Kernel 5.4.44 causes system freeze on HP MicroServer Gen8

[SnowFox]

Glad to hear the issue appears to have been solved! I just installed pve-kernel-5.4.44-2-pve on my machine.

 

[stephane de Labrusse]

I had no luck with pve-kernel-5.4.44-2-pve, I needed to roll back to the stable previous kernel. Now I have no freeze anymore, with the last kernel I just need a couple of hours to get a server down

Not sure the story is over

AMD FX(tm)-8320 Eight-Core Processor

 

[t.lamprecht]

Not sure the story is over

As our test case here and all others mentioned are working fine I'd say the story for the "cgroup_bpf_run_filter_skb NULL pointer dereference" is over, for now at least.

You can check if you see a kernel oops/panic message somewhere (tty1 or just ssh/console with dmesg -wT open.

This specific issue would have at least the following lines included:

[Wed Jul  1 13:54:11 2020] BUG: kernel NULL pointer dereference, address: 0000000000000010
[Wed Jul  1 13:54:11 2020] #PF: supervisor read access in kernel mode
[Wed Jul  1 13:54:11 2020] #PF: error_code(0x0000) - not-present page
[ ...snip .. ]
[Wed Jul  1 13:54:11 2020] RIP: 0010:__cgroup_bpf_run_filter_skb+0x26d/0x3d0

If yours doesn't it's highly probable something completely different.

AMD FX(tm)-8320 Eight-Core Processor

Quite the old processor, could be another regression. So please:

1. Check that your motherboards firmware is up to date
2. Ensure that the amd64-microcode is installed
3. get some hands on that server to ensure you get any kernel log if it crashes

If point 3 hits and you got some info please open another thread with that.

 

[stephane de Labrusse]

If yours doesn't it's highly probable something completely different.

thank for all of this, indeed I had maybe something different, will check

Highly appreciated your inputs

 

[H4R0]

pve-kernel-5.4.44-2-pve fixed the kernel panic for me running 48hours

 

[sahostking]

ok looks for me like still happening when I do an resize on lvm partitions or anything runs against lvm.
If I dont then its fine but as soon as I try to make a change to lvm partition it dies. Tried it on few servers same issue on pve-kernel-5.4.44-2-pve.

I see these stuck:
root 5938 0.0 0.0 15848 8724 ? D 08:40 0:00 /sbin/vgs --separator : --noheadings --units b --unbuffered --nosuffix --options vg_name,vg_size,vg_free,lv_count
root 7077 0.0 0.0 15848 8572 ? D 08:45 0:00 /sbin/vgs --separator : --noheadings --units b --unbuffered --nosuffix --options vg_name,vg_size,vg_free,lv_count
root 7985 0.0 0.0 6072 892 pts/0 S+ 08:49 0:00 grep vgs
root 15259 0.0 0.0 15848 8640 ? D 02:36 0:00 /sbin/vgs --separator : --noheadings --units b --unbuffered --nosuffix --options vg_name,vg_size,vg_free,lv_count
root 19983 0.0 0.0 15848 8724 ? D 02:56 0:00 /sbin/vgs --separator : --noheadings --units b --unbuffered --nosuffix --options vg_name,vg_size,vg_free,lv_count


pve-kernel-5.4.44-2-pve atleast stops the vms going down randomly anymore so thats good but still caused the above for me.

Moved back to 5.3 pve kernel which has no issues and I can resize etc. fine.

 

[t.lamprecht]

ok looks for me like still happening when I do an resize on lvm partitions or anything runs against lvm.

This seems like another issue, please open a new thread for it. Check dmesg -T for any error message.
FYI: lvresize works fine with the 5.4 kernel here, so this may also be something more specific top your setup.

 

[Killvearn]

pve-kernel-5.4.44-2-pve fixed it on AMD too (ryzen 7 1700)