dm writecache: don't split bios when overwriting contiguous cache content

If dm-writecache overwrites existing cached data, it splits the
incoming bio into many block-sized bios. The I/O scheduler does merge
these bios into one large request but this needless splitting and
merging causes performance degradation.

Fix this by avoiding bio splitting if the cache target area that is
being overwritten is contiguous.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>

diff --git a/drivers/md/dm-writecache.c b/drivers/md/dm-writecache.c
index aecc246..a440072 100644 (file)
--- a/drivers/md/dm-writecache.c
+++ b/drivers/md/dm-writecache.c
@@ -1360,14 +1360,18 @@ read_next_block:
        } else {
                do {
                        bool found_entry = false;
+                       bool search_used = false;
                        if (writecache_has_error(wc))
                                goto unlock_error;
                        e = writecache_find_entry(wc, bio->bi_iter.bi_sector, 0);
                        if (e) {
-                               if (!writecache_entry_is_committed(wc, e))
+                               if (!writecache_entry_is_committed(wc, e)) {
+                                       search_used = true;
                                        goto bio_copy;
+                               }
                                if (!WC_MODE_PMEM(wc) && !e->write_in_progress) {
                                        wc->overwrote_committed = true;
+                                       search_used = true;
                                        goto bio_copy;
                                }
                                found_entry = true;
@@ -1404,13 +1408,31 @@ bio_copy:
                                sector_t current_cache_sec = start_cache_sec + (bio_size >> SECTOR_SHIFT);
 
                                while (bio_size < bio->bi_iter.bi_size) {
-                                       struct wc_entry *f = writecache_pop_from_freelist(wc, current_cache_sec);
-                                       if (!f)
-                                               break;
-                                       write_original_sector_seq_count(wc, f, bio->bi_iter.bi_sector +
-                                                                       (bio_size >> SECTOR_SHIFT), wc->seq_count);
-                                       writecache_insert_entry(wc, f);
-                                       wc->uncommitted_blocks++;
+                                       if (!search_used) {
+                                               struct wc_entry *f = writecache_pop_from_freelist(wc, current_cache_sec);
+                                               if (!f)
+                                                       break;
+                                               write_original_sector_seq_count(wc, f, bio->bi_iter.bi_sector +
+                                                                               (bio_size >> SECTOR_SHIFT), wc->seq_count);
+                                               writecache_insert_entry(wc, f);
+                                               wc->uncommitted_blocks++;
+                                       } else {
+                                               struct wc_entry *f;
+                                               struct rb_node *next = rb_next(&e->rb_node);
+                                               if (!next)
+                                                       break;
+                                               f = container_of(next, struct wc_entry, rb_node);
+                                               if (f != e + 1)
+                                                       break;
+                                               if (read_original_sector(wc, f) !=
+                                                   read_original_sector(wc, e) + (wc->block_size >> SECTOR_SHIFT))
+                                                       break;
+                                               if (unlikely(f->write_in_progress))
+                                                       break;
+                                               if (writecache_entry_is_committed(wc, f))
+                                                       wc->overwrote_committed = true;
+                                               e = f;
+                                       }
                                        bio_size += wc->block_size;
                                        current_cache_sec += wc->block_size >> SECTOR_SHIFT;
                                }