GPU HANG: ecode 7:1:8edcfc79 for SDL test program

Maybe this is a duplicate of #2024 (closed)

Description: After X is started with startx and LXDE comes up, I can use the graphical interface, everything seems to be working, but if I start the attached very simple test SDL program I get GPU hang.

Steps to reproduce:

1. compile test program: gcc -g -O0 -Wall -Werror -I/usr/include/SDL2 -D_REENTRANT -lSDL2 -lGL test.c -o test

2a. run test program for still usable X: ./test 1

2b. run test program for not refreshing X: ./test 2

How often does the steps listed above trigger the issue: always

[root@localhost ~]# uname -m
x86_64
[root@localhost ~]# uname -r
5.8.14-200.fc32.x86_64
[root@localhost ~]# cat /etc/fedora-release
Fedora release 32 (Thirty Two)

test.c sys_class_drm_card0_error dmesg dmidecode

changed the description

Hang from the w/a. So yes same critical bug.

Grabbed a hsw gt1, and that test case is not enough to reproduce the issue for me. Oh well, it was a promising lead.

I tried to investigate it further with ./test 1, since that one does not need reboot or at least X restart.

It seems that ./test 1 does not produce the error every time, I also have to start it at most 15 times.

I added syslog printfs to SDL2 library, but the error did not appear to come from one location, even sometimes it came after the test program exited. so I guessed that the error checking/message writing is delayed.

Indeed, I found in the i915 code that the heartbeat where from the error is printed is checked only in every 2500 ms.

I also found that I can lower that parameter in /sys/class/drm/engine/rcs0/heartbeat_... setting, so I tried it even at 10 ms, but I still got random places where the error is printed into syslog.

My next guess was to somehow wait for each request completion in i915, if that is possible, but that system seems complicated to me and have not found a safe way to do that.

Please try this command, maybe it can reproduce the error for you too, because I found that manually starting the test needs more try than the scripted one:

for((i=0; i<100; i++)); do echo $i; ./test 1; done

(usually it gave me the error in the first 3 tries)

Hi Reporter, can you try to reproduce this issue on latest drm-tip.

https://cgit.freedesktop.org/drm-tip

Please post your observations on it and any steps followed to reproduce the issue.

marked this issue as a duplicate of #2024 (closed)

closed

marked this issue as related to #2024 (closed)