Report for regressions introduced during this devel cycle as of 2026-04-01

[Thorsten Leemhuis]

Hi Linus! I'm more active in tracking regressions again. And sorry, I
wanted to write this mail last week before -rc6, as I know you want all
known regressions fixed by then -- but real life got in between, Sorry.

#####################################################################
regressions introduced known for some time now which could have been
fixed before -rc6 came out (through a proper fix or a revert)
#####################################################################

------------------------------------------------------------------
  drm/amdgpu: no audio anymore
known for 4 weeks, caused by b82f0759346617 (v7.0-rc1)
https://lore.kernel.org/all/CAOKSTBs1VNBdjRZe3M2v-nC%2BpLZZ0-
KQtRUZUj6ACqb4Xe7LDg@xxxxxxxxxxxxxx/
https://gitlab.freedesktop.org/drm/amd/-/issues/5052

  Fix is "drm/amd/display: Wire up dcn10_dio_construct() for all pre-
DCN401 generations" at https://lore.kernel.org/all/20260323211343.263909-1-
sunlightlinux@xxxxxxxxx/

  According to replies the fix is now "in the next week's display driver
promotion cycle before merging", so it doesn't sounds like this will
make -rc7 (and maybe not ever -rc8/final). Not really ideal how long
this all takes after at least three people ran into this; but to be
fair, it was only bisected about 10 days ago.

------------------------------------------------------------------
  ARM: dts: imx: Colibri iMX6ULL and Colibri iMX7 no longer boot
https://lore.kernel.org/lkml/20260219151157.2549198-1-max.oss.09@xxxxxxxxx/
known for 6 weeks, caused by 8124b4a4a96b (v7.0-rc1)

  Fix is bccaf98c1077f2 ("Revert "ARM: dts: imx: move nand related
property under nand@0"") [next-20260318 (pending-fixes)] aka https://
lore.kernel.org/lkml/20260219151157.2549198-1-max.oss.09@xxxxxxxxx/

  Krzysztof already urged to apply the revert on Mar 5th and it was
merged soon after – but since then is only in -next afaics (at least in
a branch that is supposed to got to mainline in this cycle).

------------------------------------------------------------------
  PCI: pci=realloc fails, needed to get a Nvidia P40 24GB running
https://bugzilla.kernel.org/show_bug.cgi?id=221205
known for 22 days, caused by 3958bf16e2fe (v7.0-rc1 as well as v6.19.4,
v6.18.14)

  Fix is 9036bd0efcb616 ("PCI: Align head space better") [next-20260331,
afaics targeted for 7.1], which is 09/10 of a series:
https://lore.kernel.org/all/20260324165633.4583-10-ilpo.jarvinen@xxxxxxxxxxxxxxx/

  Another regression ("amdgpu with Thunderbolt eGPU bracket fails")
caused by the commit was reported a few days ago, investigation is ongoing:
https://lore.kernel.org/all/a5f23340-2b84-4734-be11-
f5a97c188195@xxxxxxxxxxxxxxxx/

------------------------------------------------------------------
  drm/msm/dpu: pink noise on the screen
https://lore.kernel.org/lkml/DH1IKLU0YZYU.2SW4WYO7H3H4R@xxxxxxxxxxxxx/
known for 19 days, caused by 2892de3f4f98 (v7.0-rc1)

  Fix is b9699dd862760e ("drm/msm/dpu: don't try using 2 LMs if only one
DSC is available") [next-20260325] aka
https://lore.kernel.org/lkml/20260317-fix-3d-dsc-
v1-1-88b54f62f659@xxxxxxxxxxxxxxxx/

------------------------------------------------------------------
  ceph: writeback can lead to a panic
https://lore.kernel.org/lkml/20260315232500.251088-1-CFSworks@xxxxxxxxx/
known for 16 days, caused by cac190c7674f (v7.0-rc1)

  Fix is 681a6d350eff10 ("ceph: fix num_ops off-by-one when crypto
allocation fails") [next-20260320] aka
https://lore.kernel.org/all/20260318023733.116789-1-CFSworks@xxxxxxxxx/

------------------------------------------------------------------
  powerpc/iommu: amdgpu driver fails to initialize
https://lore.kernel.org/linuxppc-dev/20260313142351.609bc4c3efe1184f64ca5f44@xxxxxxxx/
known for 15 days, caused by 1471c517cf7d (v7.0-rc1)

Can be fixed through a revert which was finally posted yesterday:
https://lore.kernel.org/all/20260331223022.47488-1-gbatra@xxxxxxxxxxxxx/

##################################################################
other issues worth mentioning briefly

* I assume your will get the following recent regression fix by the end
of the week through via the weekly alsa pull:
ALSA: usb-audio: Exclude Scarlett Solo 1st Gen from SKIP_IFACE_SETUP
https://lore.kernel.org/all/20260329170420.4122-1-dag@xxxxxxxxxxxx/
 Which would be great, as the culprit made it to v6.19.7 and v6.18.17,
too: https://lore.kernel.org/all/abcGli7C40kr72mB@xxxxxxx/

* I assume your will get the following recent regression fix by the end
of the week:
HID: gpd: fix report descriptor on GPD Win handheld (2f24:0137)
https://lore.kernel.org/all/20260324013847.68024-1-honjow311@xxxxxxxxx/
  To me it looks like users have to enable a new config option to avoid
a regression now, as I mentioned in the thread. No reply yet -- and
maybe I'm missing something here...

* Note sure, but I hope you will get this fix by the end of the week::
vt: discard stale unicode buffer on alt screen exit after resize
https://lore.kernel.org/all/20260327170204.29706-1-liavmordouch@xxxxxxxxx/
  Would be good, as the problem affects 6.19.10, too:
https://lore.kernel.org/all/20260327160050.31631-1-liavmordouch@xxxxxxxxx/