src/amd/vulkan/radv_debug.c

   1 /*
   2  * Copyright © 2016 Red Hat.
   3  * Copyright © 2016 Bas Nieuwenhuizen
   4  *
   5  * based in part on anv driver which is:
   6  * Copyright © 2015 Intel Corporation
   7  *
   8  * Permission is hereby granted, free of charge, to any person obtaining a
   9  * copy of this software and associated documentation files (the "Software"),
  10  * to deal in the Software without restriction, including without limitation
  11  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
  12  * and/or sell copies of the Software, and to permit persons to whom the
  13  * Software is furnished to do so, subject to the following conditions:
  14  *
  15  * The above copyright notice and this permission notice (including the next
  16  * paragraph) shall be included in all copies or substantial portions of the
  17  * Software.
  18  *
  19  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  20  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  21  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
  22  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  23  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
  24  * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
  25  * IN THE SOFTWARE.
  26  */
  27
  28 #include <stdlib.h>
  29 #include <stdio.h>
  30
  31 #include "ac_debug.h"
  32 #include "radv_debug.h"
  33 #include "radv_shader.h"
  34
  35 #define TRACE_BO_SIZE 4096
  36
  37 /* Trace BO layout (offsets are 4 bytes):
  38  *
  39  * [0]: primary trace ID
  40  * [1]: secondary trace ID
  41  * [2-3]: 64-bit GFX pipeline pointer
  42  * [4-5]: 64-bit COMPUTE pipeline pointer
  43  */
  44
  45 bool
  46 radv_init_trace(struct radv_device *device)
  47 {
  48         struct radeon_winsys *ws = device->ws;
  49
  50         device->trace_bo = ws->buffer_create(ws, TRACE_BO_SIZE, 8,
  51                                              RADEON_DOMAIN_VRAM,
  52                                              RADEON_FLAG_CPU_ACCESS);
  53         if (!device->trace_bo)
  54                 return false;
  55
  56         device->trace_id_ptr = ws->buffer_map(device->trace_bo);
  57         if (!device->trace_id_ptr)
  58                 return false;
  59
  60         memset(device->trace_id_ptr, 0, TRACE_BO_SIZE);
  61
  62         ac_vm_fault_occured(device->physical_device->rad_info.chip_class,
  63                             &device->dmesg_timestamp, NULL);
  64
  65         return true;
  66 }
  67
  68 static void
  69 radv_dump_trace(struct radv_device *device, struct radeon_winsys_cs *cs)
  70 {
  71         const char *filename = getenv("RADV_TRACE_FILE");
  72         FILE *f = fopen(filename, "w");
  73
  74         if (!f) {
  75                 fprintf(stderr, "Failed to write trace dump to %s\n", filename);
  76                 return;
  77         }
  78
  79         fprintf(f, "Trace ID: %x\n", *device->trace_id_ptr);
  80         device->ws->cs_dump(cs, f, (const int*)device->trace_id_ptr, 2);
  81         fclose(f);
  82 }
  83
  84 static void
  85 radv_dump_shader(struct radv_shader_variant *shader, gl_shader_stage stage,
  86                  FILE *f)
  87 {
  88         if (!shader)
  89                 return;
  90
  91         fprintf(f, "%s:\n%s\n\n", radv_get_shader_name(shader, stage),
  92                 shader->disasm_string);
  93 }
  94
  95 static void
  96 radv_dump_shaders(struct radv_pipeline *pipeline,
  97                   struct radv_shader_variant *compute_shader, FILE *f)
  98 {
  99         unsigned mask;
 100
 101         /* Dump active graphics shaders. */
 102         mask = pipeline->active_stages;
 103         while (mask) {
 104                 int stage = u_bit_scan(&mask);
 105
 106                 radv_dump_shader(pipeline->shaders[stage], stage, f);
 107         }
 108
 109         radv_dump_shader(compute_shader, MESA_SHADER_COMPUTE, f);
 110 }
 111
 112 static void
 113 radv_dump_graphics_state(struct radv_pipeline *graphics_pipeline,
 114                          struct radv_pipeline *compute_pipeline, FILE *f)
 115 {
 116         struct radv_shader_variant *compute_shader =
 117                 compute_pipeline ? compute_pipeline->shaders[MESA_SHADER_COMPUTE] : NULL;
 118
 119         if (!graphics_pipeline)
 120                 return;
 121
 122         radv_dump_shaders(graphics_pipeline, compute_shader, f);
 123 }
 124
 125 static void
 126 radv_dump_compute_state(struct radv_pipeline *compute_pipeline, FILE *f)
 127 {
 128         if (!compute_pipeline)
 129                 return;
 130
 131         radv_dump_shaders(compute_pipeline,
 132                           compute_pipeline->shaders[MESA_SHADER_COMPUTE], f);
 133 }
 134
 135 static struct radv_pipeline *
 136 radv_get_saved_graphics_pipeline(struct radv_device *device)
 137 {
 138         uint64_t *ptr = (uint64_t *)device->trace_id_ptr;
 139
 140         return (struct radv_pipeline *)ptr[1];
 141 }
 142
 143 static struct radv_pipeline *
 144 radv_get_saved_compute_pipeline(struct radv_device *device)
 145 {
 146         uint64_t *ptr = (uint64_t *)device->trace_id_ptr;
 147
 148         return (struct radv_pipeline *)ptr[2];
 149 }
 150
 151 static bool
 152 radv_gpu_hang_occured(struct radv_queue *queue, enum ring_type ring)
 153 {
 154         struct radeon_winsys *ws = queue->device->ws;
 155
 156         if (!ws->ctx_wait_idle(queue->hw_ctx, ring, queue->queue_idx))
 157                 return true;
 158
 159         return false;
 160 }
 161
 162 void
 163 radv_check_gpu_hangs(struct radv_queue *queue, struct radeon_winsys_cs *cs)
 164 {
 165         struct radv_pipeline *graphics_pipeline, *compute_pipeline;
 166         struct radv_device *device = queue->device;
 167         enum ring_type ring;
 168         uint64_t addr;
 169
 170         ring = radv_queue_family_to_ring(queue->queue_family_index);
 171
 172         bool hang_occurred = radv_gpu_hang_occured(queue, ring);
 173         bool vm_fault_occurred = false;
 174         if (queue->device->instance->debug_flags & RADV_DEBUG_VM_FAULTS)
 175                 vm_fault_occurred = ac_vm_fault_occured(device->physical_device->rad_info.chip_class,
 176                                                         &device->dmesg_timestamp, &addr);
 177         if (!hang_occurred && !vm_fault_occurred)
 178                 return;
 179
 180         graphics_pipeline = radv_get_saved_graphics_pipeline(device);
 181         compute_pipeline = radv_get_saved_compute_pipeline(device);
 182
 183         if (vm_fault_occurred) {
 184                 fprintf(stderr, "VM fault report.\n\n");
 185                 fprintf(stderr, "Failing VM page: 0x%08"PRIx64"\n\n", addr);
 186         }
 187
 188         switch (ring) {
 189         case RING_GFX:
 190                 radv_dump_graphics_state(graphics_pipeline, compute_pipeline,
 191                                          stderr);
 192                 break;
 193         case RING_COMPUTE:
 194                 radv_dump_compute_state(compute_pipeline, stderr);
 195                 break;
 196         default:
 197                 assert(0);
 198                 break;
 199         }
 200
 201         radv_dump_trace(queue->device, cs);
 202         abort();
 203 }
 204
 205 void
 206 radv_print_spirv(struct radv_shader_module *module, FILE *fp)
 207 {
 208         char path[] = "/tmp/fileXXXXXX";
 209         char line[2048], command[128];
 210         FILE *p;
 211         int fd;
 212
 213         /* Dump the binary into a temporary file. */
 214         fd = mkstemp(path);
 215         if (fd < 0)
 216                 return;
 217
 218         if (write(fd, module->data, module->size) == -1)
 219                 goto fail;
 220
 221         sprintf(command, "spirv-dis %s", path);
 222
 223         /* Disassemble using spirv-dis if installed. */
 224         p = popen(command, "r");
 225         if (p) {
 226                 while (fgets(line, sizeof(line), p))
 227                         fprintf(fp, "%s", line);
 228                 pclose(p);
 229         }
 230
 231 fail:
 232         close(fd);
 233         unlink(path);
 234 }