drm/msm/gpu: Add the buffer objects from the submit to the crash dump

For hangs, dump copy out the contents of the buffer objects attached to the
guilty submission and print them in the crash dump report.

Signed-off-by: Jordan Crouse <jcrouse@codeaurora.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>

Documentation/gpu/msm-crash-dump.rst

@@ -66,6 +66,20 @@ ringbuffer
 		The contents of the ring encoded as ascii85.  Only the used
 		portions of the ring will be printed.
 
+bo
+	List of buffers from the hanging submission if available.
+	Each buffer object will have a uinque iova.
+
+	iova
+		GPU address of the buffer object.
+
+	size
+		Allocated size of the buffer object.
+
+	data
+		The contents of the buffer object encoded with ascii85.  Only
+		Trailing zeros at the end of the buffer will be skipped.
+
 registers
 	Set of registers values. Each entry is on its own line enclosed
 	by brackets { }.

drivers/gpu/drm/msm/adreno/adreno_gpu.c

@@ -437,6 +437,10 @@ void adreno_gpu_state_destroy(struct msm_gpu_state *state)
 	for (i = 0; i < ARRAY_SIZE(state->ring); i++)
 		kfree(state->ring[i].data);
 
+	for (i = 0; state->bos && i < state->nr_bos; i++)
+		kvfree(state->bos[i].data);
+
+	kfree(state->bos);
 	kfree(state->comm);
 	kfree(state->cmd);
 	kfree(state->registers);
@@ -460,6 +464,39 @@ int adreno_gpu_state_put(struct msm_gpu_state *state)
 }
 
 #if defined(CONFIG_DEBUG_FS) || defined(CONFIG_DEV_COREDUMP)
+
+static void adreno_show_object(struct drm_printer *p, u32 *ptr, int len)
+{
+	char out[ASCII85_BUFSZ];
+	long l, datalen, i;
+
+	if (!ptr || !len)
+		return;
+
+	/*
+	 * Only dump the non-zero part of the buffer - rarely will any data
+	 * completely fill the entire allocated size of the buffer
+	 */
+	for (datalen = 0, i = 0; i < len >> 2; i++) {
+		if (ptr[i])
+			datalen = (i << 2) + 1;
+	}
+
+	/* Skip printing the object if it is empty */
+	if (datalen == 0)
+		return;
+
+	l = ascii85_encode_len(datalen);
+
+	drm_puts(p, "    data: !!ascii85 |\n");
+	drm_puts(p, "     ");
+
+	for (i = 0; i < l; i++)
+		drm_puts(p, ascii85_encode(ptr[i], out));
+
+	drm_puts(p, "\n");
+}
+
 void adreno_show(struct msm_gpu *gpu, struct msm_gpu_state *state,
 		struct drm_printer *p)
 {
@@ -487,19 +524,20 @@ void adreno_show(struct msm_gpu *gpu, struct msm_gpu_state *state,
 		drm_printf(p, "    wptr: %d\n", state->ring[i].wptr);
 		drm_printf(p, "    size: %d\n", MSM_GPU_RINGBUFFER_SZ);
 
-		if (state->ring[i].data && state->ring[i].data_size) {
-			u32 *ptr = (u32 *) state->ring[i].data;
-			char out[ASCII85_BUFSZ];
-			long len = ascii85_encode_len(state->ring[i].data_size);
-			int j;
+		adreno_show_object(p, state->ring[i].data,
+			state->ring[i].data_size);
+	}
 
-			drm_printf(p, "    data: !!ascii85 |\n");
-			drm_printf(p, "     ");
+	if (state->bos) {
+		drm_puts(p, "bos:\n");
 
-			for (j = 0; j < len; j++)
-				drm_printf(p, ascii85_encode(ptr[j], out));
+		for (i = 0; i < state->nr_bos; i++) {
+			drm_printf(p, "  - iova: 0x%016llx\n",
+				state->bos[i].iova);
+			drm_printf(p, "    size: %zd\n", state->bos[i].size);
 
-			drm_printf(p, "\n");
+			adreno_show_object(p, state->bos[i].data,
+				state->bos[i].size);
 		}
 	}
 

drivers/gpu/drm/msm/msm_gpu.c

@@ -318,8 +318,39 @@ static void msm_gpu_devcoredump_free(void *data)
 	msm_gpu_crashstate_put(gpu);
 }
 
-static void msm_gpu_crashstate_capture(struct msm_gpu *gpu, char *comm,
-		char *cmd)
+static void msm_gpu_crashstate_get_bo(struct msm_gpu_state *state,
+		struct msm_gem_object *obj, u64 iova, u32 flags)
+{
+	struct msm_gpu_state_bo *state_bo = &state->bos[state->nr_bos];
+
+	/* Don't record write only objects */
+
+	state_bo->size = obj->base.size;
+	state_bo->iova = iova;
+
+	/* Only store the data for buffer objects marked for read */
+	if ((flags & MSM_SUBMIT_BO_READ)) {
+		void *ptr;
+
+		state_bo->data = kvmalloc(obj->base.size, GFP_KERNEL);
+		if (!state_bo->data)
+			return;
+
+		ptr = msm_gem_get_vaddr_active(&obj->base);
+		if (IS_ERR(ptr)) {
+			kvfree(state_bo->data);
+			return;
+		}
+
+		memcpy(state_bo->data, ptr, obj->base.size);
+		msm_gem_put_vaddr(&obj->base);
+	}
+
+	state->nr_bos++;
+}
+
+static void msm_gpu_crashstate_capture(struct msm_gpu *gpu,
+		struct msm_gem_submit *submit, char *comm, char *cmd)
 {
 	struct msm_gpu_state *state;
 
@@ -335,6 +366,17 @@ static void msm_gpu_crashstate_capture(struct msm_gpu *gpu, char *comm,
 	state->comm = kstrdup(comm, GFP_KERNEL);
 	state->cmd = kstrdup(cmd, GFP_KERNEL);
 
+	if (submit) {
+		int i;
+
+		state->bos = kcalloc(submit->nr_bos,
+			sizeof(struct msm_gpu_state_bo), GFP_KERNEL);
+
+		for (i = 0; state->bos && i < submit->nr_bos; i++)
+			msm_gpu_crashstate_get_bo(state, submit->bos[i].obj,
+				submit->bos[i].iova, submit->bos[i].flags);
+	}
+
 	/* Set the active crash state to be dumped on failure */
 	gpu->crashstate = state;
 
@@ -434,7 +476,7 @@ static void recover_worker(struct work_struct *work)
 
 	/* Record the crash state */
 	pm_runtime_get_sync(&gpu->pdev->dev);
-	msm_gpu_crashstate_capture(gpu, comm, cmd);
+	msm_gpu_crashstate_capture(gpu, submit, comm, cmd);
 	pm_runtime_put_sync(&gpu->pdev->dev);
 
 	kfree(cmd);

drivers/gpu/drm/msm/msm_gpu.h

@@ -181,6 +181,12 @@ struct msm_gpu_submitqueue {
 	struct kref ref;
 };
 
+struct msm_gpu_state_bo {
+	u64 iova;
+	size_t size;
+	void *data;
+};
+
 struct msm_gpu_state {
 	struct kref ref;
 	struct timeval time;
@@ -202,6 +208,9 @@ struct msm_gpu_state {
 
 	char *comm;
 	char *cmd;
+
+	int nr_bos;
+	struct msm_gpu_state_bo *bos;
 };
 
 static inline void gpu_write(struct msm_gpu *gpu, u32 reg, u32 data)