panfrost: Add PAN_MESA_DEBUG=sync

Sometimes it's useful to get information about GPU faults in the
console, so it's synchronized with other messages.

This commit will cause Mesa to wait for completion and check if there
are any faults raised by the GPU.

Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>

src/gallium/drivers/panfrost/pan_context.h

@@ -311,7 +311,8 @@ struct mali_single_framebuffer
 panfrost_emit_sfbd(struct panfrost_batch *batch, unsigned vertex_count);
 
 mali_ptr
-panfrost_fragment_job(struct panfrost_batch *batch, bool has_draws);
+panfrost_fragment_job(struct panfrost_batch *batch, bool has_draws,
+                      struct mali_job_descriptor_header **header_cpu);
 
 void
 panfrost_shader_compile(

src/gallium/drivers/panfrost/pan_fragment.c

@@ -49,7 +49,8 @@ panfrost_initialize_surface(
  * presentations, this is supposed to correspond to eglSwapBuffers) */
 
 mali_ptr
-panfrost_fragment_job(struct panfrost_batch *batch, bool has_draws)
+panfrost_fragment_job(struct panfrost_batch *batch, bool has_draws,
+                      struct mali_job_descriptor_header **header_cpu)
 {
         struct panfrost_screen *screen = pan_screen(batch->ctx->base.screen);
 
@@ -108,5 +109,6 @@ panfrost_fragment_job(struct panfrost_batch *batch, bool has_draws)
         struct panfrost_transfer transfer = panfrost_allocate_transient(batch, sizeof(header) + sizeof(payload));
         memcpy(transfer.cpu, &header, sizeof(header));
         memcpy(transfer.cpu + sizeof(header), &payload, sizeof(payload));
+        *header_cpu = (struct mali_job_descriptor_header *)transfer.cpu;
         return transfer.gpu;
 }

src/gallium/drivers/panfrost/pan_job.c

@@ -807,7 +807,8 @@ panfrost_batch_draw_wallpaper(struct panfrost_batch *batch)
 static int
 panfrost_batch_submit_ioctl(struct panfrost_batch *batch,
                             mali_ptr first_job_desc,
-                            uint32_t reqs)
+                            uint32_t reqs,
+                            struct mali_job_descriptor_header *header)
 {
         struct panfrost_context *ctx = batch->ctx;
         struct pipe_context *gallium = (struct pipe_context *) ctx;
@@ -877,6 +878,26 @@ panfrost_batch_submit_ioctl(struct panfrost_batch *batch,
                 return errno;
         }
 
+        if (pan_debug & PAN_DBG_SYNC) {
+                u32 status;
+
+                /* Wait so we can get errors reported back */
+                drmSyncobjWait(screen->fd, &batch->out_sync->syncobj, 1,
+                               INT64_MAX, 0, NULL);
+
+                status = header->exception_status;
+
+                if (status && status != 0x1) {
+                        fprintf(stderr, "Job %" PRIx64 " failed: source ID: 0x%x access: %s exception: 0x%x (exception_status 0x%x) fault_pointer 0x%" PRIx64 " \n",
+                               first_job_desc,
+                               (status >> 16) & 0xFFFF,
+                               pandecode_exception_access((status >> 8) & 0x3),
+                               status  & 0xFF,
+                               status,
+                               header->fault_pointer);
+                }
+        }
+
         /* Trace the job if we're doing that */
         if (pan_debug & PAN_DBG_TRACE) {
                 /* Wait so we can get errors reported back */
@@ -892,17 +913,19 @@ static int
 panfrost_batch_submit_jobs(struct panfrost_batch *batch)
 {
         bool has_draws = batch->first_job.gpu;
+        struct mali_job_descriptor_header *header;
         int ret = 0;
 
         if (has_draws) {
-                ret = panfrost_batch_submit_ioctl(batch, batch->first_job.gpu, 0);
+                header = (struct mali_job_descriptor_header *)batch->first_job.cpu;
+                ret = panfrost_batch_submit_ioctl(batch, batch->first_job.gpu, 0, header);
                 assert(!ret);
         }
 
         if (batch->first_tiler.gpu || batch->clear) {
-                mali_ptr fragjob = panfrost_fragment_job(batch, has_draws);
+                mali_ptr fragjob = panfrost_fragment_job(batch, has_draws, &header);
 
-                ret = panfrost_batch_submit_ioctl(batch, fragjob, PANFROST_JD_REQ_FS);
+                ret = panfrost_batch_submit_ioctl(batch, fragjob, PANFROST_JD_REQ_FS, header);
                 assert(!ret);
         }
 

src/gallium/drivers/panfrost/pan_screen.c

@@ -59,6 +59,7 @@ static const struct debug_named_value debug_options[] = {
         {"trace",     PAN_DBG_TRACE,    "Trace the command stream"},
         {"deqp",      PAN_DBG_DEQP,     "Hacks for dEQP"},
         {"afbc",      PAN_DBG_AFBC,     "Enable non-conformant AFBC impl"},
+        {"sync",      PAN_DBG_SYNC,     "Wait for each job's completion and check for any GPU fault"},
         DEBUG_NAMED_VALUE_END
 };
 

src/gallium/drivers/panfrost/pan_util.h

@@ -32,6 +32,7 @@
 #define PAN_DBG_TRACE           0x0002
 #define PAN_DBG_DEQP            0x0004
 #define PAN_DBG_AFBC            0x0008
+#define PAN_DBG_SYNC            0x0010
 
 extern int pan_debug;
 

src/panfrost/pandecode/decode.c

@@ -490,7 +490,7 @@ pandecode_block_format(enum mali_block_format fmt)
 #undef DEFINE_CASE
 
 #define DEFINE_CASE(name) case MALI_EXCEPTION_ACCESS_## name: return ""#name
-static char *
+char *
 pandecode_exception_access(enum mali_exception_access access)
 {
         switch (access) {

src/panfrost/pandecode/public.h

@@ -49,4 +49,8 @@ pandecode_inject_mmap(uint64_t gpu_va, void *cpu, unsigned sz, const char *name)
 
 int pandecode_jc(uint64_t jc_gpu_va, bool bifrost, unsigned gpu_id);
 
+char *
+pandecode_exception_access(enum mali_exception_access access);
+
+
 #endif /* __MMAP_TRACE_H__ */