[PATCH] Btrfs: improve the loop of scrub_stripe

[PATCH] Btrfs: improve the loop of scrub_stripe

[Liu Bo]

1) Right now scrub_stripe() is looping in some unnecessary cases:
* when the found extent item's objectid has been out of the dev extent's range
  but we haven't finish scanning all the range within the dev extent
* when all the items has been processed but we haven't finish scanning all the
  range within the dev extent

In both cases, we can just finish the loop to save costs.

2) Besides, when the found extent item's length is larger than the stripe len(64k),
we don't have to release the path and search again as it'll get at the same key
used in the last loop, we can instead increase the logical cursor in place till
all space of the extent is scanned.

3) And we use 0 as the key's offset to search btree, then get to previous item
to find a smaller item, and again have to move to the next one to get the right
item.  Setting offset=-1 and previous_item() is the correct way.

4) As we won't find any checksum at offset unless this 'offset' is in a data
extent, we can just find checksum when we're really going to scrub an extent.

Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
---
 fs/btrfs/scrub.c |   79 +++++++++++++++++++++++++++++++++++++-----------------
 1 files changed, 54 insertions(+), 25 deletions(-)

diff --git a/fs/btrfs/scrub.c b/fs/btrfs/scrub.c
index 85e072b..091b57b 100644
--- a/fs/btrfs/scrub.c
+++ b/fs/btrfs/scrub.c
@@ -2236,12 +2236,12 @@ static noinline_for_stack int scrub_stripe(struct scrub_ctx *sctx,
 	u64 flags;
 	int ret;
 	int slot;
-	int i;
 	u64 nstripes;
 	struct extent_buffer *l;
 	struct btrfs_key key;
 	u64 physical;
 	u64 logical;
+	u64 logic_end;
 	u64 generation;
 	int mirror_num;
 	struct reada_control *reada1;
@@ -2255,6 +2255,7 @@ static noinline_for_stack int scrub_stripe(struct scrub_ctx *sctx,
 	u64 extent_len;
 	struct btrfs_device *extent_dev;
 	int extent_mirror_num;
+	int stop_loop;
 
 	if (map->type & (BTRFS_BLOCK_GROUP_RAID5 |
 			 BTRFS_BLOCK_GROUP_RAID6)) {
@@ -2354,8 +2355,9 @@ static noinline_for_stack int scrub_stripe(struct scrub_ctx *sctx,
 	 */
 	logical = base + offset;
 	physical = map->stripes[num].physical;
+	logic_end = logical + increment * nstripes;
 	ret = 0;
-	for (i = 0; i < nstripes; ++i) {
+	while (logical < logic_end) {
 		/*
 		 * canceled?
 		 */
@@ -2391,15 +2393,9 @@ static noinline_for_stack int scrub_stripe(struct scrub_ctx *sctx,
 			wake_up(&fs_info->scrub_pause_wait);
 		}
 
-		ret = btrfs_lookup_csums_range(csum_root, logical,
-					       logical + map->stripe_len - 1,
-					       &sctx->csum_list, 1);
-		if (ret)
-			goto out;
-
 		key.objectid = logical;
 		key.type = BTRFS_EXTENT_ITEM_KEY;
-		key.offset = (u64)0;
+		key.offset = (u64)-1;
 
 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
 		if (ret < 0)
@@ -2420,6 +2416,7 @@ static noinline_for_stack int scrub_stripe(struct scrub_ctx *sctx,
 			}
 		}
 
+		stop_loop = 0;
 		while (1) {
 			l = path->nodes[0];
 			slot = path->slots[0];
@@ -2430,18 +2427,21 @@ static noinline_for_stack int scrub_stripe(struct scrub_ctx *sctx,
 				if (ret < 0)
 					goto out;
 
+				stop_loop = 1;
 				break;
 			}
 			btrfs_item_key_to_cpu(l, &key, slot);
 
-			if (key.objectid + key.offset <= logical)
+			if (key.objectid + key.offset <= logical ||
+			    btrfs_key_type(&key) != BTRFS_EXTENT_ITEM_KEY)
 				goto next;
 
-			if (key.objectid >= logical + map->stripe_len)
+			if (key.objectid >= logical + map->stripe_len) {
+				/* out of this device extent */
+				if (key.objectid >= logic_end)
+					stop_loop = 1;
 				break;
-
-			if (btrfs_key_type(&key) != BTRFS_EXTENT_ITEM_KEY)
-				goto next;
+			}
 
 			extent = btrfs_item_ptr(l, slot,
 						struct btrfs_extent_item);
@@ -2458,22 +2458,23 @@ static noinline_for_stack int scrub_stripe(struct scrub_ctx *sctx,
 				goto next;
 			}
 
+again:
+			extent_logical = key.objectid;
+			extent_len = key.offset;
 			/*
 			 * trim extent to this stripe
 			 */
-			if (key.objectid < logical) {
-				key.offset -= logical - key.objectid;
-				key.objectid = logical;
+			if (extent_logical < logical) {
+				extent_len -= logical - extent_logical;
+				extent_logical = logical;
 			}
-			if (key.objectid + key.offset >
+			if (extent_logical + extent_len >
 			    logical + map->stripe_len) {
-				key.offset = logical + map->stripe_len -
-					     key.objectid;
+				extent_len = logical + map->stripe_len -
+					     extent_logical;
 			}
 
-			extent_logical = key.objectid;
-			extent_physical = key.objectid - logical + physical;
-			extent_len = key.offset;
+			extent_physical = extent_logical - logical + physical;
 			extent_dev = scrub_dev;
 			extent_mirror_num = mirror_num;
 			if (is_dev_replace)
@@ -2481,13 +2482,35 @@ static noinline_for_stack int scrub_stripe(struct scrub_ctx *sctx,
 						   extent_len, &extent_physical,
 						   &extent_dev,
 						   &extent_mirror_num);
+
+			ret = btrfs_lookup_csums_range(csum_root, logical,
+						logical + map->stripe_len - 1,
+						&sctx->csum_list, 1);
+			if (ret)
+				goto out;
+
 			ret = scrub_extent(sctx, extent_logical, extent_len,
 					   extent_physical, extent_dev, flags,
 					   generation, extent_mirror_num,
-					   key.objectid - logical + physical);
+					   extent_physical);
 			if (ret)
 				goto out;
 
+			if (extent_logical + extent_len <
+			    key.objectid + key.offset) {
+				logical += increment;
+				physical += map->stripe_len;
+
+				if (logical < key.objectid + key.offset) {
+					cond_resched();
+					goto again;
+				}
+
+				if (logical >= logic_end) {
+					stop_loop = 1;
+					break;
+				}
+			}
 next:
 			path->slots[0]++;
 		}
@@ -2495,8 +2518,14 @@ next:
 		logical += increment;
 		physical += map->stripe_len;
 		spin_lock(&sctx->stat_lock);
-		sctx->stat.last_physical = physical;
+		if (stop_loop)
+			sctx->stat.last_physical = map->stripes[num].physical +
+						   length;
+		else
+			sctx->stat.last_physical = physical;
 		spin_unlock(&sctx->stat_lock);
+		if (stop_loop)
+			break;
 	}
 out:
 	/* push queued extents */
-- 
1.7.7

Re: [PATCH] Btrfs: improve the loop of scrub_stripe

[Josef Bacik]

On Thu, Apr 18, 2013 at 10:54:55AM -0600, Liu Bo wrote:
> 1) Right now scrub_stripe() is looping in some unnecessary cases:
> * when the found extent item's objectid has been out of the dev extent's range
>   but we haven't finish scanning all the range within the dev extent
> * when all the items has been processed but we haven't finish scanning all the
>   range within the dev extent
> 
> In both cases, we can just finish the loop to save costs.
> 
> 2) Besides, when the found extent item's length is larger than the stripe len(64k),
> we don't have to release the path and search again as it'll get at the same key
> used in the last loop, we can instead increase the logical cursor in place till
> all space of the extent is scanned.
> 
> 3) And we use 0 as the key's offset to search btree, then get to previous item
> to find a smaller item, and again have to move to the next one to get the right
> item.  Setting offset=-1 and previous_item() is the correct way.
> 
> 4) As we won't find any checksum at offset unless this 'offset' is in a data
> extent, we can just find checksum when we're really going to scrub an extent.
> 
> Signed-off-by: Liu Bo <bo.li.liu@oracle.com>

A big chunk of this doesn't apply to btrfs-next and I don't feel comfortable
manually applying it, please rebase and resend.  Thanks,

Josef