c/mirror_zfs
1
0
Fork 0
mirror of https://git.proxmox.com/git/mirror_zfs.git synced 2024-12-25 18:59:33 +03:00
Code Issues Packages Projects Releases Wiki Activity

More speculative prefetcher improvements

Browse Source 
- Make prefetch distance adaptive: up to 4MB prefetch doubles for
every, hit same as before, but after that it grows by 1/8 every time
the prefetch read does not complete in time to satisfy the demand.
My tests show that 4MB is sufficient for wide NVMe pool to saturate
single reader thread at 2.5GB/s, while new 64MB maximum allows the
same thread to reach 1.5GB/s on wide HDD pool.  Further distance
increase may increase speed even more, but less dramatic and with
higher latency.

 - Allow early reuse of inactive prefetch streams: streams that never
saw hits can be reused immediately if there is a demand, while others
can be reused after 1s of inactivity, starting with the oldest.  After
2s of inactivity streams are deleted to free resources same as before.
This allows by several times increase strided read performance on HDD
pool in presence of simultaneous random reads, previously filling the
zfetch_max_streams limit for seconds and so blocking most of prefetch.

 - Always issue intermediate indirect block reads with SYNC priority.
Each of those reads if delayed for longer may delay up to 1024 other
block prefetches, that may be not good for wide pools.

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.
Closes #13452
This commit is contained in:
Alexander Motin 2022-05-25 13:12:52 -04:00 committed by GitHub
parent 1d89b989c1
commit 6aa8c21a2a
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
5 changed files with 133 additions and 101 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
include/sys/dbuf.h
View File

@ -329,7 +329,7 @@ typedef struct dbuf_hash_table {
krwlock_t hash_rwlocks[DBUF_RWLOCKS] ____cacheline_aligned; krwlock_t hash_rwlocks[DBUF_RWLOCKS] ____cacheline_aligned;
} dbuf_hash_table_t; } dbuf_hash_table_t;
typedef void (*dbuf_prefetch_fn)(void *, boolean_t); typedef void (*dbuf_prefetch_fn)(void *, uint64_t, uint64_t, boolean_t);
uint64_t dbuf_whichblock(const struct dnode *di, const int64_t level, uint64_t dbuf_whichblock(const struct dnode *di, const int64_t level,
const uint64_t offset); const uint64_t offset);

16
include/sys/dmu_zfetch.h
View File

@ -49,20 +49,18 @@ typedef struct zfetch {
typedef struct zstream { typedef struct zstream {
uint64_t zs_blkid; /* expect next access at this blkid */ uint64_t zs_blkid; /* expect next access at this blkid */
uint64_t zs_pf_blkid1; /* first block to prefetch */ unsigned int zs_pf_dist; /* data prefetch distance in bytes */
uint64_t zs_pf_blkid; /* block to prefetch up to */ unsigned int zs_ipf_dist; /* L1 prefetch distance in bytes */
uint64_t zs_pf_start; /* first data block to prefetch */
/* uint64_t zs_pf_end; /* data block to prefetch up to */
* We will next prefetch the L1 indirect block of this level-0 uint64_t zs_ipf_start; /* first data block to prefetch L1 */
* block id. uint64_t zs_ipf_end; /* data block to prefetch L1 up to */
*/
uint64_t zs_ipf_blkid1; /* first block to prefetch */
uint64_t zs_ipf_blkid; /* block to prefetch up to */
list_node_t zs_node; /* link for zf_stream */ list_node_t zs_node; /* link for zf_stream */
hrtime_t zs_atime; /* time last prefetch issued */ hrtime_t zs_atime; /* time last prefetch issued */
zfetch_t *zs_fetch; /* parent fetch */ zfetch_t *zs_fetch; /* parent fetch */
boolean_t zs_missed; /* stream saw cache misses */ boolean_t zs_missed; /* stream saw cache misses */
boolean_t zs_more; /* need more distant prefetch */
zfs_refcount_t zs_callers; /* number of pending callers */ zfs_refcount_t zs_callers; /* number of pending callers */
/* /*
* Number of stream references: dnode, callers and pending blocks. * Number of stream references: dnode, callers and pending blocks.

17
man/man4/zfs.4
View File

@ -487,7 +487,15 @@ However, this is limited by
.It Sy zfetch_array_rd_sz Ns = Ns Sy 1048576 Ns B Po 1 MiB Pc Pq ulong .It Sy zfetch_array_rd_sz Ns = Ns Sy 1048576 Ns B Po 1 MiB Pc Pq ulong
If prefetching is enabled, disable prefetching for reads larger than this size. If prefetching is enabled, disable prefetching for reads larger than this size.
. .
.It Sy zfetch_max_distance Ns = Ns Sy 8388608 Ns B Po 8 MiB Pc Pq uint .It Sy zfetch_min_distance Ns = Ns Sy 4194304 Ns B Po 4 MiB Pc Pq uint
Min bytes to prefetch per stream.
Prefetch distance starts from the demand access size and quickly grows to
this value, doubling on each hit.
After that it may grow further by 1/8 per hit, but only if some prefetch
since last time haven't completed in time to satisfy demand request, i.e.
prefetch depth didn't cover the read latency or the pool got saturated.
.
.It Sy zfetch_max_distance Ns = Ns Sy 67108864 Ns B Po 64 MiB Pc Pq uint
Max bytes to prefetch per stream. Max bytes to prefetch per stream.
. .
.It Sy zfetch_max_idistance Ns = Ns Sy 67108864 Ns B Po 64 MiB Pc Pq uint .It Sy zfetch_max_idistance Ns = Ns Sy 67108864 Ns B Po 64 MiB Pc Pq uint
@ -496,8 +504,11 @@ Max bytes to prefetch indirects for per stream.
.It Sy zfetch_max_streams Ns = Ns Sy 8 Pq uint .It Sy zfetch_max_streams Ns = Ns Sy 8 Pq uint
Max number of streams per zfetch (prefetch streams per file). Max number of streams per zfetch (prefetch streams per file).
. .
.It Sy zfetch_min_sec_reap Ns = Ns Sy 2 Pq uint .It Sy zfetch_min_sec_reap Ns = Ns Sy 1 Pq uint
Min time before an active prefetch stream can be reclaimed Min time before inactive prefetch stream can be reclaimed
.
.It Sy zfetch_max_sec_reap Ns = Ns Sy 2 Pq uint
Max time before inactive prefetch stream can be deleted
. .
.It Sy zfs_abd_scatter_enabled Ns = Ns Sy 1 Ns | Ns 0 Pq int .It Sy zfs_abd_scatter_enabled Ns = Ns Sy 1 Ns | Ns 0 Pq int
Enables ARC from using scatter/gather lists and forces all allocations to be Enables ARC from using scatter/gather lists and forces all allocations to be

14
module/zfs/dbuf.c
View File

@ -3185,8 +3185,10 @@ typedef struct dbuf_prefetch_arg {
static void static void
dbuf_prefetch_fini(dbuf_prefetch_arg_t *dpa, boolean_t io_done) dbuf_prefetch_fini(dbuf_prefetch_arg_t *dpa, boolean_t io_done)
{ {
if (dpa->dpa_cb != NULL) if (dpa->dpa_cb != NULL) {
dpa->dpa_cb(dpa->dpa_arg, io_done); dpa->dpa_cb(dpa->dpa_arg, dpa->dpa_zb.zb_level,
dpa->dpa_zb.zb_blkid, io_done);
}
kmem_free(dpa, sizeof (*dpa)); kmem_free(dpa, sizeof (*dpa));
} }
@ -3320,7 +3322,8 @@ dbuf_prefetch_indirect_done(zio_t *zio, const zbookmark_phys_t *zb,
dpa->dpa_zb.zb_object, dpa->dpa_curlevel, nextblkid); dpa->dpa_zb.zb_object, dpa->dpa_curlevel, nextblkid);
(void) arc_read(dpa->dpa_zio, dpa->dpa_spa, (void) arc_read(dpa->dpa_zio, dpa->dpa_spa,
bp, dbuf_prefetch_indirect_done, dpa, dpa->dpa_prio, bp, dbuf_prefetch_indirect_done, dpa,
ZIO_PRIORITY_SYNC_READ,
ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE, ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE,
&iter_aflags, &zb); &iter_aflags, &zb);
} }
@ -3455,7 +3458,8 @@ dbuf_prefetch_impl(dnode_t *dn, int64_t level, uint64_t blkid,
SET_BOOKMARK(&zb, ds != NULL ? ds->ds_object : DMU_META_OBJSET, SET_BOOKMARK(&zb, ds != NULL ? ds->ds_object : DMU_META_OBJSET,
dn->dn_object, curlevel, curblkid); dn->dn_object, curlevel, curblkid);
(void) arc_read(dpa->dpa_zio, dpa->dpa_spa, (void) arc_read(dpa->dpa_zio, dpa->dpa_spa,
&bp, dbuf_prefetch_indirect_done, dpa, prio, &bp, dbuf_prefetch_indirect_done, dpa,
ZIO_PRIORITY_SYNC_READ,
ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE, ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE,
&iter_aflags, &zb); &iter_aflags, &zb);
} }
@ -3467,7 +3471,7 @@ dbuf_prefetch_impl(dnode_t *dn, int64_t level, uint64_t blkid,
return (1); return (1);
no_issue: no_issue:
if (cb != NULL) if (cb != NULL)
cb(arg, B_FALSE); cb(arg, level, blkid, B_FALSE);
return (0); return (0);
} }

185
module/zfs/dmu_zfetch.c
View File

@ -48,9 +48,13 @@ static int zfs_prefetch_disable = B_FALSE;
/* max # of streams per zfetch */ /* max # of streams per zfetch */
static unsigned int zfetch_max_streams = 8; static unsigned int zfetch_max_streams = 8;
/* min time before stream reclaim */ /* min time before stream reclaim */
static unsigned int zfetch_min_sec_reap = 2; static unsigned int zfetch_min_sec_reap = 1;
/* max bytes to prefetch per stream (default 8MB) */ /* max time before stream delete */
unsigned int zfetch_max_distance = 8 * 1024 * 1024; static unsigned int zfetch_max_sec_reap = 2;
/* min bytes to prefetch per stream (default 4MB) */
static unsigned int zfetch_min_distance = 4 * 1024 * 1024;
/* max bytes to prefetch per stream (default 64MB) */
unsigned int zfetch_max_distance = 64 * 1024 * 1024;
/* max bytes to prefetch indirects for per stream (default 64MB) */ /* max bytes to prefetch indirects for per stream (default 64MB) */
unsigned int zfetch_max_idistance = 64 * 1024 * 1024; unsigned int zfetch_max_idistance = 64 * 1024 * 1024;
/* max number of bytes in an array_read in which we allow prefetching (1MB) */ /* max number of bytes in an array_read in which we allow prefetching (1MB) */
@ -195,74 +199,99 @@ dmu_zfetch_fini(zfetch_t *zf)
} }
/* /*
* If there aren't too many streams already, create a new stream. * If there aren't too many active streams already, create one more.
* In process delete/reuse all streams without hits for zfetch_max_sec_reap.
* If needed, reuse oldest stream without hits for zfetch_min_sec_reap or ever.
* The "blkid" argument is the next block that we expect this stream to access. * The "blkid" argument is the next block that we expect this stream to access.
* While we're here, clean up old streams (which haven't been
* accessed for at least zfetch_min_sec_reap seconds).
*/ */
static void static void
dmu_zfetch_stream_create(zfetch_t *zf, uint64_t blkid) dmu_zfetch_stream_create(zfetch_t *zf, uint64_t blkid)
{ {
zstream_t *zs_next; zstream_t *zs, *zs_next, *zs_old = NULL;
hrtime_t now = gethrtime(); hrtime_t now = gethrtime(), t;
ASSERT(MUTEX_HELD(&zf->zf_lock)); ASSERT(MUTEX_HELD(&zf->zf_lock));
/* /*
* Clean up old streams. * Delete too old streams, reusing the first found one.
*/ */
for (zstream_t *zs = list_head(&zf->zf_stream); t = now - SEC2NSEC(zfetch_max_sec_reap);
zs != NULL; zs = zs_next) { for (zs = list_head(&zf->zf_stream); zs != NULL; zs = zs_next) {
zs_next = list_next(&zf->zf_stream, zs); zs_next = list_next(&zf->zf_stream, zs);
/* /*
* Skip if still active. 1 -- zf_stream reference. * Skip if still active. 1 -- zf_stream reference.
*/ */
if (zfs_refcount_count(&zs->zs_refs) != 1) if (zfs_refcount_count(&zs->zs_refs) != 1)
continue; continue;
if (((now - zs->zs_atime) / NANOSEC) > if (zs->zs_atime > t)
zfetch_min_sec_reap) continue;
if (zs_old)
dmu_zfetch_stream_remove(zf, zs); dmu_zfetch_stream_remove(zf, zs);
else
zs_old = zs;
}
if (zs_old) {
zs = zs_old;
goto reuse;
} }
/* /*
* The maximum number of streams is normally zfetch_max_streams, * The maximum number of streams is normally zfetch_max_streams,
* but for small files we lower it such that it's at least possible * but for small files we lower it such that it's at least possible
* for all the streams to be non-overlapping. * for all the streams to be non-overlapping.
*
* If we are already at the maximum number of streams for this file,
* even after removing old streams, then don't create this stream.
*/ */
uint32_t max_streams = MAX(1, MIN(zfetch_max_streams, uint32_t max_streams = MAX(1, MIN(zfetch_max_streams,
zf->zf_dnode->dn_maxblkid * zf->zf_dnode->dn_datablksz / zf->zf_dnode->dn_maxblkid * zf->zf_dnode->dn_datablksz /
zfetch_max_distance)); zfetch_max_distance));
if (zf->zf_numstreams >= max_streams) { if (zf->zf_numstreams >= max_streams) {
t = now - SEC2NSEC(zfetch_min_sec_reap);
for (zs = list_head(&zf->zf_stream); zs != NULL;
zs = list_next(&zf->zf_stream, zs)) {
if (zfs_refcount_count(&zs->zs_refs) != 1)
continue;
if (zs->zs_atime > t)
continue;
if (zs_old == NULL || zs->zs_atime < zs_old->zs_atime)
zs_old = zs;
}
if (zs_old) {
zs = zs_old;
goto reuse;
}
ZFETCHSTAT_BUMP(zfetchstat_max_streams); ZFETCHSTAT_BUMP(zfetchstat_max_streams);
return; return;
} }
zstream_t *zs = kmem_zalloc(sizeof (*zs), KM_SLEEP); zs = kmem_zalloc(sizeof (*zs), KM_SLEEP);
zs->zs_blkid = blkid;
zs->zs_pf_blkid1 = blkid;
zs->zs_pf_blkid = blkid;
zs->zs_ipf_blkid1 = blkid;
zs->zs_ipf_blkid = blkid;
zs->zs_atime = now;
zs->zs_fetch = zf; zs->zs_fetch = zf;
zs->zs_missed = B_FALSE;
zfs_refcount_create(&zs->zs_callers); zfs_refcount_create(&zs->zs_callers);
zfs_refcount_create(&zs->zs_refs); zfs_refcount_create(&zs->zs_refs);
/* One reference for zf_stream. */ /* One reference for zf_stream. */
zfs_refcount_add(&zs->zs_refs, NULL); zfs_refcount_add(&zs->zs_refs, NULL);
zf->zf_numstreams++; zf->zf_numstreams++;
list_insert_head(&zf->zf_stream, zs); list_insert_head(&zf->zf_stream, zs);
reuse:
zs->zs_blkid = blkid;
zs->zs_pf_dist = 0;
zs->zs_pf_start = blkid;
zs->zs_pf_end = blkid;
zs->zs_ipf_dist = 0;
zs->zs_ipf_start = blkid;
zs->zs_ipf_end = blkid;
/* Allow immediate stream reuse until first hit. */
zs->zs_atime = now - SEC2NSEC(zfetch_min_sec_reap);
zs->zs_missed = B_FALSE;
zs->zs_more = B_FALSE;
} }
static void static void
dmu_zfetch_stream_done(void *arg, boolean_t io_issued) dmu_zfetch_done(void *arg, uint64_t level, uint64_t blkid, boolean_t io_issued)
{ {
(void) io_issued;
zstream_t *zs = arg; zstream_t *zs = arg;
if (io_issued && level == 0 && blkid < zs->zs_blkid)
zs->zs_more = B_TRUE;
if (zfs_refcount_remove(&zs->zs_refs, NULL) == 0) if (zfs_refcount_remove(&zs->zs_refs, NULL) == 0)
dmu_zfetch_stream_fini(zs); dmu_zfetch_stream_fini(zs);
} }
@ -284,11 +313,6 @@ dmu_zfetch_prepare(zfetch_t *zf, uint64_t blkid, uint64_t nblks,
boolean_t fetch_data, boolean_t have_lock) boolean_t fetch_data, boolean_t have_lock)
{ {
zstream_t *zs; zstream_t *zs;
int64_t pf_start, ipf_start;
int64_t pf_ahead_blks, max_blks;
int max_dist_blks, pf_nblks, ipf_nblks;
uint64_t end_of_access_blkid, maxblkid;
end_of_access_blkid = blkid + nblks;
spa_t *spa = zf->zf_dnode->dn_objset->os_spa; spa_t *spa = zf->zf_dnode->dn_objset->os_spa;
if (zfs_prefetch_disable) if (zfs_prefetch_disable)
@ -317,7 +341,7 @@ dmu_zfetch_prepare(zfetch_t *zf, uint64_t blkid, uint64_t nblks,
* A fast path for small files for which no prefetch will * A fast path for small files for which no prefetch will
* happen. * happen.
*/ */
maxblkid = zf->zf_dnode->dn_maxblkid; uint64_t maxblkid = zf->zf_dnode->dn_maxblkid;
if (maxblkid < 2) { if (maxblkid < 2) {
if (!have_lock) if (!have_lock)
rw_exit(&zf->zf_dnode->dn_struct_rwlock); rw_exit(&zf->zf_dnode->dn_struct_rwlock);
@ -345,6 +369,7 @@ dmu_zfetch_prepare(zfetch_t *zf, uint64_t blkid, uint64_t nblks,
* If the file is ending, remove the matching stream if found. * If the file is ending, remove the matching stream if found.
* If not found then it is too late to create a new one now. * If not found then it is too late to create a new one now.
*/ */
uint64_t end_of_access_blkid = blkid + nblks;
if (end_of_access_blkid >= maxblkid) { if (end_of_access_blkid >= maxblkid) {
if (zs != NULL) if (zs != NULL)
dmu_zfetch_stream_remove(zf, zs); dmu_zfetch_stream_remove(zf, zs);
@ -377,60 +402,48 @@ dmu_zfetch_prepare(zfetch_t *zf, uint64_t blkid, uint64_t nblks,
/* /*
* This access was to a block that we issued a prefetch for on * This access was to a block that we issued a prefetch for on
* behalf of this stream. Issue further prefetches for this stream. * behalf of this stream. Calculate further prefetch distances.
* *
* Normally, we start prefetching where we stopped * Start prefetch from the demand access size (nblks). Double the
* prefetching last (zs_pf_blkid). But when we get our first * distance every access up to zfetch_min_distance. After that only
* hit on this stream, zs_pf_blkid == zs_blkid, we don't * if needed increase the distance by 1/8 up to zfetch_max_distance.
* want to prefetch the block we just accessed. In this case,
* start just after the block we just accessed.
*/
pf_start = MAX(zs->zs_pf_blkid, end_of_access_blkid);
if (zs->zs_pf_blkid1 < end_of_access_blkid)
zs->zs_pf_blkid1 = end_of_access_blkid;
if (zs->zs_ipf_blkid1 < end_of_access_blkid)
zs->zs_ipf_blkid1 = end_of_access_blkid;
/*
* Double our amount of prefetched data, but don't let the
* prefetch get further ahead than zfetch_max_distance.
*/ */
unsigned int nbytes = nblks << zf->zf_dnode->dn_datablkshift;
unsigned int pf_nblks;
if (fetch_data) { if (fetch_data) {
max_dist_blks = if (unlikely(zs->zs_pf_dist < nbytes))
zfetch_max_distance >> zf->zf_dnode->dn_datablkshift; zs->zs_pf_dist = nbytes;
/* else if (zs->zs_pf_dist < zfetch_min_distance)
* Previously, we were (zs_pf_blkid - blkid) ahead. We zs->zs_pf_dist *= 2;
* want to now be double that, so read that amount again, else if (zs->zs_more)
* plus the amount we are catching up by (i.e. the amount zs->zs_pf_dist += zs->zs_pf_dist / 8;
* read just now). zs->zs_more = B_FALSE;
*/ if (zs->zs_pf_dist > zfetch_max_distance)
pf_ahead_blks = zs->zs_pf_blkid - blkid + nblks; zs->zs_pf_dist = zfetch_max_distance;
max_blks = max_dist_blks - (pf_start - end_of_access_blkid); pf_nblks = zs->zs_pf_dist >> zf->zf_dnode->dn_datablkshift;
pf_nblks = MIN(pf_ahead_blks, max_blks);
} else { } else {
pf_nblks = 0; pf_nblks = 0;
} }
if (zs->zs_pf_start < end_of_access_blkid)
zs->zs_pf_blkid = pf_start + pf_nblks; zs->zs_pf_start = end_of_access_blkid;
if (zs->zs_pf_end < end_of_access_blkid + pf_nblks)
zs->zs_pf_end = end_of_access_blkid + pf_nblks;
/* /*
* Do the same for indirects, starting from where we stopped last, * Do the same for indirects, starting where we will stop reading
* or where we will stop reading data blocks (and the indirects * data blocks (and the indirects that point to them).
* that point to them).
*/ */
ipf_start = MAX(zs->zs_ipf_blkid, zs->zs_pf_blkid); if (unlikely(zs->zs_ipf_dist < nbytes))
max_dist_blks = zfetch_max_idistance >> zf->zf_dnode->dn_datablkshift; zs->zs_ipf_dist = nbytes;
/* else
* We want to double our distance ahead of the data prefetch zs->zs_ipf_dist *= 2;
* (or reader, if we are not prefetching data). Previously, we if (zs->zs_ipf_dist > zfetch_max_idistance)
* were (zs_ipf_blkid - blkid) ahead. To double that, we read zs->zs_ipf_dist = zfetch_max_idistance;
* that amount again, plus the amount we are catching up by pf_nblks = zs->zs_ipf_dist >> zf->zf_dnode->dn_datablkshift;
* (i.e. the amount read now + the amount of data prefetched now). if (zs->zs_ipf_start < zs->zs_pf_end)
*/ zs->zs_ipf_start = zs->zs_pf_end;
pf_ahead_blks = zs->zs_ipf_blkid - blkid + nblks + pf_nblks; if (zs->zs_ipf_end < zs->zs_pf_end + pf_nblks)
max_blks = max_dist_blks - (ipf_start - zs->zs_pf_blkid); zs->zs_ipf_end = zs->zs_pf_end + pf_nblks;
ipf_nblks = MIN(pf_ahead_blks, max_blks);
zs->zs_ipf_blkid = ipf_start + ipf_nblks;
zs->zs_blkid = end_of_access_blkid; zs->zs_blkid = end_of_access_blkid;
/* Protect the stream from reclamation. */ /* Protect the stream from reclamation. */
@ -471,13 +484,13 @@ dmu_zfetch_run(zstream_t *zs, boolean_t missed, boolean_t have_lock)
mutex_enter(&zf->zf_lock); mutex_enter(&zf->zf_lock);
if (zs->zs_missed) { if (zs->zs_missed) {
pf_start = zs->zs_pf_blkid1; pf_start = zs->zs_pf_start;
pf_end = zs->zs_pf_blkid1 = zs->zs_pf_blkid; pf_end = zs->zs_pf_start = zs->zs_pf_end;
} else { } else {
pf_start = pf_end = 0; pf_start = pf_end = 0;
} }
ipf_start = MAX(zs->zs_pf_blkid1, zs->zs_ipf_blkid1); ipf_start = zs->zs_ipf_start;
ipf_end = zs->zs_ipf_blkid1 = zs->zs_ipf_blkid; ipf_end = zs->zs_ipf_start = zs->zs_ipf_end;
mutex_exit(&zf->zf_lock); mutex_exit(&zf->zf_lock);
ASSERT3S(pf_start, <=, pf_end); ASSERT3S(pf_start, <=, pf_end);
ASSERT3S(ipf_start, <=, ipf_end); ASSERT3S(ipf_start, <=, ipf_end);
@ -505,12 +518,12 @@ dmu_zfetch_run(zstream_t *zs, boolean_t missed, boolean_t have_lock)
for (int64_t blk = pf_start; blk < pf_end; blk++) { for (int64_t blk = pf_start; blk < pf_end; blk++) {
issued += dbuf_prefetch_impl(zf->zf_dnode, 0, blk, issued += dbuf_prefetch_impl(zf->zf_dnode, 0, blk,
ZIO_PRIORITY_ASYNC_READ, ARC_FLAG_PREDICTIVE_PREFETCH, ZIO_PRIORITY_ASYNC_READ, ARC_FLAG_PREDICTIVE_PREFETCH,
dmu_zfetch_stream_done, zs); dmu_zfetch_done, zs);
} }
for (int64_t iblk = ipf_start; iblk < ipf_end; iblk++) { for (int64_t iblk = ipf_start; iblk < ipf_end; iblk++) {
issued += dbuf_prefetch_impl(zf->zf_dnode, 1, iblk, issued += dbuf_prefetch_impl(zf->zf_dnode, 1, iblk,
ZIO_PRIORITY_ASYNC_READ, ARC_FLAG_PREDICTIVE_PREFETCH, ZIO_PRIORITY_ASYNC_READ, ARC_FLAG_PREDICTIVE_PREFETCH,
dmu_zfetch_stream_done, zs); dmu_zfetch_done, zs);
} }
if (!have_lock) if (!have_lock)
@ -540,6 +553,12 @@ ZFS_MODULE_PARAM(zfs_prefetch, zfetch_, max_streams, UINT, ZMOD_RW,
ZFS_MODULE_PARAM(zfs_prefetch, zfetch_, min_sec_reap, UINT, ZMOD_RW, ZFS_MODULE_PARAM(zfs_prefetch, zfetch_, min_sec_reap, UINT, ZMOD_RW,
"Min time before stream reclaim"); "Min time before stream reclaim");
ZFS_MODULE_PARAM(zfs_prefetch, zfetch_, max_sec_reap, UINT, ZMOD_RW,
"Max time before stream delete");
ZFS_MODULE_PARAM(zfs_prefetch, zfetch_, min_distance, UINT, ZMOD_RW,
"Min bytes to prefetch per stream");
ZFS_MODULE_PARAM(zfs_prefetch, zfetch_, max_distance, UINT, ZMOD_RW, ZFS_MODULE_PARAM(zfs_prefetch, zfetch_, max_distance, UINT, ZMOD_RW,
"Max bytes to prefetch per stream"); "Max bytes to prefetch per stream");