Bug 85671

Summary:	[BYT Regression]igt/gem_reloc_vs_gpu fails with ppgtt enabled
Product:	DRI	Reporter:	lu hua <huax.lu>
Component:	DRM/Intel	Assignee:	Intel GFX Bugs mailing list <intel-gfx-bugs>
Status:	CLOSED FIXED	QA Contact:	Intel GFX Bugs mailing list <intel-gfx-bugs>
Severity:	normal
Priority:	high	CC:	intel-gfx-bugs
Version:	unspecified
Hardware:	All
OS:	Linux (All)
URL:	http://patchwork.freedesktop.org/patch/36045/
Whiteboard:
i915 platform:		i915 features:

Description lu hua 2014-10-31 02:11:31 UTC

==System Environment==
--------------------------
Regression: yes
good commit: 6476486cb3570d3662e4a3870b884e75d35d1072
bad commit: 421d82916774361a4840018fc3ab3a15751c3ea6

Non-working platforms: BYT

==kernel==
--------------------------
drm-intel-nightly/bd21cf795cf5dc278f1451d0f7a597cb1d13c2ba

==Bug detailed description==
It fails on BYT with -queued and -nightly kernel. It also has bug 84859, add i915.enable_ppgtt=1, it still fails.

output:
IGT-Version: 1.8-gcba3088 (x86_64) (Linux: 3.18.0-rc2_drm-intel-nightly_bd21cf_20141030+ x86_64)
Test assertion failure function do_test, file gem_reloc_vs_gpu.c:221:
Failed assertion: test == 0xdeadbeef
mismatch in buffer 1: 0x00000000 instead of 0xdeadbeef
Subtest faulting-reloc: FAIL (0.481s)

dmesg:
[  243.292762] gem_reloc_vs_gpu: executing
[  243.304659] [drm:i915_gem_open]
[  243.305204] [drm:i915_gem_open]
[  243.305439] gem_reloc_vs_gpu: starting subtest faulting-reloc
[  243.438118] [drm:valleyview_set_rps] GPU freq request from 646 MHz (214) to 667 MHz (215)
[  243.521711] [drm:valleyview_set_rps] GPU freq request from 667 MHz (215) to 646 MHz (214)
[  243.607257] [drm:valleyview_set_rps] GPU freq request from 646 MHz (214) to 667 MHz (215)
[  243.692318] [drm:valleyview_set_rps] GPU freq request from 667 MHz (215) to 708 MHz (217)
[  243.776392] [drm:valleyview_set_rps] GPU freq request from 708 MHz (217) to 791 MHz (221)
[  243.861469] [drm:valleyview_set_rps] GPU freq request from 791 MHz (221) to 854 MHz (224)
[  249.935839] [drm] stuck on blitter ring
[  249.942053] [drm] GPU HANG: ecode 2:0x21524118, in gem_reloc_vs_gp [4378], reason: Ring hung, action: reset
[  249.942817] [drm:i915_error_work_func] resetting chip
[  249.944414] drm/i915: Resetting chip after gpu hang
[  249.944426] [drm:init_status_page] render ring hws offset: 0x00409000
[  249.946426] [drm:init_status_page] bsd ring hws offset: 0x0042d000
[  249.946608] [drm:init_status_page] blitter ring hws offset: 0x00450000
[  249.946648] [drm:i9xx_update_primary_plane] Writing base 00473000 00000000 0 0 5504
[  249.953595] [drm:intel_crtc_set_config] [CRTC:8] [FB:35] #connectors=1 (x y) (0 0)
[  249.953606] [drm:intel_set_config_compute_mode_changes] computed changes for [CRTC:8], mode_changed=0, fb_changed=0
[  249.953611] [drm:intel_modeset_stage_output_state] [CONNECTOR:28:eDP-1] to [CRTC:8]
[  249.953617] [drm:intel_crtc_set_config] [CRTC:13] [NOFB]
[  249.953621] [drm:intel_set_config_compute_mode_changes] computed changes for [CRTC:13], mode_changed=0, fb_changed=0
[  249.953625] [drm:intel_modeset_stage_output_state] [CONNECTOR:28:eDP-1] to [CRTC:8]
[  251.937584] [drm:intel_print_rc6_info] Enabling RC6 states: RC6 on
[  251.939040] [drm:valleyview_enable_rps] GPLL enabled? yes
[  251.939045] [drm:valleyview_enable_rps] GPU status: 0x0000e0d0
[  251.939049] [drm:valleyview_enable_rps] current GPU freq: 854 MHz (224)
[  251.939053] [drm:valleyview_enable_rps] setting GPU freq to 646 MHz (214)
[  251.939057] [drm:valleyview_set_rps] GPU freq request from 854 MHz (224) to 646 MHz (214)

==Reproduce steps==
---------------------------- 
1. ./gem_reloc_vs_gpu --run-subtest faulting-reloc

Comment 1 lu hua 2014-10-31 07:00:13 UTC

add i915.enable_ppgtt=0, it works well.

Comment 2 Chris Wilson 2014-10-31 13:47:22 UTC

*** Bug 85673 has been marked as a duplicate of this bug. ***

Comment 3 Chris Wilson 2014-10-31 13:47:31 UTC

*** Bug 85672 has been marked as a duplicate of this bug. ***

Comment 4 Chris Wilson 2014-10-31 13:55:04 UTC

Please try http://patchwork.freedesktop.org/patch/36045/

Comment 5 lu hua 2014-11-03 08:08:57 UTC

(In reply to Chris Wilson from comment #4)
> Please try http://patchwork.freedesktop.org/patch/36045/

Fixed by this patch.

Comment 6 Chris Wilson 2014-11-04 10:54:39 UTC

commit 7ad6e04a3d0e7c24fbb14bd7dd2f3d7168423318
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Fri Oct 31 13:53:52 2014 +0000

    drm/i915: Only mark as map-and-fenceable when bound into the GGTT
    
    We use the obj->map_and_fenceable hint for when we already have a
    valid mapping of this object in the aperture. This hint can only apply
    to the GGTT and not to the aliasing-ppGTT. One user of the hint is
    execbuffer relocation, which began to fail when it tried to follow the
    hint and perform the relocate through the non-existent GGTT mapping.
    
    Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=85671
    Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
    Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

Comment 7 lu hua 2014-11-05 06:04:30 UTC

Verified.Fixed.

Comment 8 Elizabeth 2017-10-06 14:34:16 UTC

Closing old verified.

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.