[Intel-gfx] [Regression] WARNING: drivers/gpu/drm/i915/i915_gem.c:4525 i915_gem_free_object

[Chris Wilson]

On Mon, Feb 23, 2015 at 11:12:39PM +0300, Andrey Skvortsov wrote:
> Hi, 
> 
> This warning is moved from linux-next to v4.0-rc1 now. After system boot is just a black screen.
> I ssh'ed into the machine and saved the log. I attached updated dmesg.log with drm.debug=6. Hopefully it helps. 
> If you need any other debug information, traces, core dump or something else. Feel free to ask.

The warning from free_object is annoying (and quite possibly dangerous),
but the actual hang during boot is:

[  243.876375] INFO: task Xorg:2422 blocked for more than 120 seconds.
[  243.876382]       Tainted: G        W   E   4.0.0-rc1-150223- #2
[  243.876388] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[  243.876393] Xorg            D ffff88019fc12dc0     0  2422   2180 0x00400000
[  243.876404]  ffff8800dabfe1a0 0000000000000002 ffff880194537fd8 ffff880194537ba0
[  243.876416]  ffff8800dab9e22c ffff8800dabfe1a0 ffff8800dab9e230 00000000ffffffff
[  243.876426]  ffffffff813e2479 ffff8800dab9e228 ffffffff813e26a7 0000000000000000
[  243.876438] Call Trace:
[  243.876449]  [<ffffffff813e2479>] ? schedule+0x6f/0x7c
[  243.876459]  [<ffffffff813e26a7>] ? schedule_preempt_disabled+0x15/0x21
[  243.876469]  [<ffffffff813e3347>] ? __ww_mutex_lock_slowpath+0xdf/0x1c2
[  243.876480]  [<ffffffff813e3446>] ? __ww_mutex_lock+0x1c/0x93
[  243.876541]  [<ffffffffa050e70d>] ? modeset_lock+0x8f/0xf2 [drm]
[  243.876632]  [<ffffffffa09aa0b9>] ? intel_get_load_detect_pipe+0x80/0x427 [i915]
[  243.876674]  [<ffffffffa04fd42f>] ? drm_ut_debug_printk+0x5e/0x63 [drm]
[  243.876771]  [<ffffffffa09d4661>] ? intel_tv_detect+0x115/0x43a [i915]
[  243.876783]  [<ffffffff810608d9>] ? preempt_count_sub+0xbf/0xca
[  243.876809]  [<ffffffffa05d6f24>] ? drm_helper_probe_single_connector_modes_merge_bits+0xc6/0x38d [drm_kms_helper]
[  243.876860]  [<ffffffffa0505b5d>] ? drm_mode_getconnector+0xf4/0x2ac [drm]
[  243.876900]  [<ffffffffa04fa911>] ? drm_ioctl+0x338/0x3c5 [drm]
[  243.876949]  [<ffffffffa0505a69>] ? drm_mode_getcrtc+0xb3/0xb3 [drm]
[  243.876961]  [<ffffffff81167deb>] ? fsnotify+0x314/0x35d
[  243.876973]  [<ffffffff811487be>] ? do_vfs_ioctl+0x379/0x431
[  243.876983]  [<ffffffff811488cc>] ? SyS_ioctl+0x56/0x7c
[  243.876994]  [<ffffffff813e5152>] ? system_call_fastpath+0x12/0x17

i.e. it is a mutex deadlock inside tv detect. Daniel does that make sense?
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre