfs/zil.c
index e76e5ecf1..6492dbc1c 100644
--- a/module/zfs/zil.c
+++ b/module/zfs/zil.c
@@ -38,6 +38,7 @@
#include <sys/vdev_impl.h>
#include <sys/dmu_tx.h>
#include <sys/dsl_pool.h>
+#include <sys/metaslab.h>
/*
* The zfs intent log (ZIL) saves transaction records of system calls
@@ -451,13 +452,14 @@ zil_free_log_record(zilog_t *zilog, lr_t *lrc, void *tx, uint64_t claim_txg)
}
static lwb_t *
-zil_alloc_lwb(zilog_t *zilog, blkptr_t *bp, uint64_t txg)
+zil_alloc_lwb(zilog_t *zilog, blkptr_t *bp, uint64_t txg, boolean_t fastwrite)
{
lwb_t *lwb;
lwb = kmem_cache_alloc(zil_lwb_cache, KM_PUSHPAGE);
lwb->lwb_zilog = zilog;
lwb->lwb_blk = *bp;
+ lwb->lwb_fastwrite = fastwrite;
lwb->lwb_buf = zio_buf_alloc(BP_GET_LSIZE(bp));
lwb->lwb_max_txg = txg;
lwb->lwb_zio = NULL;
@@ -489,6 +491,7 @@ zil_create(zilog_t *zilog)
dmu_tx_t *tx = NULL;
blkptr_t blk;
int error = 0;
+ boolean_t fastwrite = FALSE;
/*
* Wait for any previous destroy to complete.
@@ -516,8 +519,9 @@ zil_create(zilog_t *zilog)
BP_ZERO(&blk);
}
- error = zio_alloc_zil(zilog->zl_spa, txg, &blk, NULL,
+ error = zio_alloc_zil(zilog->zl_spa, txg, &blk,
ZIL_MIN_BLKSZ, zilog->zl_logbias == ZFS_LOGBIAS_LATENCY);
+ fastwrite = TRUE;
if (error == 0)
zil_init_log_chain(zilog, &blk);
@@ -527,7 +531,7 @@ zil_create(zilog_t *zilog)
* Allocate a log write buffer (lwb) for the first log block.
*/
if (error == 0)
- lwb = zil_alloc_lwb(zilog, &blk, txg);
+ lwb = zil_alloc_lwb(zilog, &blk, txg, fastwrite);
/*
* If we just allocated the first log block, commit our transaction
@@ -586,6 +590,10 @@ zil_destroy(zilog_t *zilog, boolean_t keep_first)
ASSERT(zh->zh_claim_txg == 0);
VERIFY(!keep_first);
while ((lwb = list_head(&zilog->zl_lwb_list)) != NULL) {
+ ASSERT(lwb->lwb_zio == NULL);
+ if (lwb->lwb_fastwrite)
+ metaslab_fastwrite_unmark(zilog->zl_spa,
+ &lwb->lwb_blk);
list_remove(&zilog->zl_lwb_list, lwb);
if (lwb->lwb_buf != NULL)
zio_buf_free(lwb->lwb_buf, lwb->lwb_sz);
@@ -826,6 +834,8 @@ zil_lwb_write_done(zio_t *zio)
*/
zio_buf_free(lwb->lwb_buf, lwb->lwb_sz);
mutex_enter(&zilog->zl_lock);
+ lwb->lwb_zio = NULL;
+ lwb->lwb_fastwrite = FALSE;
lwb->lwb_buf = NULL;
lwb->lwb_tx = NULL;
mutex_exit(&zilog->zl_lock);
@@ -854,12 +864,21 @@ zil_lwb_write_init(zilog_t *zilog, lwb_t *lwb)
zilog->zl_root_zio = zio_root(zilog->zl_spa, NULL, NULL,
ZIO_FLAG_CANFAIL);
}
+
+ /* Lock so zil_sync() doesn't fastwrite_unmark after zio is created */
+ mutex_enter(&zilog->zl_lock);
if (lwb->lwb_zio == NULL) {
+ if (!lwb->lwb_fastwrite) {
+ metaslab_fastwrite_mark(zilog->zl_spa, &lwb->lwb_blk);
+ lwb->lwb_fastwrite = 1;
+ }
lwb->lwb_zio = zio_rewrite(zilog->zl_root_zio, zilog->zl_spa,
0, &lwb->lwb_blk, lwb->lwb_buf, BP_GET_LSIZE(&lwb->lwb_blk),
zil_lwb_write_done, lwb, ZIO_PRIORITY_LOG_WRITE,
- ZIO_FLAG_CANFAIL | ZIO_FLAG_DONT_PROPAGATE, &zb);
+ ZIO_FLAG_CANFAIL | ZIO_FLAG_DONT_PROPAGATE |
+ ZIO_FLAG_FASTWRITE, &zb);
}
+ mutex_exit(&zilog->zl_lock);
}
/*
@@ -956,10 +975,8 @@ zil_lwb_write_start(zilog_t *zilog, lwb_t *lwb)
zilog->zl_prev_rotor = (zilog->zl_prev_rotor + 1) & (ZIL_PREV_BLKS - 1);
BP_ZERO(bp);
- /* pass the old blkptr in order to spread log blocks across devs */
use_slog = USE_SLOG(zilog);
- error = zio_alloc_zil(spa, txg, bp, &lwb->lwb_blk, zil_blksz,
- use_slog);
+ error = zio_alloc_zil(spa, txg, bp, zil_blksz, USE_SLOG(zilog));
if (use_slog)
{
ZIL_STAT_BUMP(zil_itx_metaslab_slog_count);
@@ -978,7 +995,7 @@ zil_lwb_write_start(zilog_t *zilog, lwb_t *lwb)
/*
* Allocate a new log write buffer (lwb).
*/
- nlwb = zil_alloc_lwb(zilog, bp, txg);
+ nlwb = zil_alloc_lwb(zilog, bp, txg, TRUE);
/* Record the block for later vdev flushing */
zil_add_block(zilog, &lwb->lwb_blk);
@@ -1625,6 +1642,9 @@ zil_sync(zilog_t *zilog, dmu_tx_t *tx)
zh->zh_log = lwb->lwb_blk;
if (lwb->lwb_buf != NULL || lwb->lwb_max_txg > txg)
break;
+
+ ASSERT(lwb->lwb_zio == NULL);
+
list_remove(&zilog->zl_lwb_list, lwb);
zio_free_zil(spa, txg, &lwb->lwb_blk);
kmem_cache_free(zil_lwb_cache, lwb);
@@ -1638,6 +1658,19 @@ zil_sync(zilog_t *zilog, dmu_tx_t *tx)
if (list_head(&zilog->zl_lwb_list) == NULL)
BP_ZERO(&zh->zh_log);
}
+
+ /*
+ * Remove fastwrite on any blocks that have been pre-allocated for
+ * the next commit. This prevents fastwrite counter pollution by
+ * unused, long-lived LWBs.
+ */
+ for (; lwb != NULL; lwb = list_next(&zilog->zl_lwb_list, lwb)) {
+ if (lwb->lwb_fastwrite && !lwb->lwb_zio) {
+ metaslab_fastwrite_unmark(zilog->zl_spa, &lwb->lwb_blk);
+ lwb->lwb_fastwrite = 0;
+ }
+ }
+
mutex_exit(&zilog->zl_lock);
}
@@ -1817,6 +1850,9 @@ zil_close(zilog_t *zilog)
lwb = list_head(&zilog->zl_lwb_list);
if (lwb != NULL) {
ASSERT(lwb == list_tail(&zilog->zl_lwb_list));
+ ASSERT(lwb->lwb_zio == NULL);
+ if (lwb->lwb_fastwrite)
+ metaslab_fastwrite_unmark(zilog->zl_spa, &lwb->lwb_blk);
list_remove(&zilog->zl_lwb_list, lwb);
zio_buf_free(lwb->lwb_buf, lwb->lwb_sz);
kmem_cache_free(zil_lwb_cache, lwb);
diff --git a/module/zfs/zio.c b/module/zfs/zio.c
index ace72a087..ce76e010c 100644
--- a/module/zfs/zio.c
+++ b/module/zfs/zio.c
@@ -1861,6 +1861,11 @@ zio_write_gang_block(zio_t *pio)
*/
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);
@@ -2270,6 +2275,7 @@ zio_dva_allocate(zio_t *zio)
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);
@@ -2333,8 +2339,8 @@ zio_dva_unallocate(zio_t *zio, zio_gang_node_t *gn, blkptr_t *bp)
* 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;
@@ -2347,14 +2353,14 @@ zio_alloc_zil(spa_t *spa, uint64_t txg, blkptr_t *new_bp, blkptr_t *old_bp,
*/
if (use_slog) {
error = metaslab_alloc(spa, spa_log_class(spa), size,
- new_bp, 1, txg, old_bp,
- METASLAB_HINTBP_AVOID | METASLAB_GANG_AVOID);
+ new_bp, 1, txg, NULL,
+ METASLAB_FASTWRITE | METASLAB_GANG_AVOID);
}
if (error) {
error = metaslab_alloc(spa, spa_normal_class(spa), size,
- new_bp, 1, txg, old_bp,
- METASLAB_HINTBP_AVOID | METASLAB_GANG_AVOID);
+ new_bp, 1, txg, NULL,
+ METASLAB_FASTWRITE | METASLAB_GANG_AVOID);
}
if (error == 0) {
@@ -3066,6 +3072,11 @@ zio_done(zio_t *zio)
zfs_ereport_free_checksum(zcr);
}
+ if (zio->io_flags & ZIO_FLAG_FASTWRITE && zio->io_bp &&
+ !BP_IS_HOLE(zio->io_bp)) {
+ 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
generated by cgit v1.2.3 (git 2.39.1) at 2024-12-04 18:30:50 +0000