[SOLVED] trap divide error / divide error: 0000 [#1] SMP kernel bug

[grin]

Hello,

This seems to be a rather critical one: under high load multisocket machines crash ceph osd [and other processes] with various division by zero errors.

The patch to fix this seems to be referenced here:
https://lkml.org/lkml/2016/7/12/297
See the thread for the details. Should apply to 4.4.x as well.

TIA

 

[fabian]

that patch is already included in our 4.4 kernel since pve-kernel-4.4.13-2-pve 4.4.13-59

 

[grin]

The repo offers only 4.4.13-58
Where should I find it?

[UPDATE] I see there's 4.4.16-64, I guess that fits your description.

 

[fabian]

yes, I only checked the git repository histroy, it might be that the 4.4.13-59 package never hit the public package repositories because it was replaced by a later kernel package altogether. for proxmox 4.x, only the most uptodate 4.4 kernel in the repository is supported.