f2fs: don't set RO when shutting down f2fs
authorJaegeuk Kim <jaegeuk@kernel.org>
Wed, 3 Apr 2024 23:07:53 +0000 (23:07 +0000)
committerJaegeuk Kim <jaegeuk@kernel.org>
Fri, 12 Apr 2024 20:58:35 +0000 (20:58 +0000)
Shutdown does not check the error of thaw_super due to readonly, which
causes a deadlock like below.

f2fs_ioc_shutdown(F2FS_GOING_DOWN_FULLSYNC)        issue_discard_thread
 - bdev_freeze
  - freeze_super
 - f2fs_stop_checkpoint()
  - f2fs_handle_critical_error                     - sb_start_write
    - set RO                                         - waiting
 - bdev_thaw
  - thaw_super_locked
    - return -EINVAL, if sb_rdonly()
 - f2fs_stop_discard_thread
  -> wait for kthread_stop(discard_thread);

Reported-by: "Light Hsieh (謝明燈)" <Light.Hsieh@mediatek.com>
Reviewed-by: Daeho Jeong <daehojeong@google.com>
Reviewed-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
fs/f2fs/super.c

index 8ac4734c2df6036ac088ca8fa1b4df6d8d08f904..df32573d1f620949880a942768729211f4f14b19 100644 (file)
@@ -4159,9 +4159,15 @@ void f2fs_handle_critical_error(struct f2fs_sb_info *sbi, unsigned char reason,
        if (shutdown)
                set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
 
-       /* continue filesystem operators if errors=continue */
-       if (continue_fs || f2fs_readonly(sb))
+       /*
+        * Continue filesystem operators if errors=continue. Should not set
+        * RO by shutdown, since RO bypasses thaw_super which can hang the
+        * system.
+        */
+       if (continue_fs || f2fs_readonly(sb) || shutdown) {
+               f2fs_warn(sbi, "Stopped filesystem due to reason: %d", reason);
                return;
+       }
 
        f2fs_warn(sbi, "Remounting filesystem read-only");
        /*