Created attachment 188201 [details] journal of error Sep 23 18:44:12 sv kernel: BUG: unable to handle kernel NULL pointer dereference at (null) Sep 23 18:44:12 sv kernel: IP: [<ffffffffa07d8b91>] get_free_stripe+0x31/0xf0 [raid456]
This is almost certainly fixed by Commit: 49895bcc7e56 ("md/raid5: don't let shrink_slab shrink too far.") I had hoped this would be in 4.1-stable by now, but it isn't. I have just sent to to stable@vger.kernel.org so it should be in 4.1.9. If you are going to apply it, it would be good to apply Commit: 2d5b569b665e ("md/raid5: avoid races when changing cache size.") first. It has a simple easy to resolve conflict.
Created attachment 188251 [details] Commit: 49895bcc7e56
Created attachment 188261 [details] Commit: 2d5b569b665e
Neil Brown, thank you for your fast reply. I was getting worried because every time the qemu VM would tax the MD RAID5 hard, my server would lock up. I use Arch Linux so I create patches from your Commits, added them to the PKGBUILD, rebuilt the linux package, installed, and rebooted. I have tried very hard now for over 2 hours to get it to lock up again. I can confirm this solved my problem. I attached the patches just to confirm which ones I tried. Thanks again, Neil.
Yes, those are the correct patches. Thanks for the confirmation that they fix the problem.