Bug 89118 - [SKL Bisected]many Ogles3conform cases core dumped
Summary: [SKL Bisected]many Ogles3conform cases core dumped
Alias: None
Product: Mesa
Classification: Unclassified
Component: Drivers/DRI/i965 (show other bugs)
Version: unspecified
Hardware: All Linux (All)
: high major
Assignee: Ben Widawsky
QA Contact: Intel 3D Bugs Mailing List
Depends on:
Reported: 2015-02-13 02:50 UTC by lu hua
Modified: 2015-03-02 06:30 UTC (History)
3 users (show)

See Also:
i915 platform:
i915 features:

Only reserve message header space once (919 bytes, patch)
2015-02-16 20:55 UTC, Ben Widawsky
Details | Splinter Review
Fix uniform pull constant payload size (1.46 KB, patch)
2015-02-20 01:50 UTC, Ben Widawsky
Details | Splinter Review

Description lu hua 2015-02-13 02:50:21 UTC
System Environment:
Platform: SKL
Libdrm:         (master)libdrm-2.4.59-8-gccbb9aa887f992359335ecf2d26919b04e14e63f
Mesa:           (master)345e8cc8496b4e6c56105c7396e80d85a37e122c
Xserver:         (master)xorg-server-1.17.0
Xf86_video_intel: (master)2.99.917-100-g5b033d638bbf2c0b841088ca75f9eb8de5852cb5
Libva:          (master)f9741725839ea144e9a6a1827f74503ee39946c3
Libva_intel_driver:  (master)9a20d6c34cb65e5b85dd16d6c8b3a215c5972b18
Kernel:   (drm-intel-nightly)b4442ee4e150506cebeee72249efc566c5f14bbe

Bug detailed description:
many Ogles3conform cases core dumped on SKL.It always has this issue.

dEQP Core GL-CTS-2.0 (0x0052484b) starting..
  target implementation = 'X11'

Test case 'ES3-CTS.shaders.uniform_block.single_basic_type.shared.lowp_mat3'..
*** Error in `./glcts': free(): invalid size: 0x0000000003d9f5f0 ***
Aborted (core dumped)

(gdb) bt
#0  0x00007ffff5e64bb9 in __GI_raise (sig=sig@entry=6) at ../nptl/sysdeps/unix/sysv/linux/raise.c:56
#1  0x00007ffff5e67fc8 in __GI_abort () at abort.c:89
#2  0x00007ffff5eac27d in __malloc_assert (assertion=assertion@entry=0x7ffff5fac8b8 "nclears >= 3", file=file@entry=0x7ffff5fac6c0 "malloc.c", line=line@entry=3266,
    function=function@entry=0x7ffff5fac9dd <__func__.11481> "__libc_calloc") at malloc.c:293
#3  0x00007ffff5eb1f92 in __libc_calloc (n=<optimized out>, elem_size=<optimized out>) at malloc.c:3266
#4  0x00007ffff430ab72 in ralloc_size (ctx=ctx@entry=0x20bfe78, size=size@entry=260) at ralloc.c:113
#5  0x00007ffff430abce in rzalloc_size (ctx=ctx@entry=0x20bfe78, size=size@entry=260) at ralloc.c:134
#6  0x00007ffff430ac9c in rzalloc_array_size (ctx=ctx@entry=0x20bfe78, size=size@entry=4, count=count@entry=65) at ralloc.c:196
#7  0x00007ffff430b8be in ra_alloc_interference_graph (regs=0x1625b08, count=65) at register_allocate.c:382
#8  0x00007ffff43db9f2 in fs_visitor::assign_regs (this=this@entry=0x7fffffffbe80, allow_spilling=allow_spilling@entry=false) at brw_fs_reg_allocate.cpp:548
#9  0x00007ffff43b747e in fs_visitor::allocate_registers (this=this@entry=0x7fffffffbe80) at brw_fs.cpp:3657
#10 0x00007ffff43b84ba in fs_visitor::run_fs (this=this@entry=0x7fffffffbe80) at brw_fs.cpp:3812
#11 0x00007ffff43b88c7 in brw_wm_fs_emit (brw=brw@entry=0x7ffff7fa1038, mem_ctx=mem_ctx@entry=0x23e9658, key=key@entry=0x7fffffffd150,
    prog_data=prog_data@entry=0x7fffffffcfb0, fp=fp@entry=0x2095f00, prog=prog@entry=0x20abf88, final_assembly_size=final_assembly_size@entry=0x7fffffffcfac) at brw_fs.cpp:3880
#12 0x00007ffff4431715 in do_wm_prog (brw=0x7ffff7fa1038, prog=0x20abf88, fp=0x2095f00, key=0x7fffffffd150) at brw_wm.c:212
#13 0x00007ffff43b7736 in brw_fs_precompile (ctx=0x7ffff7fa1038, shader_prog=0x20abf88, prog=0x2095f00) at brw_fs.cpp:3997
#14 0x00007ffff440d526 in brw_shader_precompile (sh_prog=0x20abf88, ctx=0x7ffff7fa1038) at brw_shader.cpp:65
#15 brw_link_shader (ctx=0x7ffff7fa1038, shProg=0x20abf88) at brw_shader.cpp:278
#16 0x00007ffff428abc6 in _mesa_glsl_link_shader (ctx=0x7ffff7fa1038, prog=0x20abf88) at program/ir_to_mesa.cpp:3035
#17 0x00007ffff41a9b6a in link_program (ctx=0x7ffff7fa1038, program=<optimized out>) at main/shaderapi.c:932
#18 0x0000000000eae87e in glu::Program::linkProgram(unsigned int, unsigned int, unsigned int, std::string&, unsigned long&) ()
#19 0x0000000000eadd84 in glu::Program::Program(glu::RenderContext const&, char const*, char const*) ()
#20 0x00000000007e0d32 in glcts::UniformBlockCase::iterate() ()
#21 0x0000000000f00a97 in tcu::TestCaseWrapper::iterateTestCase(tcu::TestCase*) ()
#22 0x00000000007da1a1 in glcts::TestCaseWrapper::iterateTestCase(tcu::TestCase*) ()
#23 0x0000000000f02190 in tcu::TestExecutor::iterate() ()
#24 0x0000000000ef5141 in tcu::App::iterate() ()
#25 0x0000000000541638 in main ()

==Reproduce steps==
1. xinit
2. ./glcts --deqp-case=ES3-CTS.shaders.uniform_block.single_basic_type.shared.lowp_mat3
Comment 1 Ian Romanick 2015-02-13 19:11:54 UTC
Is this a regression?  If so, is there a bisect?
Comment 2 lu hua 2015-02-15 05:46:49 UTC
Bisect shows: 0ac4c272755c75108a10a84ce33bf6a6234985d3 is the first bad commit
commit 0ac4c272755c75108a10a84ce33bf6a6234985d3
Author:     Kristian Høgsberg <krh@bitplanet.net>
AuthorDate: Wed Dec 10 14:59:26 2014 -0800
Commit:     Kristian Høgsberg <krh@bitplanet.net>
CommitDate: Thu Jan 8 10:13:32 2015 -0800

    i965/skl: Always use a header for SIMD4x2 sampler messages

    SKL+ overloads the SIMD4x2 SIMD mode to mean either SIMD8D or SIMD4x2
    depending on bit 22 in the message header.  If the bit is 0 or there is
    no header we get SIMD8D.  We always wand SIMD4x2 in vec4 and for fs pull
    constants, so use a message header in those cases and set bit 22 there.

    Based on an initial patch from Ken.

    Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
    Signed-off-by: Kristian Høgsberg <krh@bitplanet.net>
Comment 3 Ben Widawsky 2015-02-15 17:12:23 UTC
I may have already fixed this in master.

Please test master
Comment 4 lu hua 2015-02-16 02:54:42 UTC
Test on the latest mesa master branch(commit a2299bfbbd6ed96), It still exists.
dEQP Core GL-CTS-2.0 (0x0052484b) starting..
  target implementation = 'X11'

Test case 'ES3-CTS.shaders.uniform_block.single_basic_type.shared.lowp_mat3'..
glcts: malloc.c:3266: __libc_calloc: Assertion `nclears >= 3' failed.
Aborted (core dumped)
Comment 5 Ben Widawsky 2015-02-16 20:55:20 UTC
Created attachment 113543 [details] [review]
Only reserve message header space once

Please test
Comment 6 Ben Widawsky 2015-02-20 01:50:44 UTC
Created attachment 113682 [details] [review]
Fix uniform pull constant payload size

This is the patch for upstream. Ignore the previous.
Comment 7 Ben Widawsky 2015-02-22 20:29:15 UTC
I've pushed the patch to master now, can you please test?
Comment 8 lu hua 2015-03-02 06:30:06 UTC
run ./glcts --deqp-case=ES3-CTS.shaders.uniform_block.single_basic_type.shared.lowp_mat3

dEQP Core GL-CTS-2.0 (0x0052484b) starting..
  target implementation = 'X11'

Test case 'ES3-CTS.shaders.uniform_block.single_basic_type.shared.lowp_mat3'..
Vertex compile time = 11.887000 ms
Fragment compile time = 5.257000 ms
Link time = 14.381000 ms
  Pass (Pass)


Test run totals:
  Passed:        1/1 (100.00%)
  Failed:        0/1 (0.00%)
  Not supported: 0/1 (0.00%)
  Warnings:      0/1 (0.00%)
Comment 9 lu hua 2015-03-02 06:30:35 UTC

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.