Aw: Re: BUG: MTK DRM/HDMI broken on 5.13 (mt7623/bpi-r2)

[Frank Wunderlich <frank-w@public-files.de>]

> Gesendet: Donnerstag, 08. Juli 2021 um 09:22 Uhr
> Von: "Dafna Hirschfeld" <dafna.hirschfeld@collabora.com>
> We also encountered that warning on mt8173 device - Acer Chromebook R13. It happen after resuming from suspend to ram.
> We could not find a version that works and we were not able to find the fix of the bug.
> It seems like the irq isr is not called after resuming from suspend.
> Please share if you have new findings regarding that bug.

Hi,

i have not yet found a way to make the commit-history flat for running bisect without the issue of disappearing childcommits when mergecommit is out of bisect scope. so i tried to start at working 5.12.0 with mtk-drm-patches and commits from drm core (i hope i have catched them all) by cherry-picking the single commits.

c24e104c26aa 2021-06-09 drm: Lock pointer access in drm_master_release()  (HEAD -> 5.12-drm)
2aa9212803a4 2021-06-08 drm: Fix use-after-free read in drm_getunique()
23b8d6c3be47 2021-04-08 treewide: Change list_sort to use const pointers
c1e987f51f06 2021-03-26 drm/dp_mst: Drop DRM_ERROR() on kzalloc() fail in drm_dp_mst_handle_up_req()
2176a9e962be 2021-04-01 drm/drm_internal.h: Remove repeated struct declaration
fc5d92c1485d 2021-04-08 drm/syncobj: use newly allocated stub fences
23a03d271e87 2021-03-29 drm/displayid: rename displayid_hdr to displayid_header
44ef605cb08f 2021-03-29 drm/displayid: allow data blocks with 0 payload length
bbdc0aefd1b5 2021-03-29 drm/edid: use the new displayid iterator for tile info
1ee4a22d671e 2021-03-29 drm/edid: use the new displayid iterator for finding CEA extension
d9b8c26b8ddf 2021-03-29 drm/edid: use the new displayid iterator for detailed modes
d9e95df8adc8 2021-03-29 drm/displayid: add new displayid section/block iterators
2dd279949358 2021-03-29 drm/displayid: add separate drm_displayid.c
bb1a3611abc1 2021-03-29 drm/edid: make a number of functions, parameters and variables const
0b18f5b98c71 2021-03-23 drm/dp_helper: Define options for FRL training for HDMI2.1 PCON
16fbc25ab84b 2021-03-25 drm/mst: Enhance MST topology logging
bb93ad6ab4e4 2021-03-26 drm: Fix 3 typos in the inline doc
27d30189b178 2021-03-22 drm/sysfs: Convert sysfs sprintf/snprintf family to sysfs_emit
04ad4ed36cf2 2021-03-18 drm: Few typo fixes
b8821cac052f 2021-03-13 drm: Add GUD USB Display driver
d3df1b84b9ff 2021-03-13 drm/probe-helper: Check epoch counter in output_poll_execute()
298372a0cda4 2021-03-13 drm/uapi: Add USB connector type
040c9022809d 2021-03-30 drm/mediatek: Don't support hdmi connector creation
7c6582b23551 2021-03-30 drm/mediatek: Switch the hdmi bridge ops to the atomic versions
b1b43d5948b2 2021-02-03 drm/mediatek: Add missing MODULE_DEVICE_TABLE()
fe5a0ff82cfb 2021-03-13 drm/mediatek: crtc: Make config-updating atomic

result: it is still working. so at least they do not break ;)

have you found any irq-related message in dmesg (i have not found any irq-error/warning-message)?
how have you traced that?

can somebody point us to the interrupts used for pageflip/vblank "requests"? in the wait-chain i do not see them,
it seems it is called asynchronous and wait only looks at a state in the completion-struct

i have the issue on bootup, i see only a purple screen instead of fbcon/xserver and the tracebacks
on serial are very annoying  as they repeating every x seconds (maybe change to WARN_ONCE?).
But after a while it seems to stop.

imho we need a way to make the history (temporary) flat (remove parent-information from commits to merge) so that bisect have only a list and not a "tree"

regards Frank