diff --git a/module/zfs/arc.c b/module/zfs/arc.c index 5d5bcbe2b..16d970672 100644 --- a/module/zfs/arc.c +++ b/module/zfs/arc.c @@ -5789,8 +5789,8 @@ l2arc_write_buffers(spa_t *spa, l2arc_dev_t *dev, uint64_t target_sz, boolean_t *headroom_boost) { arc_buf_hdr_t *hdr, *hdr_prev, *head; - uint64_t write_asize, write_psize, write_sz, headroom, - buf_compress_minsz; + uint64_t write_asize, write_sz, headroom, buf_compress_minsz, + stats_size; void *buf_data; boolean_t full; l2arc_write_callback_t *cb; @@ -5805,7 +5805,7 @@ l2arc_write_buffers(spa_t *spa, l2arc_dev_t *dev, uint64_t target_sz, *headroom_boost = B_FALSE; pio = NULL; - write_sz = write_asize = write_psize = 0; + write_sz = write_asize = 0; full = B_FALSE; head = kmem_cache_alloc(hdr_l2only_cache, KM_PUSHPAGE); head->b_flags |= ARC_FLAG_L2_WRITE_HEAD; @@ -5842,6 +5842,7 @@ l2arc_write_buffers(spa_t *spa, l2arc_dev_t *dev, uint64_t target_sz, for (; hdr; hdr = hdr_prev) { kmutex_t *hash_lock; uint64_t buf_sz; + uint64_t buf_a_sz; if (arc_warm == B_FALSE) hdr_prev = multilist_sublist_next(mls, hdr); @@ -5870,7 +5871,15 @@ l2arc_write_buffers(spa_t *spa, l2arc_dev_t *dev, uint64_t target_sz, continue; } - if ((write_sz + hdr->b_size) > target_sz) { + /* + * Assume that the buffer is not going to be compressed + * and could take more space on disk because of a larger + * disk block size. + */ + buf_sz = hdr->b_size; + buf_a_sz = vdev_psize_to_asize(dev->l2ad_vdev, buf_sz); + + if ((write_asize + buf_a_sz) > target_sz) { full = B_TRUE; mutex_exit(hash_lock); break; @@ -5935,8 +5944,6 @@ l2arc_write_buffers(spa_t *spa, l2arc_dev_t *dev, uint64_t target_sz, * using it to denote the header's state change. */ hdr->b_l2hdr.b_daddr = L2ARC_ADDR_UNSET; - - buf_sz = hdr->b_size; hdr->b_flags |= ARC_FLAG_HAS_L2HDR; mutex_enter(&dev->l2ad_mtx); @@ -5953,6 +5960,7 @@ l2arc_write_buffers(spa_t *spa, l2arc_dev_t *dev, uint64_t target_sz, mutex_exit(hash_lock); write_sz += buf_sz; + write_asize += buf_a_sz; } multilist_sublist_unlock(mls); @@ -5971,6 +5979,19 @@ l2arc_write_buffers(spa_t *spa, l2arc_dev_t *dev, uint64_t target_sz, mutex_enter(&dev->l2ad_mtx); + /* + * Note that elsewhere in this file arcstat_l2_asize + * and the used space on l2ad_vdev are updated using b_asize, + * which is not necessarily rounded up to the device block size. + * Too keep accounting consistent we do the same here as well: + * stats_size accumulates the sum of b_asize of the written buffers, + * while write_asize accumulates the sum of b_asize rounded up + * to the device block size. + * The latter sum is used only to validate the corectness of the code. + */ + stats_size = 0; + write_asize = 0; + /* * Now start writing the buffers. We're starting at the write head * and work backwards, retracing the course of the buffer selector @@ -6024,7 +6045,7 @@ l2arc_write_buffers(spa_t *spa, l2arc_dev_t *dev, uint64_t target_sz, /* Compression may have squashed the buffer to zero length. */ if (buf_sz != 0) { - uint64_t buf_p_sz; + uint64_t buf_a_sz; wzio = zio_write_phys(pio, dev->l2ad_vdev, dev->l2ad_hand, buf_sz, buf_data, ZIO_CHECKSUM_OFF, @@ -6035,14 +6056,14 @@ l2arc_write_buffers(spa_t *spa, l2arc_dev_t *dev, uint64_t target_sz, zio_t *, wzio); (void) zio_nowait(wzio); - write_asize += buf_sz; + stats_size += buf_sz; /* * Keep the clock hand suitably device-aligned. */ - buf_p_sz = vdev_psize_to_asize(dev->l2ad_vdev, buf_sz); - write_psize += buf_p_sz; - dev->l2ad_hand += buf_p_sz; + buf_a_sz = vdev_psize_to_asize(dev->l2ad_vdev, buf_sz); + write_asize += buf_a_sz; + dev->l2ad_hand += buf_a_sz; } } @@ -6052,8 +6073,8 @@ l2arc_write_buffers(spa_t *spa, l2arc_dev_t *dev, uint64_t target_sz, ARCSTAT_BUMP(arcstat_l2_writes_sent); ARCSTAT_INCR(arcstat_l2_write_bytes, write_asize); ARCSTAT_INCR(arcstat_l2_size, write_sz); - ARCSTAT_INCR(arcstat_l2_asize, write_asize); - vdev_space_update(dev->l2ad_vdev, write_asize, 0, 0); + ARCSTAT_INCR(arcstat_l2_asize, stats_size); + vdev_space_update(dev->l2ad_vdev, stats_size, 0, 0); /* * Bump device hand to the device start if it is approaching the end.