drm/msm: add a5xx specific debugfs

Add some debugfs to dump out PFP and ME microcontroller state, as well
as some of the queues (MEQ and ROQ).  Also add a debugfs file to trigger
a GPU reset (and reloading the firmware on next submit).

Signed-off-by: Rob Clark <robdclark@gmail.com>

7 files changed, 207 insertions, 1 deletions

diff --git a/drivers/gpu/drm/msm/Makefile b/drivers/gpu/drm/msm/Makefile
index 92b3844202d2..ebe0c3d0b126 100644
--- a/drivers/gpu/drm/msm/Makefile
+++ b/drivers/gpu/drm/msm/Makefile
@@ -62,6 +62,8 @@ msm-y := \
 	msm_ringbuffer.o \
 	msm_submitqueue.o
 
+msm-$(CONFIG_DEBUG_FS) += adreno/a5xx_debugfs.o
+
 msm-$(CONFIG_DRM_FBDEV_EMULATION) += msm_fbdev.o
 msm-$(CONFIG_COMMON_CLK) += mdp/mdp4/mdp4_lvds_pll.o
 msm-$(CONFIG_COMMON_CLK) += hdmi/hdmi_pll_8960.o
diff --git a/drivers/gpu/drm/msm/adreno/a5xx_debugfs.c b/drivers/gpu/drm/msm/adreno/a5xx_debugfs.c
new file mode 100644
index 000000000000..cef09780ef17
--- /dev/null
+++ b/drivers/gpu/drm/msm/adreno/a5xx_debugfs.c
@@ -0,0 +1,188 @@
+/* Copyright (c) 2016-2017 The Linux Foundation. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 and
+ * only version 2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ */
+
+
+#include <linux/types.h>
+#include <linux/debugfs.h>
+#include <drm/drm_print.h>
+
+#include "a5xx_gpu.h"
+
+static int pfp_print(struct msm_gpu *gpu, struct drm_printer *p)
+{
+	int i;
+
+	drm_printf(p, "PFP state:\n");
+
+	for (i = 0; i < 36; i++) {
+		gpu_write(gpu, REG_A5XX_CP_PFP_STAT_ADDR, i);
+		drm_printf(p, "  %02x: %08x\n", i,
+			gpu_read(gpu, REG_A5XX_CP_PFP_STAT_DATA));
+	}
+
+	return 0;
+}
+
+static int me_print(struct msm_gpu *gpu, struct drm_printer *p)
+{
+	int i;
+
+	drm_printf(p, "ME state:\n");
+
+	for (i = 0; i < 29; i++) {
+		gpu_write(gpu, REG_A5XX_CP_ME_STAT_ADDR, i);
+		drm_printf(p, "  %02x: %08x\n", i,
+			gpu_read(gpu, REG_A5XX_CP_ME_STAT_DATA));
+	}
+
+	return 0;
+}
+
+static int meq_print(struct msm_gpu *gpu, struct drm_printer *p)
+{
+	int i;
+
+	drm_printf(p, "MEQ state:\n");
+	gpu_write(gpu, REG_A5XX_CP_MEQ_DBG_ADDR, 0);
+
+	for (i = 0; i < 64; i++) {
+		drm_printf(p, "  %02x: %08x\n", i,
+			gpu_read(gpu, REG_A5XX_CP_MEQ_DBG_DATA));
+	}
+
+	return 0;
+}
+
+static int roq_print(struct msm_gpu *gpu, struct drm_printer *p)
+{
+	int i;
+
+	drm_printf(p, "ROQ state:\n");
+	gpu_write(gpu, REG_A5XX_CP_ROQ_DBG_ADDR, 0);
+
+	for (i = 0; i < 512 / 4; i++) {
+		uint32_t val[4];
+		int j;
+		for (j = 0; j < 4; j++)
+			val[j] = gpu_read(gpu, REG_A5XX_CP_ROQ_DBG_DATA);
+		drm_printf(p, "  %02x: %08x %08x %08x %08x\n", i,
+			val[0], val[1], val[2], val[3]);
+	}
+
+	return 0;
+}
+
+static int show(struct seq_file *m, void *arg)
+{
+	struct drm_info_node *node = (struct drm_info_node *) m->private;
+	struct drm_device *dev = node->minor->dev;
+	struct msm_drm_private *priv = dev->dev_private;
+	struct drm_printer p = drm_seq_file_printer(m);
+	int (*show)(struct msm_gpu *gpu, struct drm_printer *p) =
+		node->info_ent->data;
+
+	return show(priv->gpu, &p);
+}
+
+#define ENT(n) { .name = #n, .show = show, .data = n ##_print }
+static struct drm_info_list a5xx_debugfs_list[] = {
+	ENT(pfp),
+	ENT(me),
+	ENT(meq),
+	ENT(roq),
+};
+
+/* for debugfs files that can be written to, we can't use drm helper: */
+static int
+reset_set(void *data, u64 val)
+{
+	struct drm_device *dev = data;
+	struct msm_drm_private *priv = dev->dev_private;
+	struct msm_gpu *gpu = priv->gpu;
+	struct adreno_gpu *adreno_gpu = to_adreno_gpu(gpu);
+	struct a5xx_gpu *a5xx_gpu = to_a5xx_gpu(adreno_gpu);
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EINVAL;
+
+	/* TODO do we care about trying to make sure the GPU is idle?
+	 * Since this is just a debug feature limited to CAP_SYS_ADMIN,
+	 * maybe it is fine to let the user keep both pieces if they
+	 * try to reset an active GPU.
+	 */
+
+	mutex_lock(&dev->struct_mutex);
+
+	if (adreno_gpu->pm4) {
+		release_firmware(adreno_gpu->pm4);
+		adreno_gpu->pm4 = NULL;
+	}
+
+	if (adreno_gpu->pfp) {
+		release_firmware(adreno_gpu->pfp);
+		adreno_gpu->pfp = NULL;
+	}
+	if (a5xx_gpu->pm4_bo) {
+		if (a5xx_gpu->pm4_iova)
+			msm_gem_put_iova(a5xx_gpu->pm4_bo, gpu->aspace);
+		drm_gem_object_unreference(a5xx_gpu->pm4_bo);
+		a5xx_gpu->pm4_bo = NULL;
+	}
+
+	if (a5xx_gpu->pfp_bo) {
+		if (a5xx_gpu->pfp_iova)
+			msm_gem_put_iova(a5xx_gpu->pfp_bo, gpu->aspace);
+		drm_gem_object_unreference(a5xx_gpu->pfp_bo);
+		a5xx_gpu->pfp_bo = NULL;
+	}
+
+	gpu->needs_hw_init = true;
+
+	pm_runtime_get_sync(&gpu->pdev->dev);
+	gpu->funcs->recover(gpu);
+
+	pm_runtime_put_sync(&gpu->pdev->dev);
+	mutex_unlock(&dev->struct_mutex);
+
+	return 0;
+}
+
+DEFINE_SIMPLE_ATTRIBUTE(reset_fops, NULL, reset_set, "%llx\n");
+
+
+int a5xx_debugfs_init(struct msm_gpu *gpu, struct drm_minor *minor)
+{
+	struct drm_device *dev = minor->dev;
+	struct dentry *ent;
+	int ret;
+
+	if (!minor)
+		return 0;
+
+	ret = drm_debugfs_create_files(a5xx_debugfs_list,
+			ARRAY_SIZE(a5xx_debugfs_list),
+			minor->debugfs_root, minor);
+
+	if (ret) {
+		dev_err(dev->dev, "could not install a5xx_debugfs_list\n");
+		return ret;
+	}
+
+	ent = debugfs_create_file("reset", S_IWUGO,
+		minor->debugfs_root,
+		dev, &reset_fops);
+	if (!ent)
+		return -ENOMEM;
+
+	return 0;
+}
diff --git a/drivers/gpu/drm/msm/adreno/a5xx_gpu.c b/drivers/gpu/drm/msm/adreno/a5xx_gpu.c
index 7e09d44e4a15..579c28c8c994 100644
--- a/drivers/gpu/drm/msm/adreno/a5xx_gpu.c
+++ b/drivers/gpu/drm/msm/adreno/a5xx_gpu.c
@@ -1195,6 +1195,7 @@ static const struct adreno_gpu_funcs funcs = {
 		.destroy = a5xx_destroy,
 #ifdef CONFIG_DEBUG_FS
 		.show = a5xx_show,
+		.debugfs_init = a5xx_debugfs_init,
 #endif
 		.gpu_busy = a5xx_gpu_busy,
 	},
diff --git a/drivers/gpu/drm/msm/adreno/a5xx_gpu.h b/drivers/gpu/drm/msm/adreno/a5xx_gpu.h
index 6fb8c2f9b9e4..7d71860c4bee 100644
--- a/drivers/gpu/drm/msm/adreno/a5xx_gpu.h
+++ b/drivers/gpu/drm/msm/adreno/a5xx_gpu.h
@@ -49,6 +49,10 @@ struct a5xx_gpu {
 
 #define to_a5xx_gpu(x) container_of(x, struct a5xx_gpu, base)
 
+#ifdef CONFIG_DEBUG_FS
+int a5xx_debugfs_init(struct msm_gpu *gpu, struct drm_minor *minor);
+#endif
+
 /*
  * In order to do lockless preemption we use a simple state machine to progress
  * through the process.
diff --git a/drivers/gpu/drm/msm/adreno/adreno_device.c b/drivers/gpu/drm/msm/adreno/adreno_device.c
index 62bdb7316da1..6263cb906b3c 100644
--- a/drivers/gpu/drm/msm/adreno/adreno_device.c
+++ b/drivers/gpu/drm/msm/adreno/adreno_device.c
@@ -150,6 +150,12 @@ struct msm_gpu *adreno_load_gpu(struct drm_device *dev)
 		return NULL;
 	}
 
+	if (gpu->funcs->debugfs_init) {
+		gpu->funcs->debugfs_init(gpu, dev->primary);
+		gpu->funcs->debugfs_init(gpu, dev->render);
+		gpu->funcs->debugfs_init(gpu, dev->control);
+	}
+
 	return gpu;
 }
 
diff --git a/drivers/gpu/drm/msm/msm_debugfs.c b/drivers/gpu/drm/msm/msm_debugfs.c
index 1855182c76ce..ba74cb4f94df 100644
--- a/drivers/gpu/drm/msm/msm_debugfs.c
+++ b/drivers/gpu/drm/msm/msm_debugfs.c
@@ -161,8 +161,11 @@ int msm_debugfs_init(struct drm_minor *minor)
 		return ret;
 	}
 
-	if (priv->kms->funcs->debugfs_init)
+	if (priv->kms->funcs->debugfs_init) {
 		ret = priv->kms->funcs->debugfs_init(priv->kms, minor);
+		if (ret)
+			return ret;
+	}
 
 	return ret;
 }
diff --git a/drivers/gpu/drm/msm/msm_gpu.h b/drivers/gpu/drm/msm/msm_gpu.h
index fccfccd303af..b8241179175a 100644
--- a/drivers/gpu/drm/msm/msm_gpu.h
+++ b/drivers/gpu/drm/msm/msm_gpu.h
@@ -65,6 +65,8 @@ struct msm_gpu_funcs {
 #ifdef CONFIG_DEBUG_FS
 	/* show GPU status in debugfs: */
 	void (*show)(struct msm_gpu *gpu, struct seq_file *m);
+	/* for generation specific debugfs: */
+	int (*debugfs_init)(struct msm_gpu *gpu, struct drm_minor *minor);
 #endif
 	int (*gpu_busy)(struct msm_gpu *gpu, uint64_t *value);
 };