Public bug reported:

It appears that an cgroup-OOM-killed process can cause a kernel
panic/reboot.

Affected system:
- Ubuntu/Noble
- kernel 6.8.0-138-generic
- zfs 2.2.2-0ubuntu9.5

This has been documented here:
https://github.com/openzfs/zfs/issues/15918 - "kernel BUG at mm/usercopy.c:102"

Fix is here:
https://github.com/openzfs/zfs/pull/18408 - landed in 2.4.2

Request for backport to 2.2.x is here:
https://github.com/openzfs/zfs/issues/19149

In my case the logs look like:
```
SOMETHING invoked oom-killer: gfp_mask=0xcc0(GFP_KERNEL), order=0, 
oom_score_adj=999
...
Memory cgroup stats for 
/kubepods.slice/kubepods-burstable.slice/kubepods-burstable-pod3015e4bc_2688_43c1_bcbe_a918dd63b8b9.slice:
...
usercopy: Kernel memory exposure attempt detected from vmalloc 'no area' 
(offset 0, size 21058)!
```
Only that last message points to the actual bug. The OOM itself is just noise 
that precedes the bug.

In our case, no additional logs were found, probably because the system
rebooted before ZFS had a chance to commit the logs. This has happened
twice on the same system in September now.

Filing the bug here so others can see it too. If it happens often
enough, maybe it's time for a backport.

Cheers,
Walter Doekes
OSSO B.V.

** Affects: zfs-linux (Ubuntu)
     Importance: Undecided
         Status: New

** Description changed:

  It appears that an cgroup-OOM-killed process can cause a kernel
  panic/reboot.
  
  Affected system:
  - Ubuntu/Noble
  - kernel 6.8.0-138-generic
  - zfs 2.2.2-0ubuntu9.5
  
  This has been documented here:
  https://github.com/openzfs/zfs/issues/15918 - "kernel BUG at 
mm/usercopy.c:102"
  
  Fix is here:
  https://github.com/openzfs/zfs/pull/18408 - landed in 2.4.2
  
  Request for backport to 2.2.x is here:
  https://github.com/openzfs/zfs/issues/19149
  
- In my case the looks like:
+ In my case the logs look like:
  ```
  SOMETHING invoked oom-killer: gfp_mask=0xcc0(GFP_KERNEL), order=0, 
oom_score_adj=999
  ...
  Memory cgroup stats for 
/kubepods.slice/kubepods-burstable.slice/kubepods-burstable-pod3015e4bc_2688_43c1_bcbe_a918dd63b8b9.slice:
  ...
  usercopy: Kernel memory exposure attempt detected from vmalloc 'no area' 
(offset 0, size 21058)!
  ```
  Only that last message points to the actual bug. The OOM itself is just noise 
that precedes the bug.
  
  In our case, no additional logs were found, probably because the system
  rebooted before ZFS had a chance to commit the logs. This has happened
  twice on the same system in September now.
  
  Filing the bug here so others can see it too. If it happens often
  enough, maybe it's time for a backport.
  
  Cheers,
  Walter Doekes
  OSSO B.V.

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2168414

Title:
  usercopy: Kernel memory exposure attempt detected from vmalloc

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/zfs-linux/+bug/2168414/+subscriptions


-- 
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to