Linux regressions report for mainline [2024-09-08]

[Thorsten Leemhuis]

Hi Linus! Tried to approach regression tracking a bit differently this
cycle, which is why I didn't send any reports in the past few weeks. But
as we are nearing the end of the cycle, let me quickly compile a report
for afaics unresolved regressions that were reported and bisected to a
change in this or the previous cycle.

I only mentioned those I found noteworthy for one reason or another,
most of the time because there is a lack of progress. But I track more
regression, for example 12 others from this cycle. See
https://linux-regtracking.leemhuis.info/regzbot/mainline/ for
details[1]. But people are working busily on fixing those afaics; and
none of them seem to affect a lot of users. Will send a more detailed
report next week due to the pending release.

Ciao, Thorsten

[1] ignore any pre-6.10 stuff for now, I lacked time to re-check if
there was any progress that was not properly tagged (which is frequent)
and thus missed by regzbot.


mainline regressions introduced during development of 6.11
==========================================================

* Marc Payne 18 days ago reported that btusb fails to initialize on
MT7921AUN since ccfc8948d7e4d9 ("Bluetooth: btusb: mediatek: reset the
controller before downloading the fw") [v6.11-rc1] but did not get even
a reply from a developer yet:
https://lore.kernel.org/lkml/ZsTh7Jyug7MbZsLE@xxxxxxxxxxxx/

* Mikhail Gavrilov five weeks ago reported that launching some RenPy
games causes hangs since 1b04dcca4fb10d ("drm/amd/display: Introduce
overlay cursor mode") [v6.11-rc1]. Fell through the cracks, now being
worked on.
https://lore.kernel.org/lkml/CABXGCsNgx6gQCqBq-L2P15ydaN_66sM9CgGa9GQYNzQsaa6Dkg@xxxxxxxxxxxxxx/

* Chris Li six weeks ago reported swap stress tests now running into OOM
since 33dfe9204f29b4 ("mm/gup: clear the LRU flag of a page before
adding to LRU batch") [v6.11-rc1]. Fell through the cracks, but after
prodding from my side recently posted a patch to fix this:
https://lore.kernel.org/all/CAF8kJuNP5iTj2p07QgHSGOJsiUfYpJ2f4R1Q5-3BN9JiD9W_KA@xxxxxxxxxxxxxx/
https://lore.kernel.org/lkml/20240905-lru-flag-v2-1-8a2d9046c594@xxxxxxxxxx/


mainline regressions introduced during development of 6.10
==========================================================

* Seven weeks ago someone reported that mpu6050 gyroscopes stopped
working; a fix is in -next for 15 days now and hopefully will make it to
you soon:
https://lore.kernel.org/linux-iio/0ea167a1-75d0-469d-a79f-ff2cb5e81bdc@xxxxxxxxx/
0a3b517c8089aa ("iio: imu: inv_mpu6050: fix interrupt status read for
old buggy chips") [next-20240819 (pending-fixes)]

* Two people three and a half weeks ago reported problems (like crashes
on receiving large data over virtio_net under memory and IO load) due to
f9dac92ba90810 ("virtio_ring: enable premapped mode whatever
use_dma_api") [v6.10-rc1]; there will likely be a few reverts heading
your way to fix this:
https://bugzilla.kernel.org/show_bug.cgi?id=219154
https://lore.kernel.org/netdev/8b20cc28-45a9-4643-8e87-ba164a540c0a@xxxxxxxxxx/
https://lore.kernel.org/netdev/20240906123137.108741-1-xuanzhuo@xxxxxxxxxxxxxxxxx/

* Six weeks ago someone reported massive stuttering when GPU is almost
100% load since b7a1a0ef12b819 ("drm/amd/amdgpu: add pipe1 hardware
support") [v6.10-rc1]; four weeks ago a partial revert was discussed to
fix the issue, but not merged -- and from here it looks like the issue
is not yet resolved yet and stalled for three weeks now:
https://gitlab.freedesktop.org/drm/amd/-/issues/3519

* Since about six seeks it's known that 4df96ba6676034
("drm/amd/display: Add timing pixel encoding for mst mode validation")
[v6.10-rc1] caused issues for some users with 4k displays connected via
a dock; no fix in sight yet afaics. :-/
https://lore.kernel.org/amd-gfx/872a3e83-9bc5-47de-8514-48abe8ff03c5@xxxxxxxxxxxx/
https://lore.kernel.org/stable/d74a7768e957e6ce88c27a5bece0c64dff132e24@xxxxxxxx/
https://gitlab.freedesktop.org/drm/amd/-/issues/3513

* Since about four weeks it's known that 6f31d6b643a32c ("igc: Refactor
runtime power management flow") [v6.10-rc1] causes suspend to fail on
the second attempt; at least two people seem to be affected; progress is
slow, no fix in sight, revert seems to fix this (it's not totally clear):
https://bugzilla.kernel.org/show_bug.cgi?id=219143