https://bugs.freedesktop.org/show_bug.cgi?id=93370

Bug ID: 93370 Summary: [r600g] OpenCL atomic_cmpxchg() segmentation fault on Caicos Product: Mesa Version: git Hardware: x86-64 (AMD64) OS: Linux (All) Status: NEW Severity: major Priority: medium Component: Drivers/Gallium/r600 Assignee: dri-devel@lists.freedesktop.org Reporter: rivanvx@gmail.com QA Contact: dri-devel@lists.freedesktop.org

Created attachment 120495 --> https://bugs.freedesktop.org/attachment.cgi?id=120495&action=edit Program that segfaults

I'm using mesa 8b52fa71ac3d706f27ccbddcb050aae4965839e9 and llvm svn revision 255468.

The code that segfaults is attached. Debugging with gdb gives:

Starting program: /home/vedranm/atomic [Thread debugging using libthread_db enabled] Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1". [New Thread 0x7ffff034b700 (LWP 20619)] [New Thread 0x7fffef912700 (LWP 20620)] [New Thread 0x7fffef111700 (LWP 20621)] [Thread 0x7fffef111700 (LWP 20621) exited] [Thread 0x7fffef912700 (LWP 20620) exited]

Program received signal SIGSEGV, Segmentation fault. 0x00007ffff0a4480d in radeon_lookup_buffer (csc=csc@entry=0x7ffff7f31028, bo=bo@entry=0xf103c0) at radeon_drm_cs.c:230 230 if (i == -1 || csc->relocs_bo[i].bo == bo) (gdb) backtrace #0 0x00007ffff0a4480d in radeon_lookup_buffer (csc=csc@entry=0x7ffff7f31028, bo=bo@entry=0xf103c0) at radeon_drm_cs.c:230 #1 0x00007ffff0a44931 in radeon_add_buffer (added_domains=<synthetic pointer>, priority=29, domains=<optimized out>, usage=<optimized out>, bo=0xf103c0, cs=0x7ffff7f31010) at radeon_drm_cs.c:269 #2 radeon_drm_cs_add_buffer (rcs=0x7ffff7f31010, buf=0xf103c0, usage=<optimized out>, domains=<optimized out>, priority=RADEON_PRIO_VERTEX_BUFFER) at radeon_drm_cs.c:333 #3 0x00007ffff0a65a79 in radeon_add_to_buffer_list (rctx=0x65c260, ring=0x65c580, rbo=<optimized out>, rbo=<optimized out>, priority=RADEON_PRIO_VERTEX_BUFFER, usage=RADEON_USAGE_READ) at ../../../../src/gallium/drivers/radeon/r600_cs.h:53 #4 evergreen_emit_vertex_buffers (rctx=0x65c260, state=0x65ef00, resource_offset=816, pkt_flags=<optimized out>) at evergreen_state.c:1917 #5 0x00007ffff0a62467 in r600_emit_atom (atom=0x65ef00, rctx=0x65c260) at r600_pipe.h:578 #6 compute_emit_cs (grid_layout=0xf26950, block_layout=0xf26c10, ctx=0x65c260) at evergreen_compute.c:489 #7 evergreen_launch_grid (ctx_=0x65c260, block_layout=0xf26c10, grid_layout=grid_layout@entry=0xf26950, pc=pc@entry=0, input=input@entry=0xe68850) at evergreen_compute.c:631 #8 0x00007ffff58167ac in clover::kernel::launch (this=0xf24790, q=..., grid_offset=..., grid_size=std::vector of length 1, capacity 1 = {...}, block_size=std::vector of length 1, capacity 1 = {...}) at core/kernel.cpp:81 #9 0x00007ffff5811626 in std::function<void (clover::event&)>::operator()(clover::event&) const (__args#0=..., this=0xf26820) at /usr/include/c++/5/functional:2271 #10 clover::event::trigger (this=this@entry=0xf267e0) at core/event.cpp:55 #11 0x00007ffff58124f8 in clover::hard_event::hard_event(clover::command_queue&, unsigned int, clover::ref_vectorclover::event const&, std::function<void (clover::event&)>) (this=0xf267e0, q=..., command=<optimized out>, deps=..., action=...) at core/event.cpp:126 #12 0x00007ffff57f8ae8 in clover::create<clover::hard_event, clover::command_queue&, int, clover::ref_vectorclover::event&, clEnqueueNDRangeKernel(cl_command_queue, cl_kernel, cl_uint, const size_t*, const size_t*, const size_t*, cl_uint, _cl_event* const*, _cl_event**)::<lambda(clover::event&)> > () at ./util/pointer.hpp:230 #13 clEnqueueNDRangeKernel (d_q=0x63d328, d_kern=0xf24798, dims=<optimized out>, d_grid_offset=<optimized out>, d_grid_size=<optimized out>, d_block_size=0x7fffffffc430, num_deps=0, d_deps=0x0, rd_ev=0x0) at api/kernel.cpp:293 #14 0x00000000004014a8 in main () (gdb)