#
…

2023-07-29

00:00mikelee has quit [Ping timeout: 480 seconds]
00:09mikelee has joined #asahi-gpu
00:13ourdumbfuture has joined #asahi-gpu
00:20mikelee has quit [Ping timeout: 480 seconds]
00:23ourdumbfuture has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
00:24ourdumbfuture has joined #asahi-gpu
00:43mikelee has joined #asahi-gpu
00:52mikelee has quit [Ping timeout: 480 seconds]
00:57mikelee has joined #asahi-gpu
00:57loki_val has quit []
00:57nafod has quit [Ping timeout: 480 seconds]
01:05mikelee has quit [Ping timeout: 480 seconds]
01:08mikelee has joined #asahi-gpu
01:18mikelee has quit [Ping timeout: 480 seconds]
01:20Emantor has quit [Quit: ZNC - http://znc.in]
01:21Emantor has joined #asahi-gpu
01:36ourdumbfuture has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
01:38ourdumbfuture has joined #asahi-gpu
01:46mikelee has joined #asahi-gpu
02:12mikelee has quit [Ping timeout: 480 seconds]
02:20mikelee has joined #asahi-gpu
02:28mikelee has quit [Ping timeout: 480 seconds]
02:32ourdumbfuture has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
02:32linaLooks like GNOME 44 / XWayland broke window modifiers or something like that...
02:38mikelee has joined #asahi-gpu
02:47mikelee has quit [Ping timeout: 480 seconds]
03:03WindowPain has joined #asahi-gpu
03:03WindowPa- has quit [Read error: Connection reset by peer]
03:13mikelee has joined #asahi-gpu
03:16Z750 has quit [Quit: Ping timeout (120 seconds)]
03:17Z750 has joined #asahi-gpu
03:24mikelee has quit [Ping timeout: 480 seconds]
03:26Z750 has quit [Quit: Ping timeout (120 seconds)]
03:26Z750 has joined #asahi-gpu
03:35mikelee has joined #asahi-gpu
03:36Guest7156 has quit [Quit: Bridge terminating on SIGTERM]
03:36rhysmdnz has quit [Quit: Bridge terminating on SIGTERM]
03:40rhysmdnz has joined #asahi-gpu
03:40Jamie has joined #asahi-gpu
03:40Jamie is now known as Guest7281
03:44mikelee has quit [Ping timeout: 480 seconds]
03:55mikelee has joined #asahi-gpu
04:04mikelee has quit [Ping timeout: 480 seconds]
04:19mikelee has joined #asahi-gpu
04:24Retr0id^ if I wanted to look through commits and speculate about what might've caused it, which repo(s) should I be looking in?
04:30jannauI would start with mutter (gnome's compositor) if you're reasonably sure that the gnome update broke it
04:31jannauXWayland has probably a lot less commits and there was iirc a release recently
04:33mikelee has quit [Ping timeout: 480 seconds]
04:37jannauRetr0id: can you try to downgrade to Xwayland 23.1.1
04:40Retr0idjust downgraded to 23.1.1, no change in symptoms
04:41Retr0id(I had xorg-xwayland-23.1.1-1-aarch64.pkg.tar.xz in my local cache)
05:02jannauXwayland is even more broken for me under gnome. On login I get a broken KGpg window but after that some X11 applications are completly broken, for example glxgears
05:10mikelee has joined #asahi-gpu
05:16linaThat's the modifiers issue
05:17linaApps are rendering compressed and mutter is compositing them as linear
05:17linaIt means somewhere along the line the modifier gets lost or set to zero
05:38WindowPa- has joined #asahi-gpu
05:43WindowPain has quit [Ping timeout: 480 seconds]
05:57mikelee has quit [Ping timeout: 480 seconds]
06:15stipa has quit [Remote host closed the connection]
06:25jannauglxgears seems to kill xwayland here and gnome doesn't restart it
06:25linaI think it might be a Mesa regression somewhere, it worked with some random older branch
07:00linaYeah, it's a core Mesa regression. I have a bad and good commit now (with trivial Asahi changes on top), going to bisect
07:11jannaufirefox crashes are agx related. finally one after enabling symbols:: SIGSEV in agx_bo_unreference() ../mesa-asahi-20230725/src/asahi/lib/agx_bo.c:190
07:11linaUmmmmmmmmmm
07:11linaI bisected the xwayland thing to 1d9d72b3fb905e
07:11linaWhich both has nothing to do with its own commit description (!?) and doesn't have anything to do with modifiers
07:11linaNow I'm really confused...
07:12linaBut yes, reverting it fixes xwayland on mutter
07:12linaI'm going to get lunch, but this is weird...
07:13jannauapaprently an agx_bo with dev == NULL
07:18mikelee has joined #asahi-gpu
07:21jannauthe bo looks corrupt
07:30linaDo you have a way to reproduce it?
07:30linaThat looks like a UAF...
07:32mikelee has quit [Ping timeout: 480 seconds]
07:36lina(Still feeling a bit sick so I'm going to lie down for a bit)
07:37jannaunot a good one, use firefox for hours/days. seemed to occur less often after updating asahi-20230725, asahi changes in asahi-6.4-3 might be related, see also discussion in #asahi-alt over the last copule of days
07:38jannauwhen you feel better. take care
07:56jannaumps claims the crashes are not reproducible with asahi-6.4-1 but asahi-6.4-2 with the same mesa version which seems implausible since that's just dart and dcp changes
08:37crabbedhaloablut has joined #asahi-gpu
09:17LinuxM1 has joined #asahi-gpu
09:17mikelee has joined #asahi-gpu
09:23LinuxM1 has quit [Quit: Leaving]
09:28mikelee has quit [Ping timeout: 480 seconds]
09:37mikelee has joined #asahi-gpu
10:14mikelee has quit [Ping timeout: 480 seconds]
10:16linaYeah that makes no sense
10:16linaAlso kernel changes cannot cause that kind of userspace issue
10:19jannauit's a little suspicious though that it has only be reported in #asahi-alt with asahi-6.4* and not by asahilinux users who are still on asahi-6.3*
10:19mikelee has joined #asahi-gpu
10:41mikelee has quit [Read error: Connection timed out]
10:46ourdumbfuture has joined #asahi-gpu
10:47ourdumbfuture has quit []
10:50mikelee has joined #asahi-gpu
10:59mikelee has quit [Ping timeout: 480 seconds]
11:07ourdumbfuture has joined #asahi-gpu
11:07mikelee has joined #asahi-gpu
11:08ourdumbfuture has quit []
11:12ourdumbfuture has joined #asahi-gpu
11:16mikelee has quit [Ping timeout: 480 seconds]
11:19mikelee has joined #asahi-gpu
11:29mikelee has quit [Ping timeout: 480 seconds]
11:32Retr0idfwiw I spend most of my waking hours inside of firefox and I've had no issues on 6.3
11:38mikelee has joined #asahi-gpu
11:40nsklaus has joined #asahi-gpu
11:46mikelee has quit [Ping timeout: 480 seconds]
11:55mikelee has joined #asahi-gpu
12:04mikelee has quit [Ping timeout: 480 seconds]
12:10Z750 has quit [Quit: Ping timeout (120 seconds)]
12:10Z750 has joined #asahi-gpu
12:11mikelee has joined #asahi-gpu
12:20mikelee has quit [Ping timeout: 480 seconds]
12:29mikelee has joined #asahi-gpu
12:38mikelee has quit [Ping timeout: 480 seconds]
12:39mikelee has joined #asahi-gpu
12:45ave36309 has quit []
12:47ave36309 has joined #asahi-gpu
12:49mikelee has quit [Ping timeout: 480 seconds]
12:58mikelee has joined #asahi-gpu
13:07mikelee has quit [Ping timeout: 480 seconds]
13:12ourdumbfuture has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
13:16mikelee has joined #asahi-gpu
13:25ourdumbfuture has joined #asahi-gpu
13:31mikelee has quit [Ping timeout: 480 seconds]
13:44ourdumbfuture has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
13:44alyssaclearly we need to rewrite mesa in rust
13:46alyssai agree with lina, that backtrace looks like a uaf
13:48ourdumbfuture has joined #asahi-gpu
13:52jannauwhat's the best way to track that? compile mesa with asan? running firefox under valgrind has too much noise
13:53alyssahonestly IDK
13:54jannauso rewriting mesa in rust it is ;)
14:04ella-0it was a mistake for me to start agxv in C
14:10mikelee has joined #asahi-gpu
14:15ourdumbfuture has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
14:18mikelee has quit [Ping timeout: 480 seconds]
14:37mikelee has joined #asahi-gpu
14:37possiblemeatball has joined #asahi-gpu
14:45mikelee has quit [Ping timeout: 480 seconds]
14:47mikelee has joined #asahi-gpu
15:18mikelee has quit [Ping timeout: 480 seconds]
15:22mikelee has joined #asahi-gpu
15:37mikelee has quit [Ping timeout: 480 seconds]
15:47mikelee has joined #asahi-gpu
15:50ourdumbfuture has joined #asahi-gpu
15:55mikelee has quit [Ping timeout: 480 seconds]
16:09mikelee has joined #asahi-gpu
16:22mikelee has quit [Ping timeout: 480 seconds]
16:25mikelee has joined #asahi-gpu
16:41mikelee has quit [Ping timeout: 480 seconds]
16:42rhysmdnz has quit [Quit: Bridge terminating on SIGTERM]
16:42Guest7281 has quit [Quit: Bridge terminating on SIGTERM]
16:45rhysmdnz has joined #asahi-gpu
16:46Jamie has joined #asahi-gpu
16:46Jamie is now known as Guest7316
16:50mikelee has joined #asahi-gpu
16:58mikelee has quit [Ping timeout: 480 seconds]
17:01jannauasahi-20230706 has according to asan a stack overflow in agx_link_varyings_vs_fs
17:02jannauI'll update to asahi-20230725 to see if it still reproduces
17:06jannauyes, that's easily fixed, still issues though
17:08mikelee has joined #asahi-gpu
17:10jannauthe code was that way since over a year though
17:14Venemo has quit [Quit: No Ping reply in 180 seconds.]
17:15Venemo has joined #asahi-gpu
17:17jannauevery deqp-egl tests seems to fail under asan
17:18mikelee has quit [Ping timeout: 480 seconds]
17:18jannau"Pass: 0, Fail: 3760, ..." :(
17:19jannaumaybe consistent memleaks
17:20i509vcbI wonder if there is any good reason riir the asahi/lib stuff ?
17:20i509vcbAlthough that's not very large anyways...
17:25jannauyes, looks much better with ASAN_OPTIONS=detect_leaks=0, Pass: 2067 and no fails
17:41jannauany ideas why this didn't cause problems so far?
17:43mikelee has joined #asahi-gpu
17:51mikelee has quit [Ping timeout: 480 seconds]
18:20alyssai509vcb: tbh, the scariest code is the memory manegement, etc
18:20alyssawhich is firmly tied up with common gallium or vulkan stuff
18:20alyssaso the hardest to riir
18:20alyssasomething like src/asahi/layout is very unlikely to have bugs that Rust would help with
18:20alyssait's basically just math
18:21alyssa(otoh it would be totally straightforward to riir since it's just math)
18:22i509vcbUnless you are dealing with a bunch of overflow stuff that something like u32::overflowing_add would help with, I don't see why it would really help. I do recall mesa has some macros to handle that of course
18:24mikelee has joined #asahi-gpu
18:28nafod has joined #asahi-gpu
18:29alyssaI guess panic-on-overflow would be nice for debug builds but otherwise shrug
18:32mikelee has quit [Ping timeout: 480 seconds]
18:37jannauand there is a use after free in dEQP-ES2
18:40mikelee has joined #asahi-gpu
18:48mikelee has quit [Ping timeout: 480 seconds]
18:54mikelee has joined #asahi-gpu
19:02jannauthis is probably easier with mesa internals knowledge: https://gist.github.com/jannau/7a65b5baba0b77c9ff7e924c659666f1
19:03mikelee has quit [Ping timeout: 480 seconds]
19:04jannauI gather the problem is that nothing calls agx_set_constant_buffer(..., NULL); when the program is deleted but I don't see what's supposed to do that in the _mesa_delete_program call path
19:07jeisom has joined #asahi-gpu
19:12mikelee has joined #asahi-gpu
19:22mikelee has quit [Ping timeout: 480 seconds]
19:32possiblemeatball has quit [Quit: Quit]
19:32mikelee has joined #asahi-gpu
19:36possiblemeatball has joined #asahi-gpu
19:46Z750 has quit [Quit: Ping timeout (120 seconds)]
19:47Z750 has joined #asahi-gpu
19:48novafacing has joined #asahi-gpu
19:49ourdumbfuture has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
19:50ourdumbfuture has joined #asahi-gpu
19:51jeisom has quit [Remote host closed the connection]
19:51jeisom has joined #asahi-gpu
19:52hightower2 has joined #asahi-gpu
19:53mikelee has quit [Ping timeout: 480 seconds]
19:53novafacing9 has joined #asahi-gpu
19:57jannautest passes if run standalone, but fails using deqp-runner's case list
20:00novafacing has quit [Ping timeout: 480 seconds]
20:00novafacing9 is now known as novafacing
20:01alyssajannau: that one's spicy
20:02jannaubreaks if dEQP-GLES2.functional.shaders.swizzles.vector_swizzles.mediump_vec4_wz_vertex is before it in the case list
20:02mikelee has joined #asahi-gpu
20:02alyssa> I gather the problem is that nothing calls agx_set_constant_buffer(..., NULL); when the program is deleted
20:03alyssayeah, this looks like potentially a mesa/st bug, leaving a dangling pointer
20:03ourdumbfuture has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
20:04alyssajannau: iris doesn't have this bug, because it uploads the user buf during set_constant_buffer instead of deferring to draw timw
20:05alyssathat is probably more correct anyway, maybe we are fundamentally misusing gallium by keeping the dangling pointer
20:05alyssaradeonsi too
20:05alyssaotoh, v3d does not do that
20:06alyssaso v3d (at least) is probably also affected
20:06alyssaprobably best to bring this up in #dri-devel given multiple affected drivers
20:07alyssajannau: but for asahi, I would say use const_uploader in set_const_buffer to avoid the dangling ptr
20:07alyssaprobably fster too
20:08alyssathanks in advance for the bug fix :3
20:08ourdumbfuture has joined #asahi-gpu
20:09alyssaI wonder if what we really want is a gallium-level helper to upload any user buffers
20:09alyssaso we can fix those across drivers
20:09alyssaif you don't get to it I'll probably do that on monday
20:11mikelee has quit [Ping timeout: 480 seconds]
20:12possiblemeatball has quit [Quit: Quit]
20:17jannaucomment in llvmpipe's set_constant_buffer suggest asahi's implementation is not sound
20:19mikelee has joined #asahi-gpu
20:21alyssaThanks >:
20:21possiblemeatball has joined #asahi-gpu
20:22possiblemeatball has quit []
20:27mikelee has quit [Ping timeout: 480 seconds]
20:39leio has joined #asahi-gpu
20:52mikelee has joined #asahi-gpu
21:05ourdumbfuture has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
21:08mikelee has quit [Ping timeout: 480 seconds]
21:10nsklaus has quit [Ping timeout: 480 seconds]
21:16ourdumbfuture has joined #asahi-gpu
21:16mikelee has joined #asahi-gpu
21:20jannauthat was the only remaining ASAN error in deqp-{egl,gles2,gles3}
21:26mikelee has quit [Ping timeout: 480 seconds]
21:30novafacing has quit [Ping timeout: 480 seconds]
21:35mikelee has joined #asahi-gpu
21:45mikelee has quit [Ping timeout: 480 seconds]
21:47mikelee has joined #asahi-gpu
21:55mikelee has quit [Ping timeout: 480 seconds]
22:15alyssathat's... good? maybe?
22:20mikelee has joined #asahi-gpu
22:28mikelee has quit [Ping timeout: 480 seconds]
22:38mikelee has joined #asahi-gpu
22:45Z750 has quit [Quit: Ping timeout (120 seconds)]
22:46Z750 has joined #asahi-gpu
22:47mikelee has quit [Ping timeout: 480 seconds]
22:57mikelee has joined #asahi-gpu
23:03ourdumbfuture has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
23:05mikelee has quit [Ping timeout: 480 seconds]
23:06crabbedhaloablut has quit []
23:33mikelee has joined #asahi-gpu
23:41mikelee has quit [Ping timeout: 480 seconds]
23:43ourdumbfuture has joined #asahi-gpu
23:50mikelee has joined #asahi-gpu
23:59mikelee has quit [Ping timeout: 480 seconds]
…