Probably a duplicate of https://bugs.freedesktop.org/show_bug.cgi?id=74973 (but that was an old bug) Latest git broke opencl on radeonsi. I could not start gimp (with opencl enabled) and even clinfo gives a segmetation fault. Bisected to: first bad commit: [4fd8d2df9c65396319619fa0784378600fc834d0] radeonsi: move emission of PA_SU_VTX_CNTL into emit_guardband This is my hw (glxinfo) OpenGL vendor string: X.Org OpenGL renderer string: AMD KABINI (DRM 3.26.0, 4.18.16-bfq-zstd+, LLVM 7.0.0) OpenGL core profile version string: 4.5 (Core Profile) Mesa 18.3.0-devel (git-0e0dc596a2) OpenGL core profile shading language version string: 4.50 OpenGL core profile context flags: (none) OpenGL core profile profile mask: core profile OpenGL core profile extensions: OpenGL version string: 4.5 (Compatibility Profile) Mesa 18.3.0-devel (git-0e0dc596a2) OpenGL shading language version string: 4.50 OpenGL context flags: (none) OpenGL profile mask: compatibility profile OpenGL extensions: OpenGL ES profile version string: OpenGL ES 3.2 Mesa 18.3.0-devel (git-0e0dc596a2) OpenGL ES profile shading language version string: OpenGL ES GLSL ES 3.20 Marco
Taking inspiration from: https://bugs.freedesktop.org/show_bug.cgi?id=108879 I tried to comment the call to si_clear_buffer and magically clinfo start to work again. diff --git a/src/gallium/drivers/radeonsi/si_pipe.c b/src/gallium/drivers/radeonsi/si_pipe.c index c487ef4..04678fb 100644 --- a/src/gallium/drivers/radeonsi/si_pipe.c +++ b/src/gallium/drivers/radeonsi/si_pipe.c @@ -620,10 +620,10 @@ static struct pipe_context *si_create_context(struct pipe_screen *screen, if (sctx->chip_class == CIK) { /* Clear the NULL constant buffer, because loads should return zeros. */ - uint32_t clear_value = 0; - si_clear_buffer(sctx, sctx->null_const_buf.buffer, 0, - sctx->null_const_buf.buffer->width0, - &clear_value, 4, SI_COHERENCY_SHADER); + uint32_t clear_value = 0xCCCCCCCC; + //si_clear_buffer(sctx, sctx->null_const_buf.buffer, 0, + // sctx->null_const_buf.buffer->width0, + // &clear_value, 4, SI_COHERENCY_SHADER); } return &sctx->b; fail: This is my hardware: 00:00.0 Host bridge: Advanced Micro Devices, Inc. [AMD] Family 16h Processor Root Complex 00:01.0 VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Kabini [Radeon HD 8330] 00:01.1 Audio device: Advanced Micro Devices, Inc. [AMD/ATI] Kabini HDMI/DP Audio 00:02.0 Host bridge: Advanced Micro Devices, Inc. [AMD] Family 16h Processor Function 0 00:02.1 PCI bridge: Advanced Micro Devices, Inc. [AMD] Family 16h Processor Functions 5:1 00:02.3 PCI bridge: Advanced Micro Devices, Inc. [AMD] Family 16h Processor Functions 5:1 00:02.4 PCI bridge: Advanced Micro Devices, Inc. [AMD] Family 16h Processor Functions 5:1 00:10.0 USB controller: Advanced Micro Devices, Inc. [AMD] FCH USB XHCI Controller (rev 01) 00:11.0 SATA controller: Advanced Micro Devices, Inc. [AMD] FCH SATA Controller [AHCI mode] 00:12.0 USB controller: Advanced Micro Devices, Inc. [AMD] FCH USB OHCI Controller (rev 39) 00:12.2 USB controller: Advanced Micro Devices, Inc. [AMD] FCH USB EHCI Controller (rev 39) 00:13.0 USB controller: Advanced Micro Devices, Inc. [AMD] FCH USB OHCI Controller (rev 39) 00:13.2 USB controller: Advanced Micro Devices, Inc. [AMD] FCH USB EHCI Controller (rev 39) 00:14.0 SMBus: Advanced Micro Devices, Inc. [AMD] FCH SMBus Controller (rev 3a) 00:14.2 Audio device: Advanced Micro Devices, Inc. [AMD] FCH Azalia Controller (rev 02) 00:14.3 ISA bridge: Advanced Micro Devices, Inc. [AMD] FCH LPC Bridge (rev 11) 00:18.0 Host bridge: Advanced Micro Devices, Inc. [AMD] Family 16h Processor Function 0 00:18.1 Host bridge: Advanced Micro Devices, Inc. [AMD] Family 16h Processor Function 1 00:18.2 Host bridge: Advanced Micro Devices, Inc. [AMD] Family 16h Processor Function 2 00:18.3 Host bridge: Advanced Micro Devices, Inc. [AMD] Family 16h Processor Function 3 00:18.4 Host bridge: Advanced Micro Devices, Inc. [AMD] Family 16h Processor Function 4 00:18.5 Host bridge: Advanced Micro Devices, Inc. [AMD] Family 16h Processor Function 5 01:00.0 Display controller: Advanced Micro Devices, Inc. [AMD/ATI] Sun PRO [Radeon HD 8570A/8570M] 02:00.0 Network controller: Qualcomm Atheros QCA9565 / AR9565 Wireless Network Adapter (rev 01) 03:00.0 Ethernet controller: Qualcomm Atheros QCA8172 Fast Ethernet (rev 10)
Created attachment 142654 [details] clinfo output Thi is the output of clinfo after applying the patch (commenting si_clear_buffer)
If you build radeonsi with debug sysmbols i.e. use --enable-debug you should be able to run gimp in gdb and see exactly were the segfault is. Could be as simple as a NULL check missing somewhere.
Gimp does not segfault, it just freeze: Stacktrace is this: #0 0x00007fb1f9872a79 in syscall () at /lib/x86_64-linux-gnu/libc.so.6 #1 0x00007fb1ccbfcb5c in () at /usr/lib/x86_64-linux-gnu/gallium-pipe/pipe_radeonsi.so #2 0x00007fb1ccc37271 in () at /usr/lib/x86_64-linux-gnu/gallium-pipe/pipe_radeonsi.so #3 0x00007fb1ccc39176 in () at /usr/lib/x86_64-linux-gnu/gallium-pipe/pipe_radeonsi.so #4 0x00007fb1ccc392ff in () at /usr/lib/x86_64-linux-gnu/gallium-pipe/pipe_radeonsi.so #5 0x00007fb1ccc490eb in () at /usr/lib/x86_64-linux-gnu/gallium-pipe/pipe_radeonsi.so #6 0x00007fb1ccc49961 in () at /usr/lib/x86_64-linux-gnu/gallium-pipe/pipe_radeonsi.so #7 0x00007fb1ccc0fc81 in amdgpu_winsys_create () at /usr/lib/x86_64-linux-gnu/gallium-pipe/pipe_radeonsi.so #8 0x00007fb1ccadcc07 in () at /usr/lib/x86_64-linux-gnu/gallium-pipe/pipe_radeonsi.so #9 0x00007fb1d1afb3f1 in () at /usr/lib/x86_64-linux-gnu/libMesaOpenCL.so.1 #10 0x00007fb1d1b2245c in () at /usr/lib/x86_64-linux-gnu/libMesaOpenCL.so.1 #11 0x00007fb1d1b2e058 in () at /usr/lib/x86_64-linux-gnu/libMesaOpenCL.so.1 #12 0x00007fb1d1afaa76 in () at /usr/lib/x86_64-linux-gnu/libMesaOpenCL.so.1 #13 0x00007fb1fbe1c0ca in () at /lib64/ld-linux-x86-64.so.2 #14 0x00007fb1fbe1c1d6 in () at /lib64/ld-linux-x86-64.so.2 #15 0x00007fb1fbe20253 in () at /lib64/ld-linux-x86-64.so.2 #16 0x00007fb1f98b3adf in _dl_catch_exception () at /lib/x86_64-linux-gnu/libc.so.6 #17 0x00007fb1fbe1fb1a in () at /lib64/ld-linux-x86-64.so.2 #18 0x00007fb1f9961276 in () at /lib/x86_64-linux-gnu/libdl.so.2 #19 0x00007fb1f98b3adf in _dl_catch_exception () at /lib/x86_64-linux-gnu/libc.so.6 #20 0x00007fb1f98b3b6f in _dl_catch_error () at /lib/x86_64-linux-gnu/libc.so.6 #21 0x00007fb1f9961975 in () at /lib/x86_64-linux-gnu/libdl.so.2 #22 0x00007fb1f9961331 in dlopen () at /lib/x86_64-linux-gnu/libdl.so.2 #23 0x00007fb1d2ff69af in () at /usr/lib/x86_64-linux-gnu/libOpenCL.so.1 #24 0x00007fb1d2ff76ab in clGetPlatformIDs () at /usr/lib/x86_64-linux-gnu/libOpenCL.so.1 #25 0x00007fb1fa821784 in () at /usr/lib/x86_64-linux-gnu/libgegl-0.4.so.0 #26 0x00007fb1fa822563 in () at /usr/lib/x86_64-linux-gnu/libgegl-0.4.so.0 #27 0x00007fb1fa7c3b8d in () at /usr/lib/x86_64-linux-gnu/libgegl-0.4.so.0 #28 0x00007fb1f9c26b6d in g_closure_invoke () at /usr/lib/x86_64-linux-gnu/libgobject-2.0.so.0 #29 0x00007fb1f9c398f3 in () at /usr/lib/x86_64-linux-gnu/libgobject-2.0.so.0 #30 0x00007fb1f9c42882 in g_signal_emit_valist () at /usr/lib/x86_64-linux-gnu/libgobject-2.0.so.0 #31 0x00007fb1f9c42ecf in g_signal_emit () at /usr/lib/x86_64-linux-gnu/libgobject-2.0.so.0 #32 0x00007fb1f9c2b1d4 in () at /usr/lib/x86_64-linux-gnu/libgobject-2.0.so.0 #33 0x00007fb1f9c2ab0e in () at /usr/lib/x86_64-linux-gnu/libgobject-2.0.so.0 Also clinfo get stuck with an analogous trace.
The same freeze occurs even on mesa master (19.1.0) from today. This is the stack trace (with a Ctrl+C on clinfo frozen) #0 0x00007ffff7eb4289 in syscall () from /lib/x86_64-linux-gnu/libc.so.6 #1 0x00007ffff17ae0ac in sys_futex (val3=-1, addr2=0x0, timeout=0x0, val1=2, op=9, addr1=0x5555558ddf88) at ../src/util/futex.h:38 #2 futex_wait (timeout=0x0, value=2, addr=0x5555558ddf88) at ../src/util/futex.h:50 #3 do_futex_fence_wait (fence=fence@entry=0x5555558ddf88, timeout=timeout@entry=false, abs_timeout=abs_timeout@entry=0) at ../src/util/u_queue.c:115 #4 0x00007ffff17ae6b9 in _util_queue_fence_wait (fence=fence@entry=0x5555558ddf88) at ../src/util/u_queue.c:130 #5 0x00007ffff17e9981 in util_queue_fence_wait (fence=0x5555558ddf88) at ../src/util/u_queue.h:161 #6 si_bind_compute_state (ctx=0x55555581cbf0, state=0x5555558ddf70) at ../src/gallium/drivers/radeonsi/si_compute.c:277 #7 0x00007ffff17ebd47 in si_compute_do_clear_or_copy (sctx=sctx@entry=0x55555581cbf0, dst=dst@entry=0x55555586a920, dst_offset=dst_offset@entry=0, src=src@entry=0x0, src_offset=src_offset@entry=0, size=size@entry=16, clear_value=0x7fffffffd780, clear_value_size=4, coher=SI_COHERENCY_SHADER) at ../src/gallium/drivers/radeonsi/si_compute_blit.c:161 #8 0x00007ffff17ec355 in si_clear_buffer (sctx=sctx@entry=0x55555581cbf0, dst=0x55555586a920, offset=offset@entry=0, size=16, clear_value=clear_value@entry=0x7fffffffd780, clear_value_size=clear_value_size@entry=4, coher=SI_COHERENCY_SHADER) at ../src/gallium/drivers/radeonsi/si_compute_blit.c:248 #9 0x00007ffff17b19e3 in si_create_context (screen=screen@entry=0x5555555d03a0, flags=flags@entry=0) at ../src/gallium/drivers/radeonsi/si_pipe.c:619 #10 0x00007ffff17b21f9 in radeonsi_screen_create (ws=<optimized out>, config=<optimized out>) at ../src/gallium/drivers/radeonsi/si_pipe.c:1124 #11 0x00007ffff1832f1c in amdgpu_winsys_create (fd=fd@entry=4, config=config@entry=0x7fffffffd930, screen_create=0x7ffff17b1bb0 <radeonsi_screen_create>) at ../src/gallium/winsys/amdgpu/drm/amdgpu_winsys.c:413 #12 0x00007ffff17178d7 in create_screen (fd=4, config=0x7fffffffd930) at ../src/gallium/targets/pipe-loader/pipe_radeonsi.c:15 #13 0x00007ffff6c19611 in pipe_loader_create_screen (dev=0x5555555bf260) at ../src/gallium/auxiliary/pipe-loader/pipe_loader.c:137 #14 0x00007ffff6c073bc in clover::device::device (this=0x5555555c5a50, platform=..., ldev=<optimized out>) at ../src/gallium/state_trackers/clover/core/device.cpp:47 #15 0x00007ffff6c13140 in create<clover::device, clover::platform&, pipe_loader_device*&> () at ../src/gallium/state_trackers/clover/util/pointer.hpp:230 #16 clover::platform::platform (this=0x7ffff7db94c0 <(anonymous namespace)::_clover_platform>) at ../src/gallium/state_trackers/clover/core/platform.cpp:36 #17 0x00007ffff6be3666 in __static_initialization_and_destruction_0 (__initialize_p=1, __priority=65535) at ../src/gallium/state_trackers/clover/api/platform.cpp:31 #18 _GLOBAL__sub_I_platform.cpp(void) () at ../src/gallium/state_trackers/clover/api/platform.cpp:141 #19 0x00007ffff7fe430a in ?? () from /lib64/ld-linux-x86-64.so.2 #20 0x00007ffff7fe4406 in ?? () from /lib64/ld-linux-x86-64.so.2 #21 0x00007ffff7fe8263 in ?? () from /lib64/ld-linux-x86-64.so.2 #22 0x00007ffff7ef505f in _dl_catch_exception () from /lib/x86_64-linux-gnu/libc.so.6 #23 0x00007ffff7fe7b4a in ?? () from /lib64/ld-linux-x86-64.so.2 ---Type <return> to continue, or q <return> to quit--- #24 0x00007ffff7f82256 in ?? () from /lib/x86_64-linux-gnu/libdl.so.2 #25 0x00007ffff7ef505f in _dl_catch_exception () from /lib/x86_64-linux-gnu/libc.so.6 #26 0x00007ffff7ef50ef in _dl_catch_error () from /lib/x86_64-linux-gnu/libc.so.6 #27 0x00007ffff7f82975 in ?? () from /lib/x86_64-linux-gnu/libdl.so.2 #28 0x00007ffff7f822e6 in dlopen () from /lib/x86_64-linux-gnu/libdl.so.2 #29 0x00007ffff7f8b7e0 in ?? () from /usr/lib/x86_64-linux-gnu/libOpenCL.so.1 #30 0x00007ffff7f8c4e3 in clGetPlatformIDs () from /usr/lib/x86_64-linux-gnu/libOpenCL.so.1 #31 0x000055555555a478 in ?? () #32 0x00007ffff7de409b in __libc_start_main () from /lib/x86_64-linux-gnu/libc.so.6 #33 0x000055555555a7ba in ?? ()
As previously said, commenting si_clear_buffer makes clinfo and opencl work again (with mesa master from git) ndex c6f93e7..1523919 100644 --- a/src/gallium/drivers/radeonsi/si_pipe.c +++ b/src/gallium/drivers/radeonsi/si_pipe.c @@ -616,9 +616,9 @@ static struct pipe_context *si_create_context(struct pipe_screen *screen, if (sctx->chip_class == CIK) { /* Clear the NULL constant buffer, because loads should return zeros. */ uint32_t clear_value = 0; - si_clear_buffer(sctx, sctx->null_const_buf.buffer, 0, + /*si_clear_buffer(sctx, sctx->null_const_buf.buffer, 0, sctx->null_const_buf.buffer->width0, - &clear_value, 4, SI_COHERENCY_SHADER); + &clear_value, 4, SI_COHERENCY_SHADER);*/ } return &sctx->b; fail:
This looks like a duplicate of 108879. I'm going to close this one as 108879 is more generic. KABINI si afaik CIK class GPU. Unfortunately I can't investigate as I don't have any CIK hw. *** This bug has been marked as a duplicate of bug 108879 ***
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.