06:40airlied: marysaka[d]: does dEQP-VK.binding_model.shader_access.secondary_cmd_buf.bind2.with_template.storage_buffer_dynamic.tess_ctrl.multiple_discontiguous_descriptor_sets.multiple_arbitrary_descriptors.offset_view_nonzero_dynamic_zero give a fault for you?
07:08marysaka[d]: airlied: There was some MR recently to fix those tests recently but will test in a bit https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39203
07:09marysaka[d]: if you are targeting the branchpoint, it's missed it by not much and there is a backport MR from mel here https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39440
07:30airlied: marysaka[d]: I was on the branch for you mmu test
07:30airlied: just rebasing
07:31airlied: marysaka[d]: https://paste.centos.org/view/502cf8bd is my current fix for lpt stuff
07:32marysaka[d]: hmm is that with the patch I sent yesterday?
07:32marysaka[d]: oh hmm you actually guard the invalid I guess
07:33airlied: no it's just a new patch
07:34marysaka[d]: I see will give that a try
07:36marysaka[d]: but yeah no fail on that specific test with my mmu test branch rebased on main and 25.2.7
07:36marysaka[d]: that's without your patch (bare drm-misc-next from some days ago)
07:42airlied[d]: Rebasing mesa fixed it, just started cts again
07:42marysaka[d]: okay
07:42marysaka[d]: I *thinik* that we should also change the sparse path in `nvkm_vmm_unref_sptes`
07:44marysaka[d]: actually no it's another path that is taken I forgot that
07:47marysaka[d]: Will start some CTS with it too and go make my morning tea 😄
08:23marysaka[d]: CTS is okay, will do some game testing but yeah I think the call to `sparse` should be guarded like the call to `invalid`
08:25marysaka[d]: otherwise this looks good to me
08:28phomes_[d]: I am game and CTS testing too. All good so far
08:45airlied[d]: I think sparse is its own path
08:53marysaka[d]: airlied[d]: `nvkm_vmm_sparse_ref_ptes` do call `nvkm_vmm_ref_ptes`, same for `nvkm_vmm_sparse_unref_ptes` calling `nvkm_vmm_unref_ptes` but maybe I'm missing some detail around it that would make it not needed?
09:59airlied: marysaka[d]: I think the thing is the LPT entry can only be in SPARSE or BIG_VALID
10:04marysaka[d]: I think you are right okay all good then
10:04marysaka[d]: thanks a lot for the help 😄
20:20jja2000[d]: Hmmmm something happened to this discord server and I'm not sure what
20:21HdkR: New haircut :nodders:
20:31airlied[d]: marysaka[d]: patch posted to mailing list
21:14mhenning[d]: mangodev[d]: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39464 fixes a chromium issue, although I'm not sure if it's the one you were running into or not
21:20mangodev[d]: mhenning[d]: sadly no
21:21mangodev[d]: i also still get flickering in Firefox, even after the zink fixes
21:21mhenning[d]: yeah, I can still reproduce the firefox issue
21:21mangodev[d]: in fact, earlier today, i got a zink crash in kwin itself
21:21mangodev[d]: but the logs don't say much
21:21mangodev[d]: just says that zink lost vk_device
21:21mangodev[d]: but there was no MMU fauly
21:21mangodev[d]: *fault
21:21mhenning[d]: nothing in dmesg?
21:21mangodev[d]: nope
21:22mangodev[d]: i had to restart my PC immediately after that, since it wouldn't recover
21:22mangodev[d]: happened after yesterday's patches
21:22mangodev[d]: maybe kwin just
21:22mangodev[d]: didn't like it
21:23mangodev[d]: i don't expect my system to be bug free until i reinstall though
21:23mangodev[d]: i think something's screwy with my install
21:23mangodev[d]: i just don't have the spare time to screw around with that until may or so
21:23mangodev[d]: until then, i deal with the bugs
21:25mangodev[d]: i think part of it is that my kwin is messed up beyond repair
21:25mangodev[d]: or really just anything KDE as a whole
21:26mangodev[d]: like the emoji selector being nothing but errors
21:26mangodev[d]: or random lag during games that didn't fix even when i switched back to proprietary
21:27mangodev[d]: i think my install just got lobotomized some time ago
22:25airlied[d]: marysaka[d]: can you reproduce mhenning[d] problem? maybe there is another fun path
22:26airlied[d]: I can't get the initial fault on a turing laptop or blackwell laptop
22:40marysaka[d]: I wasn't able to get it earlier will run mel reproducer again over the night and see if it triggers
22:42marysaka[d]: (I have some python script around it that retry after ~1 min if there is no crash as I was only getting one during the first minute of that test group)
23:08phomes_[d]: fwiw I am running Mel's reproducer on ada desktop
23:10marysaka[d]: I was testing on ada too
23:11marysaka[d]: I could put some ampere GPU on the test bench for testing if needed