Bug 82749

That will be the struct timespec conversion. I'll fix this up with requests.

This bug is able to reproduce on BYT on latest -nihglty(186631131a9289dad22f51315d78b9b6ac5b425f) too

root@x-bytm02:/GFX/Test/Intel_gpu_tools/intel-gpu-tools/tests# ./gem_wait_render_timeout
IGT-Version: 1.7-g5c7bcb1 (x86_64) (Linux: 3.16.0_drm-intel-nightly_186631_20140818+ x86_64)
8192 iters is enough work
Finished with 330466078 time remaining
Test assertion failure function __real_main130, file gem_wait_render_timeout.c:223:
Failed assertion: timeout == 0
Last errno: 62, Timer expired

Test failed on IVB on latest -nightly(186631131a9289dad22f51315d78b9b6ac5b425f) too.

[root@x-ivb9 tests]# ./gem_wait_render_timeout
IGT-Version: 1.7-g1bec6cb (x86_64) (Linux: 3.16.0_drm-intel-nightly_186631_20140818_debug+ x86_64)
8192 iters is enough work
Finished with 760626512 time remaining
Test assertion failure function __real_main130, file gem_wait_render_timeout.c:223:
Failed assertion: timeout == 0
Last errno: 62, Timer expired

It impacts all platforms.

This failure still able to reproduce on ILK

[root@x-e6510 tests]# ./gem_wait_render_timeout
IGT-Version: 1.7-gf473a55 (i686) (Linux: 3.17.0-rc2_drm-intel-fixes_2a592b_20140902+ i686)
1024 iters is enough work
Finished with 929066159 time remaining
Test assertion failure function __real_main120, file gem_wait_render_timeout.c:213:
Failed assertion: timeout == 0
Last errno: 62, Timer expired
error: 896679 != 0

Fixed in git://people.freedesktop.org/~ickle/linux-2.6 requests

Please retest with the branch in comment 6.

(In reply to comment #6)
> Fixed in git://people.freedesktop.org/~ickle/linux-2.6 requests

The kernel unable to boot, please refer to Bug 82750 for the failure.

commit 3a5e1e6176fb61735a98f16a80c756b3cc69f125
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Sun Aug 24 19:34:16 2014 +0100

    drm/i915: Convert a couple more INTEL_INFO-esque macros to be pointer agnostic

    Just a couple more macros that assume that they were being passed a
    struct drm_device when they want a struct drm_i915_private. Use our
    magic macro to ease transitioning over to using drm_i915_privates

    Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

Test failed on BYT on latest -nightly(eabc0c8db15f9ba4d727aee5e0612a68cafe1ab5)

root@x-byt06:/GFX/Test/Intel_gpu_tools/intel-gpu-tools/tests# ./gem_wait --run-subtest render_timeout
IGT-Version: 1.8-g764772c (x86_64) (Linux: 3.17.0-rc7_drm-intel-nightly_eabc0c_20141006+ x86_64)
8192 iters is enough work
Finished with 111317851 time remaining
Test assertion failure function render_timeout, file gem_wait.c:200:
Failed assertion: timeout == 0
Last errno: 62, Timer expired
error: 513963 != 0
Subtest render_timeout: FAIL (15.448s)
root@x-byt06:/GFX/Test/Intel_gpu_tools/intel-gpu-tools/tests# echo $?
99

Cases below still fail


root@x-byt06:/GFX/Test/Intel_gpu_tools/intel-gpu-tools/tests# ./gem_wait --run-subtest invalid-flags
IGT-Version: 1.8-g7f82289 (x86_64) (Linux: 3.17.0_drm-intel-nightly_ea4bec_20141010+ x86_64)
Test assertion failure function invalid_flags, file gem_wait.c:230:
Failed assertion: ret != 0 && errno == EINVAL
Subtest invalid-flags: FAIL (0.000s)
root@x-byt06:/GFX/Test/Intel_gpu_tools/intel-gpu-tools/tests# ./gem_wait --run-subtest render_timeout
IGT-Version: 1.8-g7f82289 (x86_64) (Linux: 3.17.0_drm-intel-nightly_ea4bec_20141010+ x86_64)
8192 iters is enough work
Finished with 110706315 time remaining
Test assertion failure function render_timeout, file gem_wait.c:200:
Failed assertion: timeout == 0
Last errno: 62, Timer expired
error: 949723 != 0
Subtest render_timeout: FAIL (15.594s)

Clarify scope with renamed tests: This bug is _only_ about the render_timeout testcase. All the other new subtests are new, any issues with them should be filed as new bugs.

(In reply to Daniel Vetter from comment #11)
> Clarify scope with renamed tests: This bug is _only_ about the
> render_timeout testcase. All the other new subtests are new, any issues with
> them should be filed as new bugs.

igt/gem_wait/invalid-flags failure is tracking by bug 85280 now.

Is this still an issue on latest kernels?

(In reply to Daniel Vetter from comment #13)
> Is this still an issue on latest kernels?

Test still fails on latest -nightly(904b8529cd69f98b98d33b533a211003f99fb1cd)

root@x-bytm02:/GFX/Test/Intel_gpu_tools/intel-gpu-tools/tests# ./gem_wait --run-subtest render_timeout
IGT-Version: 1.8-gd807891 (x86_64) (Linux: 3.18.0-rc6_drm-intel-nightly_904b85_20141127+ x86_64)
8192 iters is enough work
Finished with 330928368 time remaining
Test assertion failure function render_timeout, file gem_wait.c:200:
Failed assertion: timeout == 0
Last errno: 62, Timer expired
error: 700747 != 0
Subtest render_timeout: FAIL (13.924s)

Hm, no bisect result.

On a hunch this regression is probably due to

commit 5ed0bdf21a85d78e04f89f15ccf227562177cbd9
Author: Thomas Gleixner <tglx@linutronix.de>
Date:   Wed Jul 16 21:05:06 2014 +0000

    drm: i915: Use nsec based interfaces
    
    Use ktime_get_raw_ns() and get rid of the back and forth timespec
    conversions.
    
    Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
    Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
    Signed-off-by: John Stultz <john.stultz@linaro.org>

Can you please test this commit + its parent to check whether that's right? Test results here on my machine are somehow strange so I'd like confirmation.

Please also test the following two patches (you need both, it's two bugs):

http://patchwork.freedesktop.org/patch/37743/
http://patchwork.freedesktop.org/patch/37744/

(In reply to Daniel Vetter from comment #15)
> Hm, no bisect result.
> 
> On a hunch this regression is probably due to
> 
> commit 5ed0bdf21a85d78e04f89f15ccf227562177cbd9
> Author: Thomas Gleixner <tglx@linutronix.de>
> Date:   Wed Jul 16 21:05:06 2014 +0000
> 
>     drm: i915: Use nsec based interfaces
>     
>     Use ktime_get_raw_ns() and get rid of the back and forth timespec
>     conversions.
>     
>     Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
>     Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
>     Signed-off-by: John Stultz <john.stultz@linaro.org>
> 
> Can you please test this commit + its parent to check whether that's right?
> Test results here on my machine are somehow strange so I'd like confirmation.
Yes, 5ed0bdf21a85d78e04f89f15ccf227562177cbd9 is the first bad commit. it's parents commit(f519b1a2e08c913375324a927992bb328387f169) is good commit. 

[root@x-hsw24 tests]# ./gem_wait --run-subtest render_timeout
IGT-Version: 1.8-gdbac905 (x86_64) (Linux: 3.16.0-rc5_kcloud_f519b1_20141201+ x86_64)
8192 iters is enough work
Finished with 815483789 time remaining
Subtest render_timeout: SUCCESS (10.395s)


[root@x-hsw24 tests]# ./gem_wait --run-subtest render_timeout
IGT-Version: 1.8-gdbac905 (x86_64) (Linux: 3.16.0-rc5_kcloud_5ed0bd_20141201+ x86_64)
8192 iters is enough work
Finished with 815166097 time remaining
Test assertion failure function render_timeout, file gem_wait.c:200:
Failed assertion: timeout == 0
Last errno: 62, Timer expired
error: 1170404 != 0
Subtest render_timeout: FAIL (10.392s)

(In reply to Daniel Vetter from comment #16)
> Please also test the following two patches (you need both, it's two bugs):
> 
> http://patchwork.freedesktop.org/patch/37743/
> http://patchwork.freedesktop.org/patch/37744/

The patch unable to patch on latest -nightly(0db9cf7742874ee2c09a35b640c1bb04cb379eb6). Could you help to update the patch? Thanks.

commit 7bd0e226e313547add65c7e4b0ecf65cb981a3c6
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Thu Dec 4 11:12:54 2014 +0100

    drm/i915: compute wait_ioctl timeout correctly

commit 9cca30688043eb1b6569f794041dfeec218dffa5
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Fri Nov 28 10:29:55 2014 +0100

    drm/i915: Handle inaccurate time conversion issues

Verified on latest -nightly(34d267c2ba9c0845432baf959a2c4deed87f3ee4)

[root@x-hsw24 tests]# ./gem_wait --run-subtest render_timeout
IGT-Version: 1.8-g819e68f (x86_64) (Linux: 3.18.0_drm-intel-nightly_34d267_20141209+ x86_64)
8192 iters is enough work
Finished with 812905627 time remaining
Subtest render_timeout: SUCCESS (11.472s)

Closing (>2 years) old Verified+Fixed.