Skip to contentSkip to conversation

Btrfs: snapshot-aware defragmentation

RFC PATCH New Li Zefan 1 participant 1 message Aug 19, 2011 Btrfs file system

lore
Share & tools
lore

Li Zefan

author
#0 lore raw

RFC PATCH 2/2 Btrfs: snapshot-aware defragmentation

Note this patch is not complete. I send it out early to make sure this is the right way to implement this feature.

When we defrag a file, btrfs will read the file data into memory and mark the memory dirty. When writeback kicks off, file data will be re-written to disk. This is how defragmentation works.

To make defrag snapshot-aware:

- when mark the memory dirty, we also add a defrag flag
- when writeback starts, we check the defrag flag
- If the flag is set, we find out all the backrefs for the extent that is to be re-written
- After writeback, we change the backrefs to point to the new extent

This patch makes use of extent backref iteration functions provided by Jan Schmidt.

---
 fs/btrfs/extent_io.c        |   16 ++-
 fs/btrfs/extent_io.h        |    2 +
 fs/btrfs/file.c             |    4 +-
 fs/btrfs/free-space-cache.c |    7 +-
 fs/btrfs/inode.c            |  400 ++++++++++++++++++++++++++++++++++++++++++-
 fs/btrfs/ioctl.c            |   13 +-
 6 files changed, 419 insertions(+), 23 deletions(-)
fs/btrfs/extent_io.c +13 -3
index d418164..168f8f8 100644
@@ -928,7 +928,17 @@ int clear_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end,
 {
 	return clear_extent_bit(tree, start, end,
 				EXTENT_DIRTY | EXTENT_DELALLOC |
-				EXTENT_DO_ACCOUNTING, 0, 0, NULL, mask);
+				EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG,
+				0, 0, NULL, mask);
+}
+
+int set_extent_defrag(struct extent_io_tree *tree, u64 start, u64 end,
+		      struct extent_state **cached_state, gfp_t mask)
+{
+	return set_extent_bit(tree, start, end,
+			      EXTENT_DELALLOC | EXTENT_DIRTY |
+			      EXTENT_UPTODATE | EXTENT_DEFRAG,
+			      0, NULL, cached_state, mask);
 }
 
 int set_extent_new(struct extent_io_tree *tree, u64 start, u64 end,
@@ -2628,7 +2638,7 @@ int extent_invalidatepage(struct extent_io_tree *tree,
 	wait_on_page_writeback(page);
 	clear_extent_bit(tree, start, end,
 			 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC |
-			 EXTENT_DO_ACCOUNTING,
+			 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG,
 			 1, 1, &cached_state, GFP_NOFS);
 	return 0;
 }
@@ -2703,7 +2713,7 @@ int try_release_extent_mapping(struct extent_map_tree *map,
 			}
 			if (!test_range_bit(tree, em->start,
 					    extent_map_end(em) - 1,
-					    EXTENT_LOCKED | EXTENT_WRITEBACK,
+					    EXTENT_LOCKED,
 					    0, NULL)) {
 				remove_extent_mapping(map, em);
 				/* once for the rb tree */
fs/btrfs/extent_io.h +2 -0
index 7b2f0c3..889b675 100644
@@ -216,6 +216,8 @@ int clear_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end,
 		       gfp_t mask);
 int set_extent_delalloc(struct extent_io_tree *tree, u64 start, u64 end,
 			struct extent_state **cached_state, gfp_t mask);
+int set_extent_defrag(struct extent_io_tree *tree, u64 start, u64 end,
+			struct extent_state **cached_state, gfp_t mask);
 int find_first_extent_bit(struct extent_io_tree *tree, u64 start,
 			  u64 *start_ret, u64 *end_ret, int bits);
 struct extent_state *find_first_extent_bit_state(struct extent_io_tree *tree,
fs/btrfs/file.c +2 -2
index 658d669..e1cb0ba 100644
@@ -1129,8 +1129,8 @@ again:
 
 		clear_extent_bit(&BTRFS_I(inode)->io_tree, start_pos,
 				  last_pos - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
-				  EXTENT_DO_ACCOUNTING, 0, 0, &cached_state,
-				  GFP_NOFS);
+				  EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG,
+				  0, 0, &cached_state, GFP_NOFS);
 		unlock_extent_cached(&BTRFS_I(inode)->io_tree,
 				     start_pos, last_pos - 1, &cached_state,
 				     GFP_NOFS);
fs/btrfs/free-space-cache.c +4 -3
index 6377713..e171b45 100644
@@ -801,7 +801,8 @@ int __btrfs_write_out_cache(struct btrfs_root *root, struct inode *inode,
 		ret = -1;
 		clear_extent_bit(&BTRFS_I(inode)->io_tree, 0, bytes - 1,
 				 EXTENT_DIRTY | EXTENT_DELALLOC |
-				 EXTENT_DO_ACCOUNTING, 0, 0, NULL, GFP_NOFS);
+				 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG,
+				 0, 0, NULL, GFP_NOFS);
 		goto out;
 	}
 	leaf = path->nodes[0];
@@ -815,8 +816,8 @@ int __btrfs_write_out_cache(struct btrfs_root *root, struct inode *inode,
 			ret = -1;
 			clear_extent_bit(&BTRFS_I(inode)->io_tree, 0, bytes - 1,
 					 EXTENT_DIRTY | EXTENT_DELALLOC |
-					 EXTENT_DO_ACCOUNTING, 0, 0, NULL,
-					 GFP_NOFS);
+					 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG,
+					 0, 0, NULL, GFP_NOFS);
 			btrfs_release_path(path);
 			goto out;
 		}
fs/btrfs/inode.c +392 -8

472 lines not on this page · raw message

fs/btrfs/ioctl.c +6 -7
index 7cf0133..347fce8a 100644
@@ -924,10 +924,10 @@ again:
 	if (ordered)
 		btrfs_put_ordered_extent(ordered);
 
-	clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
-			  page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
-			  EXTENT_DO_ACCOUNTING, 0, 0, &cached_state,
-			  GFP_NOFS);
+	clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
+			  EXTENT_DIRTY | EXTENT_DELALLOC |
+			  EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
+			  &cached_state, GFP_NOFS);
 
 	if (i_done != num_pages) {
 		spin_lock(&BTRFS_I(inode)->lock);
@@ -937,9 +937,8 @@ again:
 				     (num_pages - i_done) << PAGE_CACHE_SHIFT);
 	}
 
-
-	btrfs_set_extent_delalloc(inode, page_start, page_end - 1,
-				  &cached_state);
+	set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
+			  &cached_state, GFP_NOFS);
 
 	unlock_extent_cached(&BTRFS_I(inode)->io_tree,
 			     page_start, page_end - 1, &cached_state,
Signature · 1 line
1.7.3.1