Skip to contentSkip to conversation

f2fs: Avoid double lock for cp_rwsem during checkpoint

PATCH Superseded Sayali Lokhande 2 participants 3 messages Apr 29, 2020 – Apr 30, 2020 F2FS file system

v3 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

PATCH v3 1/1 f2fs: Avoid double lock for cp_rwsem during checkpoint

There could be a scenario where f2fs_sync_node_pages gets called during checkpoint, which in turn tries to flush inline data and calls iput(). This results in deadlock as iput() tries to hold cp_rwsem, which is already held at the beginning by checkpoint->block_operations().

Call stack :

Thread A Thread B
f2fs_write_checkpoint()
- block_operations(sbi)
- f2fs_lock_all(sbi);
- down_write(&sbi->cp_rwsem);

- open()
- igrab()
- write() write inline data
- unlink()
- f2fs_sync_node_pages()
- if (is_inline_node(page))
- flush_inline_data()
- ilookup()
page = f2fs_pagecache_get_page()
if (!page)
goto iput_out;
iput_out:
-close()
-iput()
iput(inode);
- f2fs_evict_inode()
- f2fs_truncate_blocks()
- f2fs_lock_op()
- down_read(&sbi->cp_rwsem);

Fixes 399368372ed9 ("f2fs: introduce a new global lock scheme")
---
 fs/f2fs/checkpoint.c | 10 ++++------
 1 file changed, 4 insertions(+), 6 deletions(-)
fs/f2fs/checkpoint.c +4 -6
index 5ba649e..97b6378 100644
@@ -1219,21 +1219,19 @@ static int block_operations(struct f2fs_sb_info *sbi)
 		goto retry_flush_quotas;
 	}
 
-retry_flush_nodes:
 	down_write(&sbi->node_write);
 
 	if (get_pages(sbi, F2FS_DIRTY_NODES)) {
 		up_write(&sbi->node_write);
+		up_write(&sbi->node_change);
+		f2fs_unlock_all(sbi);
 		atomic_inc(&sbi->wb_sync_req[NODE]);
 		err = f2fs_sync_node_pages(sbi, &wbc, false, FS_CP_NODE_IO);
 		atomic_dec(&sbi->wb_sync_req[NODE]);
-		if (err) {
-			up_write(&sbi->node_change);
-			f2fs_unlock_all(sbi);
+		if (err)
 			goto out;
-		}
 		cond_resched();
-		goto retry_flush_nodes;
+		goto retry_flush_quotas;
 	}
 
 	/*
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project
#1 lore raw
Quoted 34 lines · Sayali Lokhande

There could be a scenario where f2fs_sync_node_pages gets called during checkpoint, which in turn tries to flush inline data and calls iput(). This results in deadlock as iput() tries to hold cp_rwsem, which is already held at the beginning by checkpoint->block_operations().

Call stack :

Thread A Thread B
f2fs_write_checkpoint()
- block_operations(sbi)
- f2fs_lock_all(sbi);
- down_write(&sbi->cp_rwsem);

- open()
- igrab()
- write() write inline data
- unlink()
- f2fs_sync_node_pages()
- if (is_inline_node(page))
- flush_inline_data()
- ilookup()
page = f2fs_pagecache_get_page()
if (!page)
goto iput_out;
iput_out:
-close()
-iput()
iput(inode);
- f2fs_evict_inode()
- f2fs_truncate_blocks()
- f2fs_lock_op()
- down_read(&sbi->cp_rwsem);

Quoted 1 line
Fixes 399368372ed9 ("f2fs: introduce a new global lock scheme")

IMO, it should be

2049d4fcb057 ("f2fs: avoid multiple node page writes due to inline_data")

It brings iput() to checkpoint process for the first time.

Thanks,

Quoted 36 lines
---
 fs/f2fs/checkpoint.c | 10 ++++------
 1 file changed, 4 insertions(+), 6 deletions(-)
fs/f2fs/checkpoint.c +4 -6
index 5ba649e..97b6378 100644
@@ -1219,21 +1219,19 @@ static int block_operations(struct f2fs_sb_info *sbi)
 		goto retry_flush_quotas;
 	}
 
-retry_flush_nodes:
 	down_write(&sbi->node_write);
 
 	if (get_pages(sbi, F2FS_DIRTY_NODES)) {
 		up_write(&sbi->node_write);
+		up_write(&sbi->node_change);
+		f2fs_unlock_all(sbi);
 		atomic_inc(&sbi->wb_sync_req[NODE]);
 		err = f2fs_sync_node_pages(sbi, &wbc, false, FS_CP_NODE_IO);
 		atomic_dec(&sbi->wb_sync_req[NODE]);
-		if (err) {
-			up_write(&sbi->node_change);
-			f2fs_unlock_all(sbi);
+		if (err)
 			goto out;
-		}
 		cond_resched();
-		goto retry_flush_nodes;
+		goto retry_flush_quotas;
 	}
 
 	/*
#2 lore raw
Quoted 35 lines · Chao Yu
Quoted 34 lines · Sayali Lokhande

There could be a scenario where f2fs_sync_node_pages gets called during checkpoint, which in turn tries to flush inline data and calls iput(). This results in deadlock as iput() tries to hold cp_rwsem, which is already held at the beginning by checkpoint->block_operations().

Call stack :

Thread A Thread B
f2fs_write_checkpoint()
- block_operations(sbi)
- f2fs_lock_all(sbi);
- down_write(&sbi->cp_rwsem);

- open()
- igrab()
- write() write inline data
- unlink()
- f2fs_sync_node_pages()
- if (is_inline_node(page))
- flush_inline_data()
- ilookup()
page = f2fs_pagecache_get_page()
if (!page)
goto iput_out;
iput_out:
-close()
-iput()
iput(inode);
- f2fs_evict_inode()
- f2fs_truncate_blocks()
- f2fs_lock_op()
- down_read(&sbi->cp_rwsem);

Quoted 8 lines
Quoted 1 line
Fixes 399368372ed9 ("f2fs: introduce a new global lock scheme")

IMO, it should be

2049d4fcb057 ("f2fs: avoid multiple node page writes due to inline_data")

It brings iput() to checkpoint process for the first time.

Thanks,

Agreed. will update it.

Quoted 36 lines
Quoted 36 lines
---
  fs/f2fs/checkpoint.c | 10 ++++------
  1 file changed, 4 insertions(+), 6 deletions(-)
fs/f2fs/checkpoint.c +4 -6
index 5ba649e..97b6378 100644
@@ -1219,21 +1219,19 @@ static int block_operations(struct f2fs_sb_info *sbi)
  		goto retry_flush_quotas;
  	}
 
-retry_flush_nodes:
  	down_write(&sbi->node_write);
 
  	if (get_pages(sbi, F2FS_DIRTY_NODES)) {
  		up_write(&sbi->node_write);
+		up_write(&sbi->node_change);
+		f2fs_unlock_all(sbi);
  		atomic_inc(&sbi->wb_sync_req[NODE]);
  		err = f2fs_sync_node_pages(sbi, &wbc, false, FS_CP_NODE_IO);
  		atomic_dec(&sbi->wb_sync_req[NODE]);
-		if (err) {
-			up_write(&sbi->node_change);
-			f2fs_unlock_all(sbi);
+		if (err)
  			goto out;
-		}
  		cond_resched();
-		goto retry_flush_nodes;
+		goto retry_flush_quotas;
  	}
 
  	/*