f2fs: fix to update user block counts in block_operations()
Chao Yu
maintainerPATCH 1/1 f2fs: fix to update user block counts in block_operations()
Commit 59c9081bc86e ("f2fs: allow write page cache when writting cp") allows write() to write data to page cache during checkpoint, so block count fields like .total_valid_block_count, .alloc_valid_block_count and .rf_node_block_count may encounter race condition as below:
CP Thread A
- write_checkpoint
- block_operations
- f2fs_down_write(&sbi->node_change)
- __prepare_cp_block
: ckpt->valid_block_count = .total_valid_block_count
- f2fs_up_write(&sbi->node_change)
- write
- f2fs_preallocate_blocks
- f2fs_map_blocks(,F2FS_GET_BLOCK_PRE_AIO)
- f2fs_map_lock
- f2fs_down_read(&sbi->node_change)
- f2fs_reserve_new_blocks
- inc_valid_block_count
: percpu_counter_add(&sbi->alloc_valid_block_count, count) : sbi->total_valid_block_count += count
- f2fs_up_read(&sbi->node_change)
- do_checkpoint
: sbi->last_valid_block_count = sbi->total_valid_block_count
: percpu_counter_set(&sbi->alloc_valid_block_count, 0)
: percpu_counter_set(&sbi->rf_node_block_count, 0)
- fsync
- need_do_checkpoint
- f2fs_space_for_roll_forward
: alloc_valid_block_count was reset to zero,
so, it may missed last data during checkpoint
Let's change to update .total_valid_block_count, .alloc_valid_block_count and .rf_node_block_count in block_operations(), then their access can be protected by .node_change and .cp_rwsem lock, so that it can avoid above race condition.
--- fs/f2fs/checkpoint.c | 11 ++++++----- 1 file changed, 6 insertions(+), 5 deletions(-)
fs/f2fs/checkpoint.c +6 -5
@@ -1298,6 +1298,12 @@ static int block_operations(struct f2fs_sb_info *sbi)* dirty node blocks and some checkpoint values by block allocation. */ __prepare_cp_block (sbi); + +/* update user_block_counts */ + sbi->last_valid_block_count = sbi->total_valid_block_count; +percpu_counter_set (&sbi->alloc_valid_block_count,0 ); +percpu_counter_set (&sbi->rf_node_block_count,0 ); +f2fs_up_write (&sbi->node_change);return err; }
@@ -1575,11 +1581,6 @@ static int do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) start_blk +=NR_CURSEG_NODE_TYPE ; } -/* update user_block_counts */ - sbi->last_valid_block_count = sbi->total_valid_block_count; -percpu_counter_set (&sbi->alloc_valid_block_count,0 ); -percpu_counter_set (&sbi->rf_node_block_count,0 ); -/* Here, we have one bio having CP pack except cp pack 2 page */ f2fs_sync_meta_pages (sbi,META ,LONG_MAX ,FS_CP_META_IO );/* Wait for all dirty meta pages to be submitted for IO */
Signature · 1 line
2.40.1
Jaegeuk Kim
maintainerQuoted 49 lines · Chao Yu
Commit
59c9081bc86e("f2fs: allow write page cache when writting cp") allowswrite()to write data to page cache during checkpoint, so block count fields like .total_valid_block_count, .alloc_valid_block_count and .rf_node_block_count may encounter race condition as below:CP Thread A
-write_checkpoint
-block_operations
-f2fs_down_write(&sbi->node_change)
-__prepare_cp_block
:ckpt->valid_block_count= .total_valid_block_count
-f2fs_up_write(&sbi->node_change)
- write
-f2fs_preallocate_blocks
-f2fs_map_blocks(,F2FS_GET_BLOCK_PRE_AIO)
-f2fs_map_lock
-f2fs_down_read(&sbi->node_change)
-f2fs_reserve_new_blocks
-inc_valid_block_count
:percpu_counter_add(&sbi->alloc_valid_block_count, count):sbi->total_valid_block_count+= count
-f2fs_up_read(&sbi->node_change)
-do_checkpoint
:sbi->last_valid_block_count=sbi->total_valid_block_count
:percpu_counter_set(&sbi->alloc_valid_block_count, 0)
:percpu_counter_set(&sbi->rf_node_block_count, 0)
- fsync
-need_do_checkpoint
-f2fs_space_for_roll_forward
:alloc_valid_block_countwas reset to zero,
so, it may missed last data during checkpointLet's change to update .total_valid_block_count, .alloc_valid_block_count and .rf_node_block_count in
block_operations(), then their access can be protected by .node_change and .cp_rwsem lock, so that it can avoid above race condition.Fixes 59c9081bc86e ("f2fs: allow write page cache when writting cp")Cc Yunlei He <[email protected]>Signed-off-by Chao Yu <[email protected]>--- fs/f2fs/checkpoint.c | 11 ++++++----- 1 file changed, 6 insertions(+), 5 deletions(-)fs/f2fs/checkpoint.c +0
-0
fs/f2fs/checkpoint.c static int block_operations(struct f2fs_sb_info *sbi) · line 1305
Quoted 8 lines
fs/f2fs/checkpoint.c +5
-0@@ -1298,6 +1298,12 @@ static int block_operations(struct f2fs_sb_info *sbi)* dirty node blocks and some checkpoint values by block allocation. */ __prepare_cp_block (sbi); + +/* update user_block_counts */ + sbi->last_valid_block_count = sbi->total_valid_block_count; +percpu_counter_set (&sbi->alloc_valid_block_count,0 ); +percpu_counter_set (&sbi->rf_node_block_count,0 );
Need to add this in __prepare_cp_block()?
Quoted 19 lines
fs/f2fs/checkpoint.c +1
-5+f2fs_up_write (&sbi->node_change);return err; }@@ -1575,11 +1581,6 @@ static int do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) start_blk +=NR_CURSEG_NODE_TYPE ; } -/* update user_block_counts */ - sbi->last_valid_block_count = sbi->total_valid_block_count; -percpu_counter_set (&sbi->alloc_valid_block_count,0 ); -percpu_counter_set (&sbi->rf_node_block_count,0 ); -/* Here, we have one bio having CP pack except cp pack 2 page */ f2fs_sync_meta_pages (sbi,META ,LONG_MAX ,FS_CP_META_IO );/* Wait for all dirty meta pages to be submitted for IO */ Signature · 1 line
2.40.1
Chao Yu
maintainerQuoted 45 lines · Jaegeuk Kim
Quoted 44 lines · Chao Yu
Commit
59c9081bc86e("f2fs: allow write page cache when writting cp") allowswrite()to write data to page cache during checkpoint, so block count fields like .total_valid_block_count, .alloc_valid_block_count and .rf_node_block_count may encounter race condition as below:CP Thread A
-write_checkpoint
-block_operations
-f2fs_down_write(&sbi->node_change)
-__prepare_cp_block
:ckpt->valid_block_count= .total_valid_block_count
-f2fs_up_write(&sbi->node_change)
- write
-f2fs_preallocate_blocks
-f2fs_map_blocks(,F2FS_GET_BLOCK_PRE_AIO)
-f2fs_map_lock
-f2fs_down_read(&sbi->node_change)
-f2fs_reserve_new_blocks
-inc_valid_block_count
:percpu_counter_add(&sbi->alloc_valid_block_count, count):sbi->total_valid_block_count+= count
-f2fs_up_read(&sbi->node_change)
-do_checkpoint
:sbi->last_valid_block_count=sbi->total_valid_block_count
:percpu_counter_set(&sbi->alloc_valid_block_count, 0)
:percpu_counter_set(&sbi->rf_node_block_count, 0)
- fsync
-need_do_checkpoint
-f2fs_space_for_roll_forward
:alloc_valid_block_countwas reset to zero,
so, it may missed last data during checkpointLet's change to update .total_valid_block_count, .alloc_valid_block_count and .rf_node_block_count in
block_operations(), then their access can be protected by .node_change and .cp_rwsem lock, so that it can avoid above race condition.Fixes 59c9081bc86e ("f2fs: allow write page cache when writting cp")Cc Yunlei He <[email protected]>Signed-off-by Chao Yu <[email protected]>--- fs/f2fs/checkpoint.c | 11 ++++++----- 1 file changed, 6 insertions(+), 5 deletions(-)
Quoted 15 lines
fs/f2fs/checkpoint.c static int block_operations(struct f2fs_sb_info *sbi) · line 1305
Quoted 13 lines
fs/f2fs/checkpoint.c +5
-0@@ -1298,6 +1298,12 @@ static int block_operations(struct f2fs_sb_info *sbi)* dirty node blocks and some checkpoint values by block allocation. */ __prepare_cp_block (sbi); + +/* update user_block_counts */ + sbi->last_valid_block_count = sbi->total_valid_block_count; +percpu_counter_set (&sbi->alloc_valid_block_count,0 ); +percpu_counter_set (&sbi->rf_node_block_count,0 );Need to add this in
__prepare_cp_block()?
Fine to me, will update in v2.
Thanks,
Quoted 20 lines
Quoted 20 lines
fs/f2fs/checkpoint.c +1
-5+f2fs_up_write (&sbi->node_change);return err; }@@ -1575,11 +1581,6 @@ static int do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) start_blk +=NR_CURSEG_NODE_TYPE ; } -/* update user_block_counts */ - sbi->last_valid_block_count = sbi->total_valid_block_count; -percpu_counter_set (&sbi->alloc_valid_block_count,0 ); -percpu_counter_set (&sbi->rf_node_block_count,0 ); -/* Here, we have one bio having CP pack except cp pack 2 page */ f2fs_sync_meta_pages (sbi,META ,LONG_MAX ,FS_CP_META_IO );/* Wait for all dirty meta pages to be submitted for IO */ Signature · 1 line
2.40.1