MFC r286551: 5694 traverse_prefetcher does not prefetch enough

Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: Alex Reece <alex@delphix.com>
Reviewed by: Christopher Siden <christopher.siden@delphix.com>
Reviewed by: Josef 'Jeff' Sipek <josef.sipek@nexenta.com>
Reviewed by: Bayard Bell <buffer.g.overflow@gmail.com>
Approved by: Garrett D'Amore <garrett@damore.org>
Author: George Wilson <george.wilson@delphix.com>

illumos/illumos-gate@34d7ce052c4565b078f73b95ccbd49274e98edaa

1 files changed, 9 insertions, 10 deletions

diff --git a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dmu_traverse.c b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dmu_traverse.c
index 9891099..f716442 100644
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dmu_traverse.c
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dmu_traverse.c
@@ -39,13 +39,12 @@
 #include <sys/callb.h>
 #include <sys/zfeature.h>
 
-int zfs_pd_blks_max = 100;
+int32_t zfs_pd_bytes_max = 50 * 1024 * 1024;	/* 50MB */
 
 typedef struct prefetch_data {
 	kmutex_t pd_mtx;
 	kcondvar_t pd_cv;
-	int pd_blks_max;
-	int pd_blks_fetched;
+	int32_t pd_bytes_fetched;
 	int pd_flags;
 	boolean_t pd_cancel;
 	boolean_t pd_exited;
@@ -250,11 +249,12 @@ traverse_visitbp(traverse_data_t *td, const dnode_phys_t *dnp,
 	}
 
 	if (pd != NULL && !pd->pd_exited && prefetch_needed(pd, bp)) {
+		uint64_t size = BP_GET_LSIZE(bp);
 		mutex_enter(&pd->pd_mtx);
-		ASSERT(pd->pd_blks_fetched >= 0);
-		while (pd->pd_blks_fetched == 0 && !pd->pd_exited)
+		ASSERT(pd->pd_bytes_fetched >= 0);
+		while (pd->pd_bytes_fetched < size && !pd->pd_exited)
 			cv_wait(&pd->pd_cv, &pd->pd_mtx);
-		pd->pd_blks_fetched--;
+		pd->pd_bytes_fetched -= size;
 		cv_broadcast(&pd->pd_cv);
 		mutex_exit(&pd->pd_mtx);
 	}
@@ -447,7 +447,7 @@ traverse_prefetcher(spa_t *spa, zilog_t *zilog, const blkptr_t *bp,
 	prefetch_data_t *pfd = arg;
 	arc_flags_t aflags = ARC_FLAG_NOWAIT | ARC_FLAG_PREFETCH;
 
-	ASSERT(pfd->pd_blks_fetched >= 0);
+	ASSERT(pfd->pd_bytes_fetched >= 0);
 	if (pfd->pd_cancel)
 		return (SET_ERROR(EINTR));
 
@@ -455,9 +455,9 @@ traverse_prefetcher(spa_t *spa, zilog_t *zilog, const blkptr_t *bp,
 		return (0);
 
 	mutex_enter(&pfd->pd_mtx);
-	while (!pfd->pd_cancel && pfd->pd_blks_fetched >= pfd->pd_blks_max)
+	while (!pfd->pd_cancel && pfd->pd_bytes_fetched >= zfs_pd_bytes_max)
 		cv_wait(&pfd->pd_cv, &pfd->pd_mtx);
-	pfd->pd_blks_fetched++;
+	pfd->pd_bytes_fetched += BP_GET_LSIZE(bp);
 	cv_broadcast(&pfd->pd_cv);
 	mutex_exit(&pfd->pd_mtx);
 
@@ -529,7 +529,6 @@ traverse_impl(spa_t *spa, dsl_dataset_t *ds, uint64_t objset, blkptr_t *rootbp,
 		td.td_hole_birth_enabled_txg = 0;
 	}
 
-	pd.pd_blks_max = zfs_pd_blks_max;
 	pd.pd_flags = flags;
 	mutex_init(&pd.pd_mtx, NULL, MUTEX_DEFAULT, NULL);
 	cv_init(&pd.pd_cv, NULL, CV_DEFAULT, NULL);