Repository navigation
[NOT FOR UPSTREAM] drm/xe/lnl: Drop force_probe requirement - #5110
Conversation
… GPU bind" This reverts commit 0f49dde. Signed-off-by: Kai Vehmanen <kai.vehmanen@linux.intel.com>
The driver is ready for testing with SOF for display audio use-cases, so force_probe requirement can be dropped. Signed-off-by: Kai Vehmanen <kai.vehmanen@linux.intel.com>
@plbossart I have checked all are updated |
|
SOFCI TEST |
|
Seems we are hitting a Xe driver load error on one system, so need more testing before this can be merged. https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/1631 is similar (same error message), but not the same use-case. |
|
SOFCI TEST |
|
We may have inconsistent GuC versions across devices - we still haven't productized Ansible :-( |
|
The results look ok now. There's one fail on https://sof-ci.01.org/linuxpr/PR5110/build4324/devicetest/index.html?model=LNLM_RVP_HDA&testcase=check-sof-logger I can't explain (no dmesg), but rest of the tests are good so appears Xe works on that DUT and we have SOF probing. So @ssavati and @marc-hb if you don't object, I'd proceed with merging this as handle any DUT specific issues as we find them. |
|
Fine by me since this does not add more red that would hide actual audio issues. |
|
SOFCI TEST (re-running because some devices were unavailable) |
|
Very strange, this PR seems to break... https://sof-ci.01.org/linuxpr/PR5110/build4326/devicetest/index.html?model=CML_HEL_RT5682-ipc3&testcase=check-sof-logger Today's daily build 44658 did NOT fail and this test either rarely or never fails in that configuration. This recent PR did not fail either: https://sof-ci.01.org/linuxpr/PR5140/build4314/devicetest/index.html |
|
SOFCI TEST EDIT: https://sof-ci.01.org/linuxpr/PR5110/build4338/devicetest/index.html looks like it's passing now. The sof-logger has never been 100% reliable, so maybe it was just incredibly bad luck that it failed twice back to back in the same configuration...? |
| static const struct xe_device_desc lnl_desc = { | ||
| PLATFORM(XE_LUNARLAKE), | ||
| .has_display = true, | ||
| .require_force_probe = true, |
There was a problem hiding this comment.
Don't we have similar patch in linux-next to sherry-pick?
There was a problem hiding this comment.
Let me check, when I submitted this (one month ago), there wasn't anything in linx-next yet. Btw, sherry-picking sounds fun :)
There was a problem hiding this comment.
@ujfalusi nope, not yet https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/tree/drivers/gpu/drm/xe/xe_pci.c#n338 ... still requires force_probe as of linux-next 20240816
|
It's Friday night now but Monday 18th would be a great time to merge this because the test results are stable and generally good except for the two known alsabat and suspend/resume issues on LNL. |
fredoh9
left a comment
fredoh9
left a comment
There was a problem hiding this comment.
I'm able to test legacy HDA on LNL with this PR.
Without this, legacy HDA probe didn't work.
|
@kv2019i, so, we declare xe stable even if the display guys are not comfortable with it? ;) |
|
Good news: the sky didn't fall. Same LNL failures as usual into today's daily run 44893 A big caveat though: we disabled |
Enable the Xe driver on all test devices and enable gpu_bind in SOF.