1 files changed, 105 insertions, 3 deletions

diff --git a/drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c b/drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c
index 7501849ed15d..55f443328d8e 100644
--- a/drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c
+++ b/drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c
@@ -42,7 +42,15 @@ struct a6xx_gpu_state {
 	struct a6xx_gpu_state_obj *cx_debugbus;
 	int nr_cx_debugbus;
 
+	struct msm_gpu_state_bo *gmu_log;
+	struct msm_gpu_state_bo *gmu_hfi;
+	struct msm_gpu_state_bo *gmu_debug;
+
+	s32 hfi_queue_history[2][HFI_HISTORY_SZ];
+
 	struct list_head objs;
+
+	bool gpu_initialized;
 };
 
 static inline int CRASHDUMP_WRITE(u64 *in, u32 reg, u32 val)
@@ -777,12 +785,12 @@ static void a6xx_get_gmu_registers(struct msm_gpu *gpu,
 	struct a6xx_gpu *a6xx_gpu = to_a6xx_gpu(adreno_gpu);
 
 	a6xx_state->gmu_registers = state_kcalloc(a6xx_state,
-		2, sizeof(*a6xx_state->gmu_registers));
+		3, sizeof(*a6xx_state->gmu_registers));
 
 	if (!a6xx_state->gmu_registers)
 		return;
 
-	a6xx_state->nr_gmu_registers = 2;
+	a6xx_state->nr_gmu_registers = 3;
 
 	/* Get the CX GMU registers from AHB */
 	_a6xx_get_gmu_registers(gpu, a6xx_state, &a6xx_gmu_reglist[0],
@@ -800,6 +808,45 @@ static void a6xx_get_gmu_registers(struct msm_gpu *gpu,
 		&a6xx_state->gmu_registers[2], false);
 }
 
+static struct msm_gpu_state_bo *a6xx_snapshot_gmu_bo(
+		struct a6xx_gpu_state *a6xx_state, struct a6xx_gmu_bo *bo)
+{
+	struct msm_gpu_state_bo *snapshot;
+
+	snapshot = state_kcalloc(a6xx_state, 1, sizeof(*snapshot));
+	if (!snapshot)
+		return NULL;
+
+	snapshot->iova = bo->iova;
+	snapshot->size = bo->size;
+	snapshot->data = kvzalloc(snapshot->size, GFP_KERNEL);
+	if (!snapshot->data)
+		return NULL;
+
+	memcpy(snapshot->data, bo->virt, bo->size);
+
+	return snapshot;
+}
+
+static void a6xx_snapshot_gmu_hfi_history(struct msm_gpu *gpu,
+					  struct a6xx_gpu_state *a6xx_state)
+{
+	struct adreno_gpu *adreno_gpu = to_adreno_gpu(gpu);
+	struct a6xx_gpu *a6xx_gpu = to_a6xx_gpu(adreno_gpu);
+	struct a6xx_gmu *gmu = &a6xx_gpu->gmu;
+	unsigned i, j;
+
+	BUILD_BUG_ON(ARRAY_SIZE(gmu->queues) != ARRAY_SIZE(a6xx_state->hfi_queue_history));
+
+	for (i = 0; i < ARRAY_SIZE(gmu->queues); i++) {
+		struct a6xx_hfi_queue *queue = &gmu->queues[i];
+		for (j = 0; j < HFI_HISTORY_SZ; j++) {
+			unsigned idx = (j + queue->history_idx) % HFI_HISTORY_SZ;
+			a6xx_state->hfi_queue_history[i][j] = queue->history[idx];
+		}
+	}
+}
+
 #define A6XX_GBIF_REGLIST_SIZE   1
 static void a6xx_get_registers(struct msm_gpu *gpu,
 		struct a6xx_gpu_state *a6xx_state,
@@ -937,6 +984,12 @@ struct msm_gpu_state *a6xx_gpu_state_get(struct msm_gpu *gpu)
 
 	a6xx_get_gmu_registers(gpu, a6xx_state);
 
+	a6xx_state->gmu_log = a6xx_snapshot_gmu_bo(a6xx_state, &a6xx_gpu->gmu.log);
+	a6xx_state->gmu_hfi = a6xx_snapshot_gmu_bo(a6xx_state, &a6xx_gpu->gmu.hfi);
+	a6xx_state->gmu_debug = a6xx_snapshot_gmu_bo(a6xx_state, &a6xx_gpu->gmu.debug);
+
+	a6xx_snapshot_gmu_hfi_history(gpu, a6xx_state);
+
 	/* If GX isn't on the rest of the data isn't going to be accessible */
 	if (!a6xx_gmu_gx_is_on(&a6xx_gpu->gmu))
 		return &a6xx_state->base;
@@ -950,7 +1003,8 @@ struct msm_gpu_state *a6xx_gpu_state_get(struct msm_gpu *gpu)
 	 * write out GPU state, so we need to skip this when the SMMU is
 	 * stalled in response to an iova fault
 	 */
-	if (!stalled && !a6xx_crashdumper_init(gpu, &_dumper)) {
+	if (!stalled && !gpu->needs_hw_init &&
+	    !a6xx_crashdumper_init(gpu, &_dumper)) {
 		dumper = &_dumper;
 	}
 
@@ -967,6 +1021,8 @@ struct msm_gpu_state *a6xx_gpu_state_get(struct msm_gpu *gpu)
 	if (snapshot_debugbus)
 		a6xx_get_debugbus(gpu, a6xx_state);
 
+	a6xx_state->gpu_initialized = !gpu->needs_hw_init;
+
 	return  &a6xx_state->base;
 }
 
@@ -978,6 +1034,12 @@ static void a6xx_gpu_state_destroy(struct kref *kref)
 	struct a6xx_gpu_state *a6xx_state = container_of(state,
 			struct a6xx_gpu_state, base);
 
+	if (a6xx_state->gmu_log)
+		kvfree(a6xx_state->gmu_log->data);
+
+	if (a6xx_state->gmu_hfi)
+		kvfree(a6xx_state->gmu_hfi->data);
+
 	list_for_each_entry_safe(obj, tmp, &a6xx_state->objs, node)
 		kfree(obj);
 
@@ -1189,8 +1251,48 @@ void a6xx_show(struct msm_gpu *gpu, struct msm_gpu_state *state,
 	if (IS_ERR_OR_NULL(state))
 		return;
 
+	drm_printf(p, "gpu-initialized: %d\n", a6xx_state->gpu_initialized);
+
 	adreno_show(gpu, state, p);
 
+	drm_puts(p, "gmu-log:\n");
+	if (a6xx_state->gmu_log) {
+		struct msm_gpu_state_bo *gmu_log = a6xx_state->gmu_log;
+
+		drm_printf(p, "    iova: 0x%016llx\n", gmu_log->iova);
+		drm_printf(p, "    size: %zu\n", gmu_log->size);
+		adreno_show_object(p, &gmu_log->data, gmu_log->size,
+				&gmu_log->encoded);
+	}
+
+	drm_puts(p, "gmu-hfi:\n");
+	if (a6xx_state->gmu_hfi) {
+		struct msm_gpu_state_bo *gmu_hfi = a6xx_state->gmu_hfi;
+		unsigned i, j;
+
+		drm_printf(p, "    iova: 0x%016llx\n", gmu_hfi->iova);
+		drm_printf(p, "    size: %zu\n", gmu_hfi->size);
+		for (i = 0; i < ARRAY_SIZE(a6xx_state->hfi_queue_history); i++) {
+			drm_printf(p, "    queue-history[%u]:", i);
+			for (j = 0; j < HFI_HISTORY_SZ; j++) {
+				drm_printf(p, " %d", a6xx_state->hfi_queue_history[i][j]);
+			}
+			drm_printf(p, "\n");
+		}
+		adreno_show_object(p, &gmu_hfi->data, gmu_hfi->size,
+				&gmu_hfi->encoded);
+	}
+
+	drm_puts(p, "gmu-debug:\n");
+	if (a6xx_state->gmu_debug) {
+		struct msm_gpu_state_bo *gmu_debug = a6xx_state->gmu_debug;
+
+		drm_printf(p, "    iova: 0x%016llx\n", gmu_debug->iova);
+		drm_printf(p, "    size: %zu\n", gmu_debug->size);
+		adreno_show_object(p, &gmu_debug->data, gmu_debug->size,
+				&gmu_debug->encoded);
+	}
+
 	drm_puts(p, "registers:\n");
 	for (i = 0; i < a6xx_state->nr_registers; i++) {
 		struct a6xx_gpu_state_obj *obj = &a6xx_state->registers[i];