08:34 <krh> Good Isa spec discussion

14:08 <alyssa> manhattan seems to be triggering tons of resource shadowing flushes ... that seems not great

14:08 <alyssa> and staging<-->AFBC blits

14:10 <alyssa> things are way faster if I skip the shadowing flushes, results visually seem ok but that of course doesn't mean much

14:11 <alyssa> 23fps->25fps overall, that's substantially of course

14:15 <alyssa> curiously deqp-gles2 passes without the flush

17:50 <srslypascal> Hi, on 2 of my machines, I'm seeing a very strange but reproducible crash pattern with kernel 5.18.x built with the "linux-hardened" patchset that's currently maintained by Levente Polyak. I assume that the hardened patchset does not *cause* but merely trigger the bug due to it acting a bit more paranoid with memory errors than a vanilla kernel. The crashes seem to be related to both the machine having an AMD GPU (using the amdgpu

17:50 <srslypascal> module) and the machine loading/using the snd_hda_intel module. I've got two machines that the same Dell model, but one of them has an NVidia GPU and the other has an AMD GPU, and the crash occurs only on the machine with the AMD GPU. Also, the crash does not occur when I either boot with "modprobe.blacklist=snd_hda_intel" or when I boot with "snd_hda_intel.snoop=1". Interestingly, booting with "modprobe.blacklist=amdgpu" does *not*

17:50 <srslypascal> prevent the crashes. I'm currently trying to narrow it down to a specific patch, but bisecting between kernel version branches really sucks since I'd have to adapt the hardening patchset for each bisect step… :/

17:54 <srslypascal> I have already tried reverting commits 69458e2c27800da7697c87ed908b65323ef3f3bd, 6317f7449348a897483a2b4841f7a9190745c81b and acd289e04a0a1f52bea7ff1129b365626059e3c2 but that didn't help. I'm currently trying to build a kernel with commits fefee95488412796b293d28c948be6fce63d149b, 327e8ba54a212f707a68670c9372747b7a32bb92, c9db8a30d9f091aa571b5fb7c3f434cde107b02c and 00fd7cfad0548b6b7234c93370076f9b9c2e39f8 reverted. Any hints/guesses

17:54 <srslypascal> what to try if this doesn't help either?

18:05 <srslypascal> I've also made screenshots of the error messages when the crashes occur (the two screenshots from today were made with "loglevel=7" and "snd_hda_intel.single_cmd=1"):

18:05 <srslypascal> https://remotehost.online/snd_hda_intel/snd_hda_intel-2022-05-30-182702.jpg

18:05 <srslypascal> https://remotehost.online/snd_hda_intel/snd_hda_intel-2022-06-04-184242.jpg

18:05 <srslypascal> https://remotehost.online/snd_hda_intel/snd_hda_intel-2022-06-04-184320.jpg

18:17 <srslypascal> The above pictures are from my non-Dell machine where the crash occurs. The crashes do not occur with 5.17.x kernels.

18:18 <srslypascal> (*5.17.x kernels with the linux-hardened patchset applied)

19:08 <srslypascal> reverting those 4 patches didn't help either :(

19:11 <srslypascal> oh and booting with "amdgpu.audio=0" didn't help either (though that was to be expected if blacklisting amdgpu didn't help either)