Control: forwarded -1 https://github.com/lxc/lxcfs/issues/608
Control: tags -1 + confirmed - moreinfo

  I finally had a free day to investigate this, and I tracked down the
issue: too many CPUs! :)

  32bit arm systems report an additional line in /proc/cpuinfo, which
isn't accounted for in lxcfs. That causes lxcfs to consume an
intentional buffer padding, until at about the 13th CPU the internal
buffer is consumed and lxcfs bails on trying to report a /proc/cpuinfo
to the container.

  Further details are in the linked bug report, but for now if the
armel and armhf VMs are limited to 12 CPUs, I think everything should
work as expected. Once there's a fix upstream, it can be cherry-picked
back to the version of lxcfs for bookworm in a point update.

Mathias

Attachment: signature.asc
Description: This is a digitally signed message part

Reply via email to