Forwarded: [PATCH] f2fs: fix hung task in block_operations during checkpoint

[syzbot <syzbot+4235e4d7b6fd75704528@syzkaller.appspotmail.com>]

For archival purposes, forwarding an incoming command email to
linux-kernel@vger.kernel.org, syzkaller-bugs@googlegroups.com.

***

Subject: [PATCH] f2fs: fix hung task in block_operations during checkpoint
Author: kartikey406@gmail.com

#syz test: git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git master

        f2fs_sync_inode_meta() can return 0 (success) even when
        f2fs_update_inode_page() fails and triggers f2fs_stop_checkpoint().
        This happens because the error flag check only occurs at the start
        of each loop iteration, not after f2fs_update_inode_page() returns.

        When I/O errors occur:
        1. f2fs_update_inode_page() retries 8 times then calls
           f2fs_stop_checkpoint(), which sets CP_ERROR_FLAG
        2. f2fs_sync_inode_meta() returns 0 without checking the error flag
        3. block_operations() sees success and loops back to retry_flush_quotas
        4. Dirty inodes remain on list (sync failed), loop repeats forever
        5. Checkpoint never completes, waiters block indefinitely

        This causes hung tasks when operations like unlink wait for checkpoint
        completion while holding locks that other tasks need.

        Fix by checking f2fs_cp_error() after processing each inode in
        f2fs_sync_inode_meta() to detect errors from f2fs_update_inode_page().

        Reported-by: syzbot+4235e4d7b6fd75704528@syzkaller.appspotmail.com
        Closes: https://syzkaller.appspot.com/bug?extid=4235e4d7b6fd75704528
        Signed-off-by: Deepanshu Kartikey <kartikey406@gmail.com>
---
 fs/f2fs/checkpoint.c | 11 ++++++++---
 1 file changed, 8 insertions(+), 3 deletions(-)

diff --git a/fs/f2fs/checkpoint.c b/fs/f2fs/checkpoint.c
index bbe07e3a6c75..b0b5b792e092 100644
--- a/fs/f2fs/checkpoint.c
+++ b/fs/f2fs/checkpoint.c
@@ -1659,9 +1659,10 @@ int f2fs_write_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc)
 			return 0;
 		f2fs_warn(sbi, "Start checkpoint disabled!");
 	}
+	printk("f2fs_cp: 1 before cp_global_sem\n");
 	if (cpc->reason != CP_RESIZE)
 		f2fs_down_write(&sbi->cp_global_sem);
-
+	 printk("f2fs_cp: 2 after cp_global_sem\n");
 	stat_cp_time(cpc, CP_TIME_LOCK);
 
 	if (!is_sbi_flag_set(sbi, SBI_IS_DIRTY) &&
@@ -1669,16 +1670,18 @@ int f2fs_write_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc)
 		((cpc->reason & CP_DISCARD) && !sbi->discard_blks)))
 		goto out;
 	if (unlikely(f2fs_cp_error(sbi))) {
+		printk("f2fs_cp: 3 cp_error detected early\n");
 		err = -EIO;
 		goto out;
 	}
 
 	trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "start block_ops");
-
+	printk("f2fs_cp: 4 before block_operations\n");
 	err = block_operations(sbi);
+	printk("f2fs_cp: 5 after block_operations err=%d cp_error=%d\n", err, f2fs_cp_error(sbi));
 	if (err)
 		goto out;
-
+	//printk("f2fs_cp: 6 before do_checkpoint\n");
 	stat_cp_time(cpc, CP_TIME_OP_LOCK);
 
 	trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish block_ops");
@@ -1724,8 +1727,10 @@ int f2fs_write_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc)
 
 	/* save inmem log status */
 	f2fs_save_inmem_curseg(sbi);
+	printk("f2fs_cp: 4 before block_operations\n");
 
 	err = do_checkpoint(sbi, cpc);
+	printk("f2fs_cp: 7 after do_checkpoint err=%d\n", err);
 	if (err) {
 		f2fs_err(sbi, "do_checkpoint failed err:%d, stop checkpoint", err);
 		f2fs_bug_on(sbi, !f2fs_cp_error(sbi));
-- 
2.43.0