Bug 91520

Summary:	[radeonsi, PIGLIT, regression] spec@arb_gpu_shader_fp64@execution@fs-indirect-temp-double-{dst,src} triggers SIGSEGV
Product:	Mesa	Reporter:	Kai <kai>
Component:	Drivers/Gallium/radeonsi	Assignee:	Default DRI bug account <dri-devel>
Status:	RESOLVED NOTOURBUG	QA Contact:	Default DRI bug account <dri-devel>
Severity:	normal
Priority:	medium	Keywords:	regression
Version:	git
Hardware:	x86-64 (AMD64)
OS:	Linux (All)
Whiteboard:
i915 platform:		i915 features:
Attachments:	GDB: backtraces and register info for crashes

Description Kai 2015-08-01 10:42:22 UTC

Created attachment 117480 [details]
GDB: backtraces and register info for crashes

For two builds now I'm seeing the two Piglit tests spec@arb_gpu_shader_fp64@execution@fs-indirect-temp-double-{src,dst} crash in radeonsi_dri.so.

dmesg shows:
> shader_runner[26631]: segfault at 2017f9ba1 ip 00007f16519bd800 sp 00007ffc79c82250 error 6 in radeonsi_dri.so[7f165140f000+7d3000]

Attached you'll find the GDB output (full backtrace, full register info) for the crashes triggered by both tests with my current stack (see below for details).

The last known good Mesa version was Git:master/bb9d59aed5 (build date 2015-07-26). The first known failing build was Git:master/fdb8487613 (build date 2015-07-30). The LLVM revision used for the fdb8487613 build was r243627 (LLVM SVN trunk, 3.8-devel). Sadly I can't seem to find the LLVM revision used by the bb9d59aed5 build anylonger. But it was probably a build of a revision from that day, since I'm using the snapshot builds from <http://llvm.org/apt>, which are done twice per day.

My current stack (Debian testing as a base):
GPU: Hawaii PRO [Radeon R9 290] (ChipID = 0x67b1)
Mesa: Git:master/5d29eaef85
libdrm: 2.4.62-1
LLVM: SVN:trunk/r243678 (3.8 devel)
X.Org: 2:1.17.1-2
Linux: 4.1.3
Firmware: <https://secure.freedesktop.org/~agd5f/radeon_ucode/hawaii/>
> 286640da3d90d7b51bdb038b65addc47  hawaii_ce.bin
> 161105a73f7dfb2fca513327491c32d6  hawaii_mc.bin
> d6195059ea724981c9acd3abd6ee5166  hawaii_me.bin
> ad511d31a4fe3147c8d80b8f6770b8d5  hawaii_mec.bin
> 63eae3f33c77aadbc6ed1a09a2aed81e  hawaii_pfp.bin
> 5b72c73acf0cbd0cbb639302f65bc7dc  hawaii_rlc.bin
> f00de91c24b3520197e1ddb85d99c34a  hawaii_sdma1.bin
> 8e16f749d62b150d0d1f580d71bc4348  hawaii_sdma.bin
> 7b6ca5302b56bd35bf52804919d57e63  hawaii_smc.bin
> 9f2ba7e720e2af4d7605a9a4fd903513  hawaii_uvd.bin
> b0f2a043e72fbf265b2f858b8ddbdb09  hawaii_vce.bin
libclc: Git:master/7958b0202b
DDX: Git:master/b6d871bf29

Comment 1 Michel Dänzer 2015-08-03 02:56:05 UTC

I had the same problem with libelf1 0.163-4 (was fine with libelf1 0.159-4.2). I rebuilt Mesa against libelfg0-dev instead of libelf-dev, and the problem is gone for me.

If you can confirm this, please file a bug against libelf1.

Comment 2 Kai 2015-08-03 15:30:10 UTC

(In reply to Michel Dänzer from comment #1)
> I had the same problem with libelf1 0.163-4 (was fine with libelf1
> 0.159-4.2). I rebuilt Mesa against libelfg0-dev instead of libelf-dev, and
> the problem is gone for me.
> 
> If you can confirm this, please file a bug against libelf1.

Indeed, libelfg0 (0.8.13-5) works! Thanks for the pointer.

Comment 3 Kai 2015-08-03 15:56:17 UTC

Only as reference, if others hit this too: the Debian bug report can be found at <https://bugs.debian.org/794488>.

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.