[PATCH 1/2] f2fs: fix to drop all dirty meta/node pages during umount()

[PATCH 1/2] f2fs: fix to drop all dirty meta/node pages during umount()

[Chao Yu]

For cp error case, there will be dirty meta/node pages remained after
f2fs_write_checkpoint() in f2fs_put_super(), drop them explicitly, and
do sanity check on reference count of dirty pages and inflight IOs.

Signed-off-by: Chao Yu <chao@kernel.org>
---
 fs/f2fs/super.c | 18 ++++++++++++++++--
 1 file changed, 16 insertions(+), 2 deletions(-)

diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
index 7499ebabaf35..d1bce753f0d2 100644
--- a/fs/f2fs/super.c
+++ b/fs/f2fs/super.c
@@ -1573,6 +1573,7 @@ static void f2fs_put_super(struct super_block *sb)
 {
 	struct f2fs_sb_info *sbi = F2FS_SB(sb);
 	int i;
+	int err = 0;
 	bool done;
 
 	/* unregister procfs/sysfs entries in advance to avoid race case */
@@ -1599,7 +1600,7 @@ static void f2fs_put_super(struct super_block *sb)
 		struct cp_control cpc = {
 			.reason = CP_UMOUNT,
 		};
-		f2fs_write_checkpoint(sbi, &cpc);
+		err = f2fs_write_checkpoint(sbi, &cpc);
 	}
 
 	/* be sure to wait for any on-going discard commands */
@@ -1608,7 +1609,7 @@ static void f2fs_put_super(struct super_block *sb)
 		struct cp_control cpc = {
 			.reason = CP_UMOUNT | CP_TRIMMED,
 		};
-		f2fs_write_checkpoint(sbi, &cpc);
+		err = f2fs_write_checkpoint(sbi, &cpc);
 	}
 
 	/*
@@ -1625,6 +1626,19 @@ static void f2fs_put_super(struct super_block *sb)
 
 	f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
 
+	if (err) {
+		truncate_inode_pages_final(NODE_MAPPING(sbi));
+		truncate_inode_pages_final(META_MAPPING(sbi));
+	}
+
+	for (i = 0; i < NR_COUNT_TYPE; i++) {
+		if (!get_pages(sbi, i))
+			continue;
+		f2fs_err(sbi, "detect filesystem reference count leak during "
+			"umount, type: %d, count: %lld", i, get_pages(sbi, i));
+		f2fs_bug_on(sbi, 1);
+	}
+
 	f2fs_bug_on(sbi, sbi->fsync_node_num);
 
 	f2fs_destroy_compress_inode(sbi);
-- 
2.40.1

[PATCH 2/2] f2fs: avoid dead loop in f2fs_issue_checkpoint()

[Chao Yu]

generic/082 reports a bug as below:

__schedule+0x332/0xf60
schedule+0x6f/0xf0
schedule_timeout+0x23b/0x2a0
wait_for_completion+0x8f/0x140
f2fs_issue_checkpoint+0xfe/0x1b0
f2fs_sync_fs+0x9d/0xb0
sync_filesystem+0x87/0xb0
dquot_load_quota_sb+0x41b/0x460
dquot_load_quota_inode+0xa5/0x130
dquot_quota_on+0x4b/0x60
f2fs_quota_on+0xe3/0x1b0
do_quotactl+0x483/0x700
__x64_sys_quotactl+0x15c/0x310
do_syscall_64+0x3f/0x90
entry_SYSCALL_64_after_hwframe+0x72/0xdc

The root casue is race case as below:

Thread A			Kworker			IRQ
- write()
: write data to quota.user file

				- writepages
				 - f2fs_submit_page_write
				  - __is_cp_guaranteed return false
				  - inc_page_count(F2FS_WB_DATA)
				 - submit_bio
- quotactl(Q_QUOTAON)
 - f2fs_quota_on
  - dquot_quota_on
   - dquot_load_quota_inode
    - vfs_setup_quota_inode
    : inode->i_flags |= S_NOQUOTA
							- f2fs_write_end_io
							 - __is_cp_guaranteed return true
							 - dec_page_count(F2FS_WB_CP_DATA)
    - dquot_load_quota_sb
     - f2fs_sync_fs
      - f2fs_issue_checkpoint
       - do_checkpoint
        - f2fs_wait_on_all_pages(F2FS_WB_CP_DATA)
        : loop due to F2FS_WB_CP_DATA count is negative

Calling filemap_fdatawrite() and filemap_fdatawait() to keep all data
clean before quota file setup.

Signed-off-by: Chao Yu <chao@kernel.org>
---
 fs/f2fs/super.c | 15 +++++++++++++--
 1 file changed, 13 insertions(+), 2 deletions(-)

diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
index d1bce753f0d2..0bfa98d053bd 100644
--- a/fs/f2fs/super.c
+++ b/fs/f2fs/super.c
@@ -2927,15 +2927,26 @@ static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
 		return -EBUSY;
 	}
 
+	if (path->dentry->d_sb != sb)
+		return -EXDEV;
+
 	err = f2fs_quota_sync(sb, type);
 	if (err)
 		return err;
 
-	err = dquot_quota_on(sb, type, format_id, path);
+	inode = d_inode(path->dentry);
+
+	err = filemap_fdatawrite(inode->i_mapping);
 	if (err)
 		return err;
 
-	inode = d_inode(path->dentry);
+	err = filemap_fdatawait(inode->i_mapping);
+		if (err)
+			return err;
+
+	err = dquot_quota_on(sb, type, format_id, path);
+	if (err)
+		return err;
 
 	inode_lock(inode);
 	F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
-- 
2.40.1

Re: [f2fs-dev] [PATCH 1/2] f2fs: fix to drop all dirty meta/node pages during umount()

[patchwork-bot+f2fs]

Hello:

This series was applied to jaegeuk/f2fs.git (dev)
by Jaegeuk Kim <jaegeuk@kernel.org>:

On Sun, 28 May 2023 15:47:12 +0800 you wrote:
> For cp error case, there will be dirty meta/node pages remained after
> f2fs_write_checkpoint() in f2fs_put_super(), drop them explicitly, and
> do sanity check on reference count of dirty pages and inflight IOs.
> 
> Signed-off-by: Chao Yu <chao@kernel.org>
> ---
>  fs/f2fs/super.c | 18 ++++++++++++++++--
>  1 file changed, 16 insertions(+), 2 deletions(-)

Here is the summary with links:
  - [f2fs-dev,1/2] f2fs: fix to drop all dirty meta/node pages during umount()
    https://git.kernel.org/jaegeuk/f2fs/c/50cfdbd772cf
  - [f2fs-dev,2/2] f2fs: avoid dead loop in f2fs_issue_checkpoint()
    (no matching commit)

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html