1 |
On Wednesday, 11 December 2019 06:31:04 CET Andrew Udvare wrote: |
2 |
> I have been getting relatively consistent kernel panics on some call to |
3 |
> find_css_set and sometimes a stack trace that mentions cgroups. |
4 |
|
5 |
Can you provide the full kernel panic (a picture taken with phone or camera |
6 |
attached to an email is fine as well) and tell us at which point this happens? |
7 |
|
8 |
> On 5.4.0 I don't get this same crash and I added blocking of |
9 |
> auto-loading nvidia under the ramdisk just in case that's the issue, as |
10 |
> I was sometimes getting a similar crash on 5.4.0. |
11 |
|
12 |
I am "still" on 5.3.14, but not encountering any kernel panics myself. |
13 |
|
14 |
> /etc/default/grub: |
15 |
> |
16 |
> GRUB_PRELOAD_MODULES=lvm |
17 |
> GRUB_CMDLINE_LINUX="init=/usr/lib/systemd/systemd |
18 |
|
19 |
Not using systemd myself. |
20 |
|
21 |
> systemd.legacy_systemd_cgroup_controller=yes rd.driver.blacklist=nvidia |
22 |
> rd.driver.blacklist=nvidia_modeset rd.driver.blacklist=nvidia_drm" |
23 |
> GRUB_GFXPAYLOAD_LINUX="keep" |
24 |
> |
25 |
> The reason the legacy argument is there is because Docker won't work |
26 |
> under the new cgroups, for now. |
27 |
> |
28 |
> I have a couple of modules, but the one that sticks out most is nvidia. |
29 |
> This is the one I see in the stack trace. I have not seen a bug report |
30 |
> on Gentoo or Nvidia's end. |
31 |
|
32 |
Nvidia (and some other modules) "taint" the kernel. Any module that "taints" |
33 |
the kernel WILL be listed when a kernel-panic occurs. I never found that info |
34 |
really helpful as I never had those modules to be the cause. |
35 |
|
36 |
> For now I've masked >5.4.0 gentoo-sources. |
37 |
> |
38 |
> Anyone else getting a similar issue? |
39 |
|
40 |
No, but an upgrade to 5.4.x is on my todo-list. |
41 |
|
42 |
-- |
43 |
Joost |