c/mirror_zfs
1
0
Fork 0
mirror of https://git.proxmox.com/git/mirror_zfs.git synced 2026-05-24 11:18:52 +03:00
Code Issues Packages Projects Releases Wiki Activity

Prefetch on deadlists merge

Browse Source 
During snapshot deletion ZFS may issue several reads for each deadlist
to merge them into next snapshot's or pool's bpobj.  Number of the dead
lists increases with number of snapshots.  On HDD pools it may take
significant time during which sync thread is blocked.

This patch introduces prescient prefetch of required blocks for up to
128 deadlists ahead.  Tests show reduction of time required to delete
dataset with 720 snapshots with randomly overwritten file on wide HDD
pool from 75-85 to 22-28 seconds.

Reviewed-by: Allan Jude <allan@klarasystems.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by:	Alexander Motin <mav@FreeBSD.org>
Sponsored by:	iXsystems, Inc.
Issue #14276 
Closes #14402
This commit is contained in:
Alexander Motin
2023-01-25 14:30:24 -05:00
committed by GitHub
parent c85ac731a0
commit dc5c8006f6
3 changed files with 131 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files
module/zfs/bpobj.c
+63 -2
View File
@@ -663,14 +663,13 @@ bpobj_enqueue_subobj(bpobj_t *bpo, uint64_t subobj, dmu_tx_t *tx)
}
VERIFY3U(0, ==, bpobj_open(&subbpo, bpo->bpo_os, subobj));
VERIFY3U(0, ==, bpobj_space(&subbpo, &used, &comp, &uncomp));
if (bpobj_is_empty(&subbpo)) {
/* No point in having an empty subobj. */
bpobj_close(&subbpo);
bpobj_free(bpo->bpo_os, subobj, tx);
return;
}
VERIFY3U(0, ==, bpobj_space(&subbpo, &used, &comp, &uncomp));
mutex_enter(&bpo->bpo_lock);
dmu_buf_will_dirty(bpo->bpo_dbuf, tx);
@@ -780,6 +779,68 @@ bpobj_enqueue_subobj(bpobj_t *bpo, uint64_t subobj, dmu_tx_t *tx)
}
/*
* Prefetch metadata required for bpobj_enqueue_subobj().
*/
void
bpobj_prefetch_subobj(bpobj_t *bpo, uint64_t subobj)
{
dmu_object_info_t doi;
bpobj_t subbpo;
uint64_t subsubobjs;
boolean_t copy_subsub = B_TRUE;
boolean_t copy_bps = B_TRUE;
ASSERT(bpobj_is_open(bpo));
ASSERT(subobj != 0);
if (subobj == dmu_objset_pool(bpo->bpo_os)->dp_empty_bpobj)
return;
if (bpobj_open(&subbpo, bpo->bpo_os, subobj) != 0)
return;
if (bpobj_is_empty(&subbpo)) {
bpobj_close(&subbpo);
return;
}
subsubobjs = subbpo.bpo_phys->bpo_subobjs;
bpobj_close(&subbpo);
if (subsubobjs != 0) {
if (dmu_object_info(bpo->bpo_os, subsubobjs, &doi) != 0)
return;
if (doi.doi_max_offset > doi.doi_data_block_size)
copy_subsub = B_FALSE;
}
if (dmu_object_info(bpo->bpo_os, subobj, &doi) != 0)
return;
if (doi.doi_max_offset > doi.doi_data_block_size || !copy_subsub)
copy_bps = B_FALSE;
if (copy_subsub && subsubobjs != 0) {
if (bpo->bpo_phys->bpo_subobjs) {
dmu_prefetch(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs, 0,
bpo->bpo_phys->bpo_num_subobjs * sizeof (subobj), 1,
ZIO_PRIORITY_ASYNC_READ);
}
dmu_prefetch(bpo->bpo_os, subsubobjs, 0, 0, 1,
ZIO_PRIORITY_ASYNC_READ);
}
if (copy_bps) {
dmu_prefetch(bpo->bpo_os, bpo->bpo_object, 0,
bpo->bpo_phys->bpo_num_blkptrs * sizeof (blkptr_t), 1,
ZIO_PRIORITY_ASYNC_READ);
dmu_prefetch(bpo->bpo_os, subobj, 0, 0, 1,
ZIO_PRIORITY_ASYNC_READ);
} else if (bpo->bpo_phys->bpo_subobjs) {
dmu_prefetch(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs, 0,
bpo->bpo_phys->bpo_num_subobjs * sizeof (subobj), 1,
ZIO_PRIORITY_ASYNC_READ);
}
}
void
bpobj_enqueue(bpobj_t *bpo, const blkptr_t *bp, boolean_t bp_freed,
dmu_tx_t *tx)