#asahi-dev · 2023-06-02 12:25
marcanfd leaks are nothing new#asahi-dev · 2023-06-02 12:25
marcanexpecting a clean fd environment without explicitly cleaning up seems fragile to me, tbh#asahi-dev · 2023-06-02 12:23
marcannot sure who owns the inotify one#asahi-dev · 2023-06-02 12:22
marcanso *shrug*#asahi-dev · 2023-06-02 12:22
marcanthere hasn't been a pulse release since then though#asahi-dev · 2023-06-02 12:20
marcanseems the PA one is already fixed#asahi-dev · 2023-06-02 12:20
marcanhttps://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/merge_requests/751#asahi-dev · 2023-06-02 12:14
marcanif you're so concerned feel free to track it down and fix it ;)#asahi-dev · 2023-06-02 12:13
marcanthe pulseaudio leak is probably the pulseaudio lib's fault#asahi-dev · 2023-06-02 12:13
marcanhttps://bugs.kde.org/show_bug.cgi?id=440990#asahi-dev · 2023-06-02 12:12
marcanthe dmabuf leaks are fixed and could cause a memory leak. the pulseaudio/inotify leaks are probably harmless.#asahi-dev · 2023-06-02 12:04
marcanyes#asahi-dev · 2023-06-02 12:03
marcanI'll have to take a deeper look at this later, too tired to do to go hunting for it#asahi-dev · 2023-06-02 12:02
marcanif you just want it to work#asahi-dev · 2023-06-02 12:02
marcanI bet writing 1 to that on some/all CPUs will fix it#asahi-dev · 2023-06-02 12:02
marcananyway, I recommend playing with /sys/devices/system/cpu/cpu?/cpuidle/state1/disable#asahi-dev · 2023-06-02 12:02
marcanI should also just try it under newer firmware, because it's entirely possible this is an iBoot bug. we're running the first version M2 shipped with after all.#asahi-dev · 2023-06-02 12:01
marcanmaybe I should go back to macos and see if there are any magic pokes we missed...#asahi-dev · 2023-06-02 12:00
marcanon both clusters#asahi-dev · 2023-06-02 12:00
marcanit's like a bunch of CPUs are just dead#asahi-dev · 2023-06-02 12:00
marcanI managed to get a debug shell in the broken state, but I don't know what's up#asahi-dev · 2023-06-02 11:35
marcanstarting to wonder if it's thermal#asahi-dev · 2023-06-02 10:49
marcanokay so definitely something going wrong there#asahi-dev · 2023-06-02 10:48
marcanheey a message#asahi-dev · 2023-06-02 10:48
marcan^CTTY> HV: Failed to rendezvous, missing CPUs: 0x76#asahi-dev · 2023-06-02 10:48
marcanReturning from exception#asahi-dev · 2023-06-02 10:44
marcanso it's definitely the watchdog, and it's not SMC thinking the world is on fire#asahi-dev · 2023-06-02 10:43
marcanno reboot though, it just hangs#asahi-dev · 2023-06-02 10:43
marcanrepro'd under the hypervisor... and the hypervisor died too :(#asahi-dev · 2023-06-02 10:33
marcanand then wdt reboots us#asahi-dev · 2023-06-02 10:33
marcanI think we are deadlocking on something, probably related to cpuidle#asahi-dev · 2023-06-02 10:33
marcancrashed again while plugged in, I don't think it's power delivery#asahi-dev · 2023-06-02 10:28
marcanwatchdog_stop_on_unregister doesn't seem to be working, I unbound the device and it rebooted me#asahi-dev · 2023-06-02 10:26
marcanyeah, so then it's still consistent with wdt timing out#asahi-dev · 2023-06-02 10:23
marcanhow often does the kernel kick the watchdog? could still be that#asahi-dev · 2023-06-02 10:23
marcan24 seconds this time#asahi-dev · 2023-06-02 10:20
marcanI would say it's about 15 seconds before the reboot#asahi-dev · 2023-06-02 10:18
marcanfun.#asahi-dev · 2023-06-02 10:18
marcanit just crashed#asahi-dev · 2023-06-02 10:18
marcanoh wait#asahi-dev · 2023-06-02 10:17
marcaneric_engestrom: maybe also your meson configuration?#asahi-dev · 2023-06-02 10:17
marcanI just survived two mesa compiles on j413 on battery...#asahi-dev · 2023-06-02 10:15
marcaneric_engestrom: can you paste me `grep -a . /proc/device-tree/chosen/asahi*` somewhere?#asahi-dev · 2023-06-02 10:11
marcansid-linux__: I've been hard rebooting machines dozens of times daily during testing#asahi-dev · 2023-06-02 10:04
marcanit probably should, writing plugins for every single device seems silly :p#asahi-dev · 2023-06-02 10:01
marcaneric_engestrom: fwiw, modulo the peculiarities of what thresholds are supported, this is all a completely standard kernel sysfs interface and I would expect it to be supported by generic code, not anything apple-specific#asahi-dev · 2023-06-02 10:00
marcanyes, but your only options are 80 and 100#asahi-dev · 2023-06-02 09:58
marcanyou could create a udev rule or something like that#asahi-dev · 2023-06-02 09:56
marcanit does not persist across reboots if you use KDE to set it#asahi-dev · 2023-06-02 09:56
marcanare you sure the charge limit is enabled?