On 06/03/2025 17:37, Malte Schröder wrote: > On 16/02/2025 17:12, Malte Schröder wrote: >> On 16/02/2025 14:10, Matthew Wilcox wrote: >>> On Sun, Feb 16, 2025 at 12:26:06AM +0100, Malte Schröder wrote: >>>> On 15/02/2025 18:11, Matthew Wilcox wrote: >>>>> On Sat, Feb 15, 2025 at 01:34:33PM +0100, Malte Schröder wrote: >>>>>> Hi, >>>>>> I am getting stuff freezing randomly since 6.14-rc. I do not have a clear way to >>>>> When you say "since 6.14-rc", what exactly do you mean? 6.13 is fine >>>>> and 6.14-rc2 is broken? Or some other version? >>>> 6.13 and 6.13 + bcachefs-master was fine. Issue started with 6.14-rc1. >>> That's interesting. >>> >>>>> This seems very similar to all of these syzbot reports: >>>>> https://lore.kernel.org/linux-bcachefs/Z6-o5A4Y-rf7Hq8j@xxxxxxxxxxxxxxxxxxxx/ >>>>> >>>>> Fortunately, syzbot thinks it's bisected one of them: >>>>> https://lore.kernel.org/linux-bcachefs/67b0bf29.050a0220.6f0b7.0010.GAE@xxxxxxxxxx/ >>>>> >>>>> Can you confirm? >>>> >From my limited understanding of how bcachefs works I do not think this >>>> commit is the root cause of this issue. That commit just changes the >>>> autofix flags, so it might just uncover some other issue in fsck code. >>>> Also I've been running that code before the 6.14 merge without issues. >>> If you have time to investigate this, seeing if you can reproduce this on >>> commit 141526548052 and then (if it does reproduce) bisecting between that >>> and v6.13-rc3 might lead us to the real commit that's causing the problem. >>> >> I will try. But I will need to find a way to reliably reproduce my issue >> first. > I did not find a reliable way to reproduce this issue. It happens like > every few weeks, so bisecting is not an option for me. Sorry. > > > It is also is hard to distinguish, which kind of freeze I just > encountered. I also found issues with apparmor vs. resume (just > reported) and I have a feeling something is going on in amdgpu-land but > can't quite pinpoint it, yet. > > > /Malte > I get the feeling this issue lies deeper. I get this about once a week, but also when there is nothing really going on IO wise. May there be a deeper issue with rcu? Who to involve? /Malte
