diff options
Diffstat (limited to 'usr/src/uts/common/fs/zfs/arc.c')
| -rw-r--r-- | usr/src/uts/common/fs/zfs/arc.c | 48 |
1 files changed, 33 insertions, 15 deletions
diff --git a/usr/src/uts/common/fs/zfs/arc.c b/usr/src/uts/common/fs/zfs/arc.c index 5a899aa1c7..af70eee950 100644 --- a/usr/src/uts/common/fs/zfs/arc.c +++ b/usr/src/uts/common/fs/zfs/arc.c @@ -380,6 +380,13 @@ int zfs_arc_shrink_shift = 0; int zfs_arc_p_min_shift = 0; int zfs_arc_average_blocksize = 8 * 1024; /* 8KB */ +/* + * ARC dirty data constraints for arc_tempreserve_space() throttle + */ +uint_t zfs_arc_dirty_limit_percent = 50; /* total dirty data limit */ +uint_t zfs_arc_anon_limit_percent = 25; /* anon block dirty limit */ +uint_t zfs_arc_pool_dirty_percent = 20; /* each pool's anon allowance */ + boolean_t zfs_compressed_arc_enabled = B_TRUE; /* @@ -5848,12 +5855,10 @@ arc_write(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp, arc_buf_t *buf, } static int -arc_memory_throttle(uint64_t reserve, uint64_t txg) +arc_memory_throttle(spa_t *spa, uint64_t reserve, uint64_t txg) { #ifdef _KERNEL uint64_t available_memory = ptob(freemem); - static uint64_t page_load = 0; - static uint64_t last_txg = 0; #if defined(__i386) available_memory = @@ -5863,9 +5868,9 @@ arc_memory_throttle(uint64_t reserve, uint64_t txg) if (freemem > physmem * arc_lotsfree_percent / 100) return (0); - if (txg > last_txg) { - last_txg = txg; - page_load = 0; + if (txg > spa->spa_lowmem_last_txg) { + spa->spa_lowmem_last_txg = txg; + spa->spa_lowmem_page_load = 0; } /* * If we are in pageout, we know that memory is already tight, @@ -5873,18 +5878,19 @@ arc_memory_throttle(uint64_t reserve, uint64_t txg) * continue to let page writes occur as quickly as possible. */ if (curproc == proc_pageout) { - if (page_load > MAX(ptob(minfree), available_memory) / 4) + if (spa->spa_lowmem_page_load > + MAX(ptob(minfree), available_memory) / 4) return (SET_ERROR(ERESTART)); /* Note: reserve is inflated, so we deflate */ - page_load += reserve / 8; + atomic_add_64(&spa->spa_lowmem_page_load, reserve / 8); return (0); - } else if (page_load > 0 && arc_reclaim_needed()) { + } else if (spa->spa_lowmem_page_load > 0 && arc_reclaim_needed()) { /* memory is low, delay before restarting */ ARCSTAT_INCR(arcstat_memory_throttle_count, 1); return (SET_ERROR(EAGAIN)); } - page_load = 0; -#endif + spa->spa_lowmem_page_load = 0; +#endif /* _KERNEL */ return (0); } @@ -5896,7 +5902,7 @@ arc_tempreserve_clear(uint64_t reserve) } int -arc_tempreserve_space(uint64_t reserve, uint64_t txg) +arc_tempreserve_space(spa_t *spa, uint64_t reserve, uint64_t txg) { int error; uint64_t anon_size; @@ -5923,7 +5929,7 @@ arc_tempreserve_space(uint64_t reserve, uint64_t txg) * in order to compress/encrypt/etc the data. We therefore need to * make sure that there is sufficient available memory for this. */ - error = arc_memory_throttle(reserve, txg); + error = arc_memory_throttle(spa, reserve, txg); if (error != 0) return (error); @@ -5931,12 +5937,24 @@ arc_tempreserve_space(uint64_t reserve, uint64_t txg) * Throttle writes when the amount of dirty data in the cache * gets too large. We try to keep the cache less than half full * of dirty blocks so that our sync times don't grow too large. + * + * In the case of one pool being built on another pool, we want + * to make sure we don't end up throttling the lower (backing) + * pool when the upper pool is the majority contributor to dirty + * data. To insure we make forward progress during throttling, we + * also check the current pool's net dirty data and only throttle + * if it exceeds zfs_arc_pool_dirty_percent of the anonymous dirty + * data in the cache. + * * Note: if two requests come in concurrently, we might let them * both succeed, when one of them should fail. Not a huge deal. */ + uint64_t total_dirty = reserve + arc_tempreserve + anon_size; + uint64_t spa_dirty_anon = spa_dirty_data(spa); - if (reserve + arc_tempreserve + anon_size > arc_c / 2 && - anon_size > arc_c / 4) { + if (total_dirty > arc_c * zfs_arc_dirty_limit_percent / 100 && + anon_size > arc_c * zfs_arc_anon_limit_percent / 100 && + spa_dirty_anon > anon_size * zfs_arc_pool_dirty_percent / 100) { uint64_t meta_esize = refcount_count(&arc_anon->arcs_esize[ARC_BUFC_METADATA]); uint64_t data_esize = |