Hi Salvatore,

Am Mi den 27. Dez 2023 um 21:24 schrieb Salvatore Bonaccorso:

I do realize, but given we have nobody else reporting similar
behaviour we need to rely on you bisecting the breaking change so it
might be reported upstream. But that said, in meanwhile we have
6.6.8-1 uploaded to unstable. It would be great if you can report back
if that version resolves the issue.

I already installed the kernel and test it. As a site note, it does not
work with dkms anymore as the header links are in /usr/lib/modules where
no kernel is able to find them.

However, I will keep an eye on it and see if the bug is fixed (as I did
with the kernels in between.) Give me some time as the bug is not that relyable.

If it's still reproducibe, check first that it's as well reproducible
with an untained kernel because othwerise an upstream report might not
be accepted.

On that system I use untained kernel (I did not on a different system
with nvidia, but there I use my own kernel and the bug is related to amd
GPU.)

Regards
Klaus
--
Klaus Ethgen http://www.ethgen.ch/
pub 4096R/4E20AF1C 2011-05-16 Klaus Ethgen <Klaus@Ethgen.ch> Fingerprint: 85D4 CA42 952C 949B 1753 62B3 79D0 B06F 4E20 AF1C

-----BEGIN PGP SIGNATURE-----
Comment: Charset: ISO-8859-1

iQGzBAABCgAdFiEEMWF28vh4/UMJJLQEpnwKsYAZ9qwFAmWMiZQACgkQpnwKsYAZ 9qzl0Qv/bdmazKnvPruMUf7+elXoRgsrRexq+psTz84L/t5U4ej0PisWavjlFojL rK4Ww6pZ6PtCSU1WXLPEnQ7icsWJuCNmLJB8iatMqoU/wXI84hbwWjd1k5Y2UQH8 AfQkawYhoDPuZIDT+99yQRUdXLo2sBqI64/UXKpakllnZ56I64pulJzRbzsHGYVo I3kp9TLt4hK/G6px7o36/LNU4c5lSxvFRDpyi/D1ZWMOHD3ewPuSf1QA2vUd4XLB mR0e7I+gvsjNA70oc6li8gK8ZJjC4//32zkOsw7n9i17QRXue5w8O9UtTTv0S1W6 n2KQwJuMpYqN3/STwCwi/9F0qfN8FsIyTws7Y/ctedmRzMZjJJeQqeiKws+NOsJK 99I58DK5qz/h5Xx5w1zbkFTS+NOj5pj6SN+ur0eU0fv5O9fW1/CP9BrPNjiwH0T7 588ZMYvhew+lTqgm2LEQkdEPY7RNgmagbuVuoGJoUWJQW4OGL6FOc8gZWNGwQB6W
O2DoXnkB
=c5V+
-----END PGP SIGNATURE-----

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Hi Klaus,

On Sat, Oct 21, 2023 at 08:34:55AM +0100, Klaus Ethgen wrote:

Hi,

Am Do den 19. Okt 2023 um 20:46 schrieb Salvatore Bonaccorso:

On Thu, Oct 12, 2023 at 06:57:20AM +0100, Klaus Ethgen wrote:

Package: src:linux
Version: 6.5.6-1
Severity: critical
Tags: security
X-Debbugs-Cc: Debian Security Team <team@security.debian.org>

It is not fully clear for me, where exactly this bug happens. First I
was thinking about xscreensaver but that package got not updated for ages. The bug happens with updates from kernel 6.4.0 to 6.5.0.

So you are saying this happens solely after switching from 6.4.y
series to 6.5.y series. Thus I assume 6.5.3-1 in testing as well
exposes the issue.

Might be but I cannot test that due to the other AMD display related
bug.

I use xscreensaver with fvwm3 on my amd laptop. xscreensaver is set up
to only blank the screen.

I first thought, that it does not happen with fvwm2 but I also see it
with fvwm2 but not that often.

When I lock the screen and press a key or moving the mouse, everything
is fine. But when I go to suspend too ram after locking and waking up
the laptop, the password dialog gets showed as usual but I can see the full desktop content with probably sensitive material on in. Although, I cannot interact with the desktop, it is a security break to reveal the content without authenticating.

It might be related, when I have a PSI chat window on the screen but on different desktop, it gets moved to the current one. That definitively also came with the new kernel.

Can you please attach as well the kernel log once you triggered the behaviour? Anything suspicious logged?

I could. But there is no hint and no unusual log entry.

Next, can you bisect the kernel between a good known upstream version
and 6.5.6? Can you as well test 6.5.7 upstream to see if it fixes the issue?

That would take many time to recompile kernel, test it for several hours
and try again.

I do realize, but given we have nobody else reporting similar
behaviour we need to rely on you bisecting the breaking change so it
might be reported upstream. But that said, in meanwhile we have
6.6.8-1 uploaded to unstable. It would be great if you can report back
if that version resolves the issue.

If it's still reproducibe, check first that it's as well reproducible
with an untained kernel because othwerise an upstream report might not
be accepted.

Can you check that?

Regards,
Salvatore

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Hi Salvatore,

Thanks for helping keeping the bug alive.

Am Mi den 27. Dez 2023 um 21:24 schrieb Salvatore Bonaccorso:

That would take many time to recompile kernel, test it for several hours and try again.

I do realize, but given we have nobody else reporting similar
behaviour we need to rely on you bisecting the breaking change so it
might be reported upstream. But that said, in meanwhile we have
6.6.8-1 uploaded to unstable. It would be great if you can report back
if that version resolves the issue.

Unfortunately it is not fixed. But the bug did show up after longer
sleep only.

If it's still reproducibe, check first that it's as well reproducible
with an untained kernel because othwerise an upstream report might not
be accepted.

Can you check that?

It is reproducible with an untainted kernel but not easily. I have no
way to reliable trigger the bug but it happens always after longer
sleeps.

With the new (6.6.8) kernel I see also two further "bugs":
- Drawing on the screen sometimes create artefacts for lines that stay
on the screen for some time. Maybe that is related as I think it
happens on AMDGPU...

- The second bug is not related. I connect a logitech MX Anywhere 3 via
bluetooth. With that the rate of the mouse wheel did switch to a very
unusable low rate. I can switch the wheel to high speed but when I go
back to the working kernel, that is unusable fast. Also with the
broken kernel, even with high speed, it delay the mouse wheel
somewhat. As there are many stacks in between, I have no idea, which
part of the stack produces that error. Maybe you might help?

Regards
Klaus
--
Klaus Ethgen http://www.ethgen.ch/
pub 4096R/4E20AF1C 2011-05-16 Klaus Ethgen <Klaus@Ethgen.ch> Fingerprint: 85D4 CA42 952C 949B 1753 62B3 79D0 B06F 4E20 AF1C

-----BEGIN PGP SIGNATURE-----
Comment: Charset: ISO-8859-1

iQGzBAABCgAdFiEEMWF28vh4/UMJJLQEpnwKsYAZ9qwFAmWSmLYACgkQpnwKsYAZ 9qyknwwAv7qJel+CHz/Ww27276PzPGllrX242I1paCexEzmSX9o84eVeOK6eVN5q grD11Gdt1Gtd57CixGLztUcEaOzNVUwtdaS6zm9kdZLde9QLk9Q1AjZ8m6YA7LV0 Ofs25efdqQ/VRyK5xMzpy26fY5Gw0JCkRriJ05YOIiwRWBvZopc8yzaYdJX/eDG+ Uo5uGKB7diurQ5lEBsJWBt49u2b+YhZKy9xe+ZPVTqNUinDZ9GedKu+WMlI9GSt9 wwaBtDZZfqv+tQ+X0ubr7aZ1sFKx9amZN1zv/TiS9+Rc0fodlRyAPhyzoV4LFrXw om9XGpBa4oFPVPFUAyL78yoJ70znbLHcB3YhvjQvCpf3lICui5K4W/BdCir6spXZ pAnZBQS7XkuSxxEMy4YspwmnuiHBMvNx+nrmxsNU4gCim9vokihG+hW7r7r4S+Vh nq50MyByO8BJ5lcRGo+mtl7vaCWTrZCQ7pim4myML3yOy2hf5qfjByG+sUEPcdSR
JY4t1EDf
=r3YS
-----END PGP SIGNATURE-----

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Control: severity -1 normal

Hi Klaus

On Thu, Oct 12, 2023 at 06:57:20AM +0100, Klaus Ethgen wrote:

-- System Information:
Debian Release: trixie/sid
merged-usr: no

I just realized that this system is in an unsupported state. Bookworm
and later is not longer supported without merged-/usr, see https://www.debian.org/releases/stable/amd64/release-notes/ch-information.en.html#a-merged-usr-is-now-required

Please reinstall from scratch and report back if it is still broken.

Maybe please also describe how you got into this state, where /lib is
not a symlink to /usr/lib.

Bastian

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)