Re: [PATCH 6.1.y 6.6.y 0/3] mm/filemap: fix page cache corruption with large folios

[Yafang Shao <laoar.shao@xxxxxxxxx>]

Hello,

After applying this patch series to our 6.1 kernel and enabling XFS 
large folios, we encountered random core dumps in one of our Hadoop 
services. These core dumps occurred sporadically over several weeks, and 
it took us a long time to identify XFS large folios as the root cause.

Reverting commit 6795801366da0cd3d99e27c37f020a8f16714886 ("xfs: Support 
large folios") eliminated the issue. However, we have yet to determine 
the exact cause, as there are no warnings or errors in the kernel logs 
when the core dump occurs. Additionally, we have not found a reliable 
way to reproduce it consistently. While we have managed to trigger the 
issue within a few days on our test servers, this has not provided 
significant insights.

At this point, XFS large folios appear to be unreliable in the 6.1.y 
stable kernel.

We would appreciate any suggestions, such as adding debug messages to 
the kernel source code, to help us diagnose the root cause.


Regards

Yafang