[Bug 105821] New: Hard lock after running radeon.ko built-in tests

[bugzilla-daemon at bugzilla.kernel.org]

https://bugzilla.kernel.org/show_bug.cgi?id=105821

            Bug ID: 105821
           Summary: Hard lock after running radeon.ko built-in tests
           Product: Drivers
           Version: 2.5
    Kernel Version: 4.2
          Hardware: x86-64
                OS: Linux
              Tree: Mainline
            Status: NEW
          Severity: normal
          Priority: P1
         Component: Video(DRI - non Intel)
          Assignee: drivers_video-dri at kernel-bugs.osdl.org
          Reporter: iive at yahoo.com
        Regression: No

Created attachment 190021
  --> https://bugzilla.kernel.org/attachment.cgi?id=190021&action=edit
Kernel logs from running both test at the same time and from running the second
test only..

Built-in tests are enabled by the module "option radeon test=3". The integer
value bits are used as flags what test to run on the module load.

Running both test takes about 1-2 minutes to complete. The first one turns off
the display and trashes the system in a way that the first key pressed hangs it
hard - no flashing LEDs, no Magic SysRq.

---
The first test moves data from system ram to video ram and back.
It runs for about a minute. During that time the monitor is turning off and on
again (DELLs do that if DPMS signaling is not exactly right, it's connected on
DVI port). The rest of the system seems to function. Boot loading seams to
finish and sometimes I can salvage complete kernel message log. Pressing
CapsLock turns on its LED, but second press have no effect. Even pressing Magic
SysRq first doesn't provide any control of the system. It hangs hard and only
reset helps.

The seconds test is about syncing between GPU rings.
It seems to finish without error if run after the first one.
However if I run it alone with "test=2", I get a bunch of timeouts, "GPU
lockup" and other errors. When run alone the test doesn't prevent setting of
working graphic mode.

My videocard is Radeon HD5670 (Redwood).
I have confirmation that the same happens on Tahiti chipsets.

I have tried 3.17 (the oldest kernel I had) and 4.2.3 (latest release at this
date), the bug is present on all of them. (Hang may differ, but this seems
related on build parameters.)

I've reported the bug to freedesktop bugzilla a month ago, but so far nobody
have looked into it.
https://bugs.freedesktop.org/show_bug.cgi?id=91961

-- 
You are receiving this mail because:
You are watching the assignee of the bug.