Kernel oops in vc4_overflow_mem_work

anholt / linux

Other

134 stars 24 forks source link

Kernel oops in vc4_overflow_mem_work #115

Closed anyc closed 6 years ago

anyc commented 6 years ago

I get exactly the same kernel oops as described in #114 but at a random time after boot. I'd say around 30 minutes to several hours after booting the device. I discussed this issue already with @lategoodbye in my original report at https://github.com/raspberrypi/linux/issues/2217. I try to summarize it here:

I use a self-compiled kernel but it also occurs with a raspbian image
also occurs with kernel 4.9.36 and another RPi2 and a RPi3
I use the RPi for a slideshow with large photographs but it also happens without the app
I see a lot of "Resetting GPU" messages and the oops seems to occur ~50 ms after one of those messages
I do not own an official power supply but I used two different, "non-cheap" supplies with different cables

If it helps, I can also send you a link to the SD card image.

anyc commented 6 years ago

The same also happens with an original RPi power supply. The last output from a custom surveillance script calling vcgencmd every 10s:

frequency(1)=250000000
frequency(45)=600000000
temp=43.3'C
throttled=0x0

lategoodbye commented 6 years ago

First step to fix this issue, would be a reproducible scenario which also works for us.

Is there a chance to reproduce this issue with a default raspbian without any custom applications?

anyc commented 6 years ago

Right now raspbian is running for 9h without the app and without oops. I'll let it run for a little longer.

I can provide you the raspbian image with just the app and pictures. I just don't want to share them publicly in original resolution.

Thank you!

anyc commented 6 years ago

I tried different runs during the last days but the results are all the same. Raspbian without the app works. Both images cause an oops after a random amount of time with the app running.

lategoodbye commented 6 years ago

@anyc

Please try the following patches (keep in mind they are written for current mainline, so maybe they need rework): https://patchwork.kernel.org/patch/10052207/ https://patchwork.kernel.org/patch/10052209/

anyc commented 6 years ago

It looks good so far. No oops after 8h with the app running.

anholt commented 6 years ago

I've merged the patch for upstream and sent the PR for 4.9 downstream. If your testing says the patch fixed it, please close. (and thanks for testing!)

anyc commented 6 years ago

So, after several weeks there was no reported issue. Thank you!