| Summary: | system hang on "radeon_hwmon_get_pwm1_enable" | ||
|---|---|---|---|
| Product: | Mesa | Reporter: | Thomas DEBESSE <dev> |
| Component: | Drivers/Gallium/radeonsi | Assignee: | Default DRI bug account <dri-devel> |
| Status: | RESOLVED FIXED | QA Contact: | Default DRI bug account <dri-devel> |
| Severity: | normal | ||
| Priority: | medium | ||
| Version: | git | ||
| Hardware: | x86-64 (AMD64) | ||
| OS: | Linux (All) | ||
| Whiteboard: | |||
| i915 platform: | i915 features: | ||
| Attachments: |
complete dmesg log
verbose lspci output for the GPU possible fix attachment-7674-0.html |
||
|
Description
Thomas DEBESSE
2015-10-18 15:14:47 UTC
Created attachment 118963 [details]
complete dmesg log
Created attachment 118964 [details]
verbose lspci output for the GPU
Created attachment 118978 [details] [review] possible fix The attached patch should fix it. pwm control is only available with dpm, and you've disabled dpm via the kernel command line. Created attachment 118987 [details] attachment-7674-0.html > The attached patch should fix it. pwm control is only available with dpm, and you've disabled dpm via the kernel command line. Oh yes, the last grub update (fortuitously made the morning before I change my GPU for the one described above) introduced a "radeon.dpm=0" default kernel command line option, probably to workaround that other bug https://bugzilla.kernel.org/show_bug.cgi?id=103271 Thank you for your fast answer, I will try both (the patch and the kernel command line option). On which source tree this patch must be applied? -- Thomas DEBESSE (In reply to Thomas DEBESSE from comment #4) > > On which source tree this patch must be applied? It was against my drm-next tree, but should apply to any recent kernel. So, some news! I've just tested with radeon.dpm=1, but less than 10 minute after the boot the graphic display completely hangs with graphic glitches (the system was available from ssh for example, but was not able to reboot), the command "cat /sys/class/hwmon/hwmon0/pwm1_enable" works (I did it *after* the graphical hang, so the graphical hang was not a consequence of this command). So, the glitch is another bug, but I now verified that "cat /sys/class/hwmon/hwmon0/pwm1_enable" is not a crash cause when radeon.dpm=1. Also, I tried to load amdgpu instead of radeon module, blacklinsting radeon, then I verified that with amdgpu.dpm=0 there is no "/sys/class/hwmon/hwmon0/pwm1_enable" file and with amdgpu.dpm=1, there is a "/sys/class/hwmon/hwmon0/pwm1_enable" file and I can "cat" it. So, the bug was not reproduced in amdgpu module, which is a good news. I've not yet tried to recompile the kernel to test your patch. I'm now running Ubuntu 15.10, of course it changes nothing about this bug, but they ship by default a 4.2 kernel so I can use my distro tools to recompile the kernel and do it cleanly, applying the patch on the official Ubuntu kernel tree. Hi, I just recompiled my kernel applying your patch and it works, you fixed that issue! I cannot anymore "cat /sys/class/hwmon/hwmon0/pwm1_enable" with radeon.dpm=0 so my system does not hang when I run pwmconfig. :-) Thanks a lot for your fast response and your efficient help. About the other bug, the unexpected hang when using radeon.dpm=1 , I added some comments on bug #68059 . I close the current bug as RESOLVED/FIXED. |
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.