Btrfs: cache the extent state everywhere we possibly can
PATCH 1/1 Btrfs: cache the extent state everywhere we possibly can
This patch just goes through and fixes everybody that does
lock_extent()
blahunlock_extent()
to use
lock_extent_bits()
blahunlock_extent_cached()
and pass around a extent_state so we only have to do the searches once per function. This gives me about a 3 mb/s boots on my random write test. I have not converted some things, like the relocation and ioctl's, since they aren't heavily used and the relocation stuff is in the middle of being re-written. I also changed the clear_extent_bit() to only unset the cached state if we are clearing EXTENT_LOCKED and related stuff, so we can do things like this
lock_extent_bits()
clear delalloc bitsunlock_extent_cached()
without losing our cached state. I tested this thoroughly and turned on LEAK_DEBUG to make sure we weren't leaking extent states, everything worked out fine.
--- fs/btrfs/ctree.h | 3 +- fs/btrfs/disk-io.c | 8 ++- fs/btrfs/extent-tree.c | 11 +++-- fs/btrfs/extent_io.c | 44 ++++++++++++------- fs/btrfs/extent_io.h | 4 +- fs/btrfs/file.c | 23 ++++++---- fs/btrfs/inode.c | 111 ++++++++++++++++++++++++++++++------------------ fs/btrfs/ioctl.c | 2 +- fs/btrfs/relocation.c | 2 +- 9 files changed, 130 insertions(+), 78 deletions(-)
fs/btrfs/ctree.h +2 -1
@@ -2309,7 +2309,8 @@ int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans, u32 min_type);int btrfs_start_delalloc_inodes (struct btrfs_root *root,int delay_iput); -int btrfs_set_extent_delalloc (struct inode *inode,u64 start,u64 end); +int btrfs_set_extent_delalloc (struct inode *inode,u64 start,u64 end, +struct extent_state **cached_state);int btrfs_writepages (struct address_space *mapping,struct writeback_control *wbc); int btrfs_create_subvol_root(struct btrfs_trans_handle *trans,
fs/btrfs/disk-io.c +5 -3
@@ -263,12 +263,14 @@ static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,static int verify_parent_transid (struct extent_io_tree *io_tree,struct extent_buffer *eb,u64 parent_transid) { +struct extent_state *cached_state =NULL ;int ret;if (!parent_transid ||btrfs_header_generation (eb) == parent_transid)return 0 ; -lock_extent (io_tree, eb->start, eb->start + eb->len -1 ,GFP_NOFS ); +lock_extent_bits (io_tree, eb->start, eb->start + eb->len -1 , +0 , &cached_state,GFP_NOFS );if (extent_buffer_uptodate (io_tree, eb) &&btrfs_header_generation (eb) == parent_transid) { ret =0 ;
@@ -284,8 +286,8 @@ static int verify_parent_transid(struct extent_io_tree *io_tree, ret =1 ;clear_extent_buffer_uptodate (io_tree, eb); out: -unlock_extent (io_tree, eb->start, eb->start + eb->len -1 , -GFP_NOFS ); +unlock_extent_cached (io_tree, eb->start, eb->start + eb->len -1 , + &cached_state,GFP_NOFS );return ret; }
fs/btrfs/extent-tree.c +7 -4
@@ -6558,6 +6558,7 @@ static noinline int invalidate_extent_cache(struct btrfs_root *root,struct btrfs_key key;struct inode *inode =NULL ;struct btrfs_file_extent_item *fi; +struct extent_state *cached_state =NULL ;u64 num_bytes;u64 skip_objectid =0 ;u32 nritems;
@@ -6586,12 +6587,14 @@ static noinline int invalidate_extent_cache(struct btrfs_root *root, } num_bytes =btrfs_file_extent_num_bytes (leaf, fi); -lock_extent (&BTRFS_I (inode)->io_tree, key.offset, - key.offset + num_bytes -1 ,GFP_NOFS ); +lock_extent_bits (&BTRFS_I (inode)->io_tree, key.offset, + key.offset + num_bytes -1 ,0 , &cached_state, +GFP_NOFS );btrfs_drop_extent_cache (inode, key.offset, key.offset + num_bytes -1 ,1 ); -unlock_extent (&BTRFS_I (inode)->io_tree, key.offset, - key.offset + num_bytes -1 ,GFP_NOFS ); +unlock_extent_cached (&BTRFS_I (inode)->io_tree, key.offset, + key.offset + num_bytes -1 , &cached_state, +GFP_NOFS );cond_resched (); }iput (inode);
fs/btrfs/extent_io.c +28 -16
116 lines not on this page · raw message
fs/btrfs/extent_io.h +3 -1
@@ -164,6 +164,8 @@ int lock_extent(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask);int lock_extent_bits (struct extent_io_tree *tree,u64 start,u64 end,int bits,struct extent_state **cached,gfp_t mask);int unlock_extent (struct extent_io_tree *tree,u64 start,u64 end,gfp_t mask); +int unlock_extent_cached (struct extent_io_tree *tree,u64 start,u64 end, +struct extent_state **cached,gfp_t mask);int try_lock_extent (struct extent_io_tree *tree,u64 start,u64 end,gfp_t mask); int extent_read_full_page(struct extent_io_tree *tree, struct page *page,
@@ -197,7 +199,7 @@ int clear_extent_ordered(struct extent_io_tree *tree, u64 start, u64 end,int clear_extent_ordered_metadata (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, -gfp_t mask); +struct extent_state **cached_state,gfp_t mask);int set_extent_ordered (struct extent_io_tree *tree,u64 start,u64 end,gfp_t mask); int find_first_extent_bit(struct extent_io_tree *tree, u64 start,
fs/btrfs/file.c +14 -9
@@ -123,7 +123,8 @@ static noinline int dirty_and_release_pages(struct btrfs_trans_handle *trans, root->sectorsize - 1) & ~((u64)root->sectorsize - 1); end_of_last_block = start_pos + num_bytes -1 ; - err =btrfs_set_extent_delalloc (inode, start_pos, end_of_last_block); + err =btrfs_set_extent_delalloc (inode, start_pos, end_of_last_block, +NULL );if (err)return err;
@@ -751,6 +752,7 @@ static noinline int prepare_pages(struct btrfs_root *root, struct file *file, loff_t pos, unsigned long first_index, unsigned long last_index, size_t write_bytes) { + struct extent_state *cached_state = NULL;int i;unsigned long index = pos >>PAGE_CACHE_SHIFT ;struct inode *inode =fdentry (file)->d_inode;
@@ -779,16 +781,18 @@ again: }if (start_pos < inode->i_size) {struct btrfs_ordered_extent *ordered; -lock_extent (&BTRFS_I (inode)->io_tree, - start_pos, last_pos -1 ,GFP_NOFS ); +lock_extent_bits (&BTRFS_I (inode)->io_tree, + start_pos, last_pos -1 ,0 , &cached_state, +GFP_NOFS ); ordered =btrfs_lookup_first_ordered_extent (inode, last_pos -1 );if (ordered && ordered->file_offset + ordered->len > start_pos && ordered->file_offset < last_pos) {btrfs_put_ordered_extent (ordered); -unlock_extent (&BTRFS_I (inode)->io_tree, - start_pos, last_pos -1 ,GFP_NOFS ); +unlock_extent_cached (&BTRFS_I (inode)->io_tree, + start_pos, last_pos -1 , + &cached_state,GFP_NOFS );for (i =0 ; i < num_pages; i++) {unlock_page (pages[i]);page_cache_release (pages[i]);
@@ -800,12 +804,13 @@ again:if (ordered)btrfs_put_ordered_extent (ordered); -clear_extent_bits (&BTRFS_I (inode)->io_tree, start_pos, +clear_extent_bit (&BTRFS_I (inode)->io_tree, start_pos, last_pos -1 ,EXTENT_DIRTY |EXTENT_DELALLOC | -EXTENT_DO_ACCOUNTING , +EXTENT_DO_ACCOUNTING ,0 ,0 , &cached_state,GFP_NOFS ); -unlock_extent (&BTRFS_I (inode)->io_tree, - start_pos, last_pos -1 ,GFP_NOFS ); +unlock_extent_cached (&BTRFS_I (inode)->io_tree, + start_pos, last_pos -1 , &cached_state, +GFP_NOFS ); }for (i =0 ; i < num_pages; i++) {clear_page_dirty_for_io (pages[i]);
fs/btrfs/inode.c +69 -42
306 lines not on this page · raw message
fs/btrfs/ioctl.c +1 -1
@@ -538,7 +538,7 @@ again:*/ clear_page_dirty_for_io (page); -btrfs_set_extent_delalloc (inode, page_start, page_end); +btrfs_set_extent_delalloc (inode, page_start, page_end,NULL );set_page_dirty (page);unlock_extent (io_tree, page_start, page_end,GFP_NOFS );unlock_page (page);
fs/btrfs/relocation.c +1 -1
@@ -2659,7 +2659,7 @@ static int relocate_file_extent_cluster(struct inode *inode, EXTENT_BOUNDARY, GFP_NOFS); nr++; } -btrfs_set_extent_delalloc (inode, page_start, page_end); +btrfs_set_extent_delalloc (inode, page_start, page_end,NULL );set_page_dirty (page); dirty_page++;
Signature · 1 line
1.6.6
PATCH 1/1 Btrfs: cache the extent state everywhere we possibly can V2
This patch just goes through and fixes everybody that does
lock_extent()
blahunlock_extent()
to use
lock_extent_bits()
blahunlock_extent_cached()
and pass around a extent_state so we only have to do the searches once per function. This gives me about a 3 mb/s boots on my random write test. I have not converted some things, like the relocation and ioctl's, since they aren't heavily used and the relocation stuff is in the middle of being re-written. I also changed the clear_extent_bit() to only unset the cached state if we are clearing EXTENT_LOCKED and related stuff, so we can do things like this
lock_extent_bits()
clear delalloc bitsunlock_extent_cached()
without losing our cached state. I tested this thoroughly and turned on LEAK_DEBUG to make sure we weren't leaking extent states, everything worked out fine.
V2: fixed verify_parent_transid to use the cached state for extent_buffer_uptodate and clear_extent_buffer_uptodate
--- fs/btrfs/ctree.h | 3 +- fs/btrfs/disk-io.c | 15 ++++--- fs/btrfs/extent-tree.c | 11 +++-- fs/btrfs/extent_io.c | 61 ++++++++++++++++---------- fs/btrfs/extent_io.h | 10 +++- fs/btrfs/file.c | 23 ++++++---- fs/btrfs/inode.c | 111 ++++++++++++++++++++++++++++++------------------ fs/btrfs/ioctl.c | 2 +- fs/btrfs/relocation.c | 2 +- 9 files changed, 148 insertions(+), 90 deletions(-)
fs/btrfs/ctree.h +2 -1
@@ -2309,7 +2309,8 @@ int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans, u32 min_type);int btrfs_start_delalloc_inodes (struct btrfs_root *root,int delay_iput); -int btrfs_set_extent_delalloc (struct inode *inode,u64 start,u64 end); +int btrfs_set_extent_delalloc (struct inode *inode,u64 start,u64 end, +struct extent_state **cached_state);int btrfs_writepages (struct address_space *mapping,struct writeback_control *wbc); int btrfs_create_subvol_root(struct btrfs_trans_handle *trans,
fs/btrfs/disk-io.c +9 -6
@@ -263,13 +263,15 @@ static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,static int verify_parent_transid (struct extent_io_tree *io_tree,struct extent_buffer *eb,u64 parent_transid) { +struct extent_state *cached_state =NULL ;int ret;if (!parent_transid ||btrfs_header_generation (eb) == parent_transid)return 0 ; -lock_extent (io_tree, eb->start, eb->start + eb->len -1 ,GFP_NOFS ); -if (extent_buffer_uptodate (io_tree, eb) && +lock_extent_bits (io_tree, eb->start, eb->start + eb->len -1 , +0 , &cached_state,GFP_NOFS ); +if (extent_buffer_uptodate (io_tree, eb, cached_state) &&btrfs_header_generation (eb) == parent_transid) { ret =0 ; goto out;
@@ -282,10 +284,10 @@ static int verify_parent_transid(struct extent_io_tree *io_tree, (unsigned long long)btrfs_header_generation(eb)); } ret =1 ; -clear_extent_buffer_uptodate (io_tree, eb); +clear_extent_buffer_uptodate (io_tree, eb, &cached_state); out: -unlock_extent (io_tree, eb->start, eb->start + eb->len -1 , -GFP_NOFS ); +unlock_extent_cached (io_tree, eb->start, eb->start + eb->len -1 , + &cached_state,GFP_NOFS );return ret; }
@@ -2436,7 +2438,8 @@ int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid)int ret;struct inode *btree_inode = buf->first_page->mapping->host; - ret =extent_buffer_uptodate (&BTRFS_I (btree_inode)->io_tree, buf); + ret =extent_buffer_uptodate (&BTRFS_I (btree_inode)->io_tree, buf, +NULL );if (!ret)return ret;
fs/btrfs/extent-tree.c +7 -4
@@ -6558,6 +6558,7 @@ static noinline int invalidate_extent_cache(struct btrfs_root *root,struct btrfs_key key;struct inode *inode =NULL ;struct btrfs_file_extent_item *fi; +struct extent_state *cached_state =NULL ;u64 num_bytes;u64 skip_objectid =0 ;u32 nritems;
@@ -6586,12 +6587,14 @@ static noinline int invalidate_extent_cache(struct btrfs_root *root, } num_bytes =btrfs_file_extent_num_bytes (leaf, fi); -lock_extent (&BTRFS_I (inode)->io_tree, key.offset, - key.offset + num_bytes -1 ,GFP_NOFS ); +lock_extent_bits (&BTRFS_I (inode)->io_tree, key.offset, + key.offset + num_bytes -1 ,0 , &cached_state, +GFP_NOFS );btrfs_drop_extent_cache (inode, key.offset, key.offset + num_bytes -1 ,1 ); -unlock_extent (&BTRFS_I (inode)->io_tree, key.offset, - key.offset + num_bytes -1 ,GFP_NOFS ); +unlock_extent_cached (&BTRFS_I (inode)->io_tree, key.offset, + key.offset + num_bytes -1 , &cached_state, +GFP_NOFS );cond_resched (); }iput (inode);
fs/btrfs/extent_io.c +38 -23
171 lines not on this page · raw message
fs/btrfs/extent_io.h +7 -3
@@ -164,6 +164,8 @@ int lock_extent(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask);int lock_extent_bits (struct extent_io_tree *tree,u64 start,u64 end,int bits,struct extent_state **cached,gfp_t mask);int unlock_extent (struct extent_io_tree *tree,u64 start,u64 end,gfp_t mask); +int unlock_extent_cached (struct extent_io_tree *tree,u64 start,u64 end, +struct extent_state **cached,gfp_t mask);int try_lock_extent (struct extent_io_tree *tree,u64 start,u64 end,gfp_t mask); int extent_read_full_page(struct extent_io_tree *tree, struct page *page,
@@ -197,7 +199,7 @@ int clear_extent_ordered(struct extent_io_tree *tree, u64 start, u64 end,int clear_extent_ordered_metadata (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, -gfp_t mask); +struct extent_state **cached_state,gfp_t mask);int set_extent_ordered (struct extent_io_tree *tree,u64 start,u64 end,gfp_t mask); int find_first_extent_bit(struct extent_io_tree *tree, u64 start,
@@ -282,9 +284,11 @@ int test_extent_buffer_dirty(struct extent_io_tree *tree,int set_extent_buffer_uptodate (struct extent_io_tree *tree,struct extent_buffer *eb);int clear_extent_buffer_uptodate (struct extent_io_tree *tree, -struct extent_buffer *eb); +struct extent_buffer *eb, +struct extent_state **cached_state);int extent_buffer_uptodate (struct extent_io_tree *tree, -struct extent_buffer *eb); +struct extent_buffer *eb, +struct extent_state *cached_state); int map_extent_buffer(struct extent_buffer *eb, unsigned long offset, unsigned long min_len, char **token, char **map, unsigned long *map_start,
fs/btrfs/file.c +14 -9
@@ -123,7 +123,8 @@ static noinline int dirty_and_release_pages(struct btrfs_trans_handle *trans, root->sectorsize - 1) & ~((u64)root->sectorsize - 1); end_of_last_block = start_pos + num_bytes -1 ; - err =btrfs_set_extent_delalloc (inode, start_pos, end_of_last_block); + err =btrfs_set_extent_delalloc (inode, start_pos, end_of_last_block, +NULL );if (err)return err;
@@ -751,6 +752,7 @@ static noinline int prepare_pages(struct btrfs_root *root, struct file *file, loff_t pos, unsigned long first_index, unsigned long last_index, size_t write_bytes) { + struct extent_state *cached_state = NULL;int i;unsigned long index = pos >>PAGE_CACHE_SHIFT ;struct inode *inode =fdentry (file)->d_inode;
@@ -779,16 +781,18 @@ again: }if (start_pos < inode->i_size) {struct btrfs_ordered_extent *ordered; -lock_extent (&BTRFS_I (inode)->io_tree, - start_pos, last_pos -1 ,GFP_NOFS ); +lock_extent_bits (&BTRFS_I (inode)->io_tree, + start_pos, last_pos -1 ,0 , &cached_state, +GFP_NOFS ); ordered =btrfs_lookup_first_ordered_extent (inode, last_pos -1 );if (ordered && ordered->file_offset + ordered->len > start_pos && ordered->file_offset < last_pos) {btrfs_put_ordered_extent (ordered); -unlock_extent (&BTRFS_I (inode)->io_tree, - start_pos, last_pos -1 ,GFP_NOFS ); +unlock_extent_cached (&BTRFS_I (inode)->io_tree, + start_pos, last_pos -1 , + &cached_state,GFP_NOFS );for (i =0 ; i < num_pages; i++) {unlock_page (pages[i]);page_cache_release (pages[i]);
@@ -800,12 +804,13 @@ again:if (ordered)btrfs_put_ordered_extent (ordered); -clear_extent_bits (&BTRFS_I (inode)->io_tree, start_pos, +clear_extent_bit (&BTRFS_I (inode)->io_tree, start_pos, last_pos -1 ,EXTENT_DIRTY |EXTENT_DELALLOC | -EXTENT_DO_ACCOUNTING , +EXTENT_DO_ACCOUNTING ,0 ,0 , &cached_state,GFP_NOFS ); -unlock_extent (&BTRFS_I (inode)->io_tree, - start_pos, last_pos -1 ,GFP_NOFS ); +unlock_extent_cached (&BTRFS_I (inode)->io_tree, + start_pos, last_pos -1 , &cached_state, +GFP_NOFS ); }for (i =0 ; i < num_pages; i++) {clear_page_dirty_for_io (pages[i]);
fs/btrfs/inode.c +69 -42
306 lines not on this page · raw message
fs/btrfs/ioctl.c +1 -1
@@ -538,7 +538,7 @@ again:*/ clear_page_dirty_for_io (page); -btrfs_set_extent_delalloc (inode, page_start, page_end); +btrfs_set_extent_delalloc (inode, page_start, page_end,NULL );set_page_dirty (page);unlock_extent (io_tree, page_start, page_end,GFP_NOFS );unlock_page (page);
fs/btrfs/relocation.c +1 -1
@@ -2659,7 +2659,7 @@ static int relocate_file_extent_cluster(struct inode *inode, EXTENT_BOUNDARY, GFP_NOFS); nr++; } -btrfs_set_extent_delalloc (inode, page_start, page_end); +btrfs_set_extent_delalloc (inode, page_start, page_end,NULL );set_page_dirty (page); dirty_page++;
Signature · 1 line
1.6.6