Closed marc-hb closed 3 months ago
More context:
@plbossart , @bardliao can we merge this cherry-pick in topic/sof-dev
? It gets rid of a lot of red.
Summarizing a chat with @msatwood (thx!)
When multiple drivers compete for the same (PCI) IDs, udev
loads all the non-blocklisted ones. This is the case for i915
and xe
on TGL, MTL and others, example:
some-MTL$ modinfo i915 | grep 7D55
alias: pci:v00008086d00007D55sv*sd*bc03sc*i*
some-MTL$ modinfo xe | grep 7D55
alias: pci:v00008086d00007D55sv*sd*bc03sc*i*
This gives an opportunity for competing drivers to either "fight it out", or peacefully share models based on some sub-IDs, etc.
In most of our Ubuntu configurations, i915
gets placed in the initramfs and loaded first. xe
gets loaded by udev
at some point later but notices it came second and admits defeat. It stays loaded and visible in lsmod
but in an less tested, inactive state.... which triggered this error.
In other words this fix and PR is a "better NO-OP"
An empty sysctl table entry was inadvertently left behind for observation sysctl. The breaks on 6.11 with the following errors:
[ 219.654850] sysctl table check failed: dev/xe/(null) procname is null [ 219.654862] sysctl table check failed: dev/xe/(null) No proc_handler
Drop the empty entry.
Fixes: 8169b2097d88 ("drm/xe/uapi: Rename xe perf layer as xe observation layer") Closes: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/2419
Reviewed-by: Umesh Nerlige Ramappa umesh.nerlige.ramappa@intel.com