Warning Trace seen at https://elixir.bootlin.com/linux/v6.6.69/source/fs/ext4/ext4_jbd2.c#L73 post remount-ro

Warning Trace seen at https://elixir.bootlin.com/linux/v6.6.69/source/fs/ext4/ext4_jbd2.c#L73 post remount-ro

[Shramana Chakraborty]

Hi All,

I am following up on this bug, https://bugzilla.kernel.org/show_bug.cgi?id=220299, I raised last year regarding warning traces seen during systemd shutdown.

Warning Trace seen at https://elixir.bootlin.com/linux/v6.6.69/source/fs/ext4/ext4_jbd2.c#L73 is hit by the writeback system even though there is nothing to write and the filesystem has been mounted as readonly. 
An ext4 fs is mounted with journal=data and a large file is written to it. Then the fs is mounted as readonly-ro post a sync. At this point of time, data consistency is preserved through the sync itself followed by a sync from the remount-ro action. 
All the dirty pages should be marked as clean in the cache. Yet post the remount-ro, it hits https://elixir.bootlin.com/linux/v6.6.69/source/fs/ext4/ext4_jbd2.c#L73 and prints the trace as the fs is readonly now. 
Even though there is nothing to writeback, the wb_writeback is scheduled every 5 secs or so and the same trace can be seen if changed from WARN_ON_ONCE to WARN_ON.

I found that this was reported earlier as well via https://lkml.org/lkml/2024/5/17/983. 

This is still seen on 6.6.x LTS kernel with rootfs mounted with data=journal option. I wonder whether/how this could be addressed.

Thanks and Regards
Shramana Chakraborty