[PATCH] drm/i915: Add debugfs file to dump entire logical context

[PATCH] drm/i915: Add debugfs file to dump entire logical context

[armin.c.reese]

From: Armin Reese <armin.c.reese@intel.com>

The new 'i915_context_dump' file generates a hex dump of the
entire logical context DRM object.  It is useful for
validating the contents of the default context set up by
the golden state batch buffer.

Signed-off-by: Armin Reese <armin.c.reese@intel.com>
---
 drivers/gpu/drm/i915/i915_debugfs.c | 92 ++++++++++++++++++++++++++++++-------
 1 file changed, 76 insertions(+), 16 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index a79f83c..7807c14 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -46,6 +46,11 @@ enum {
 	PINNED_LIST,
 };
 
+enum {
+	LRC_CONTEXT_DUMP,
+	FULL_CONTEXT_DUMP,
+};
+
 static const char *yesno(int v)
 {
 	return v ? "yes" : "no";
@@ -120,6 +125,52 @@ static inline const char *get_global_flag(struct drm_i915_gem_object *obj)
 }
 
 static void
+dump_32_obj(struct seq_file *m, struct drm_i915_gem_object *obj)
+{
+	struct page *page;
+	size_t size; /* In bytes */
+	size_t start_dword, end_dword, end_row_dword; /* In uint32_t offsets */
+	int i, num_pages;
+	uint32_t *obj_ptr;
+
+	if (i915_gem_object_get_pages(obj))
+		return;
+
+	size = obj->base.size; /* In bytes */
+	num_pages = (size + PAGE_SIZE - 1) / PAGE_SIZE;
+
+	for (i = 0; i < num_pages; i++) {
+		page = i915_gem_object_get_page(obj, i);
+		drm_clflush_pages(&page, 1);
+
+		start_dword = (i * PAGE_SIZE) / sizeof(uint32_t);
+		end_dword = start_dword + (PAGE_SIZE / sizeof(uint32_t));
+		if ((end_dword * sizeof(uint32_t)) > size)
+			end_dword = size / sizeof(uint32_t);
+
+		obj_ptr = (uint32_t *)kmap_atomic(page);
+
+		while (start_dword < end_dword) {
+			end_row_dword = start_dword + 8;
+			if (end_row_dword > end_dword)
+				end_row_dword = end_dword;
+			seq_printf(m, "0x%08lx:  ",
+					start_dword * sizeof(uint32_t));
+			while (start_dword < end_row_dword) {
+				seq_printf(m, "0x%08x ",
+						*(obj_ptr +
+						(start_dword &
+						 (PAGE_SIZE - 1))));
+				start_dword++;
+			}
+			seq_puts(m, "\n");
+		}
+
+		kunmap_atomic(obj_ptr);
+	}
+}
+
+static void
 describe_obj(struct seq_file *m, struct drm_i915_gem_object *obj)
 {
 	struct i915_vma *vma;
@@ -1773,9 +1824,10 @@ static int i915_context_status(struct seq_file *m, void *unused)
 	return 0;
 }
 
-static int i915_dump_lrc(struct seq_file *m, void *unused)
+static int i915_dump_lrc(struct seq_file *m, void *data)
 {
 	struct drm_info_node *node = (struct drm_info_node *) m->private;
+	uintptr_t dump_flag = (uintptr_t) node->info_ent->data;
 	struct drm_device *dev = node->minor->dev;
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	struct intel_engine_cs *ring;
@@ -1795,24 +1847,31 @@ static int i915_dump_lrc(struct seq_file *m, void *unused)
 		for_each_ring(ring, dev_priv, i) {
 			struct drm_i915_gem_object *ctx_obj = ctx->engine[i].state;
 
-			if (ring->default_context == ctx)
+			if ((ring->default_context == ctx) &&
+				(dump_flag == LRC_CONTEXT_DUMP))
 				continue;
 
 			if (ctx_obj) {
-				struct page *page = i915_gem_object_get_page(ctx_obj, 1);
-				uint32_t *reg_state = kmap_atomic(page);
-				int j;
-
-				seq_printf(m, "CONTEXT: %s %u\n", ring->name,
-						intel_execlists_ctx_id(ctx_obj));
-
-				for (j = 0; j < 0x600 / sizeof(u32) / 4; j += 4) {
-					seq_printf(m, "\t[0x%08lx] 0x%08x 0x%08x 0x%08x 0x%08x\n",
-					i915_gem_obj_ggtt_offset(ctx_obj) + 4096 + (j * 4),
-					reg_state[j], reg_state[j + 1],
-					reg_state[j + 2], reg_state[j + 3]);
+				if (dump_flag == LRC_CONTEXT_DUMP) {
+					struct page *page = i915_gem_object_get_page(ctx_obj, 1);
+					uint32_t *reg_state = kmap_atomic(page);
+					int j;
+
+					seq_printf(m, "CONTEXT: %s %u\n", ring->name,
+							intel_execlists_ctx_id(ctx_obj));
+
+					for (j = 0; j < 0x600 / sizeof(u32) / 4; j += 4) {
+						seq_printf(m, "\t[0x%08lx] 0x%08x 0x%08x 0x%08x 0x%08x\n",
+						i915_gem_obj_ggtt_offset(ctx_obj) + 4096 + (j * 4),
+						reg_state[j], reg_state[j + 1],
+						reg_state[j + 2], reg_state[j + 3]);
+					}
+					kunmap_atomic(reg_state);
+				} else if (dump_flag == FULL_CONTEXT_DUMP) {
+					seq_printf(m, "Full Context Dump: %s %u\n", ring->name,
+							intel_execlists_ctx_id(ctx_obj));
+					dump_32_obj(m, ctx_obj);
 				}
-				kunmap_atomic(reg_state);
 
 				seq_putc(m, '\n');
 			}
@@ -4187,7 +4246,8 @@ static const struct drm_info_list i915_debugfs_list[] = {
 	{"i915_opregion", i915_opregion, 0},
 	{"i915_gem_framebuffer", i915_gem_framebuffer_info, 0},
 	{"i915_context_status", i915_context_status, 0},
-	{"i915_dump_lrc", i915_dump_lrc, 0},
+	{"i915_context_dump", i915_dump_lrc, 0, (void *)FULL_CONTEXT_DUMP},
+	{"i915_dump_lrc", i915_dump_lrc, 0, (void *)LRC_CONTEXT_DUMP},
 	{"i915_execlists", i915_execlists, 0},
 	{"i915_gen6_forcewake_count", i915_gen6_forcewake_count_info, 0},
 	{"i915_swizzle_info", i915_swizzle_info, 0},
-- 
1.9.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH] drm/i915: Add debugfs file to dump entire logical context

[Volkin, Bradley D]

On Thu, Oct 30, 2014 at 04:03:23PM -0700, armin.c.reese@intel.com wrote:
> From: Armin Reese <armin.c.reese@intel.com>
> 
> The new 'i915_context_dump' file generates a hex dump of the
> entire logical context DRM object.  It is useful for
> validating the contents of the default context set up by
> the golden state batch buffer.

This patch needs a changelog providing brief descriptions of what
has changed in each new version that you send.

> 
> Signed-off-by: Armin Reese <armin.c.reese@intel.com>
> ---
>  drivers/gpu/drm/i915/i915_debugfs.c | 92 ++++++++++++++++++++++++++++++-------
>  1 file changed, 76 insertions(+), 16 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
> index a79f83c..7807c14 100644
> --- a/drivers/gpu/drm/i915/i915_debugfs.c
> +++ b/drivers/gpu/drm/i915/i915_debugfs.c
> @@ -46,6 +46,11 @@ enum {
>  	PINNED_LIST,
>  };
>  
> +enum {
> +	LRC_CONTEXT_DUMP,

I think this should be something like REG_STATE_CONTEXT_DUMP or similar.
It better describes what we're actually dumping in this case.

> +	FULL_CONTEXT_DUMP,
> +};
> +
>  static const char *yesno(int v)
>  {
>  	return v ? "yes" : "no";
> @@ -120,6 +125,52 @@ static inline const char *get_global_flag(struct drm_i915_gem_object *obj)
>  }
>  
>  static void
> +dump_32_obj(struct seq_file *m, struct drm_i915_gem_object *obj)
> +{
> +	struct page *page;
> +	size_t size; /* In bytes */
> +	size_t start_dword, end_dword, end_row_dword; /* In uint32_t offsets */
> +	int i, num_pages;
> +	uint32_t *obj_ptr;
> +
> +	if (i915_gem_object_get_pages(obj))
> +		return;
> +
> +	size = obj->base.size; /* In bytes */
> +	num_pages = (size + PAGE_SIZE - 1) / PAGE_SIZE;
> +
> +	for (i = 0; i < num_pages; i++) {
> +		page = i915_gem_object_get_page(obj, i);

I believe the preferred way to walk the list of pages is:

	struct sg_page_iter sg_iter;
	for_each_sg_page(obj->pages->sgl, &sg_iter, obj->pages->nents, 0) {
		page = sg_page_iter_page(&sg_iter);

If we make this change, num_pages is no longer used.

> +		drm_clflush_pages(&page, 1);
> +
> +		start_dword = (i * PAGE_SIZE) / sizeof(uint32_t);
> +		end_dword = start_dword + (PAGE_SIZE / sizeof(uint32_t));
> +		if ((end_dword * sizeof(uint32_t)) > size)
> +			end_dword = size / sizeof(uint32_t);

The size of a GEM object is always a multiple of the page size, so you
can assume that it's safe to just dump the entirety of each page in the
object's page list. I think you can also use this fact to simplify the
looping below. And you might consider doing obj_ptr++ instead of the
bitwise 'and'.

Brad

> +
> +		obj_ptr = (uint32_t *)kmap_atomic(page);
> +
> +		while (start_dword < end_dword) {
> +			end_row_dword = start_dword + 8;
> +			if (end_row_dword > end_dword)
> +				end_row_dword = end_dword;
> +			seq_printf(m, "0x%08lx:  ",
> +					start_dword * sizeof(uint32_t));
> +			while (start_dword < end_row_dword) {
> +				seq_printf(m, "0x%08x ",
> +						*(obj_ptr +
> +						(start_dword &
> +						 (PAGE_SIZE - 1))));
> +				start_dword++;
> +			}
> +			seq_puts(m, "\n");
> +		}
> +
> +		kunmap_atomic(obj_ptr);
> +	}
> +}
> +
> +static void
>  describe_obj(struct seq_file *m, struct drm_i915_gem_object *obj)
>  {
>  	struct i915_vma *vma;
> @@ -1773,9 +1824,10 @@ static int i915_context_status(struct seq_file *m, void *unused)
>  	return 0;
>  }
>  
> -static int i915_dump_lrc(struct seq_file *m, void *unused)
> +static int i915_dump_lrc(struct seq_file *m, void *data)
>  {
>  	struct drm_info_node *node = (struct drm_info_node *) m->private;
> +	uintptr_t dump_flag = (uintptr_t) node->info_ent->data;
>  	struct drm_device *dev = node->minor->dev;
>  	struct drm_i915_private *dev_priv = dev->dev_private;
>  	struct intel_engine_cs *ring;
> @@ -1795,24 +1847,31 @@ static int i915_dump_lrc(struct seq_file *m, void *unused)
>  		for_each_ring(ring, dev_priv, i) {
>  			struct drm_i915_gem_object *ctx_obj = ctx->engine[i].state;
>  
> -			if (ring->default_context == ctx)
> +			if ((ring->default_context == ctx) &&
> +				(dump_flag == LRC_CONTEXT_DUMP))
>  				continue;
>  
>  			if (ctx_obj) {
> -				struct page *page = i915_gem_object_get_page(ctx_obj, 1);
> -				uint32_t *reg_state = kmap_atomic(page);
> -				int j;
> -
> -				seq_printf(m, "CONTEXT: %s %u\n", ring->name,
> -						intel_execlists_ctx_id(ctx_obj));
> -
> -				for (j = 0; j < 0x600 / sizeof(u32) / 4; j += 4) {
> -					seq_printf(m, "\t[0x%08lx] 0x%08x 0x%08x 0x%08x 0x%08x\n",
> -					i915_gem_obj_ggtt_offset(ctx_obj) + 4096 + (j * 4),
> -					reg_state[j], reg_state[j + 1],
> -					reg_state[j + 2], reg_state[j + 3]);
> +				if (dump_flag == LRC_CONTEXT_DUMP) {
> +					struct page *page = i915_gem_object_get_page(ctx_obj, 1);
> +					uint32_t *reg_state = kmap_atomic(page);
> +					int j;
> +
> +					seq_printf(m, "CONTEXT: %s %u\n", ring->name,
> +							intel_execlists_ctx_id(ctx_obj));
> +
> +					for (j = 0; j < 0x600 / sizeof(u32) / 4; j += 4) {
> +						seq_printf(m, "\t[0x%08lx] 0x%08x 0x%08x 0x%08x 0x%08x\n",
> +						i915_gem_obj_ggtt_offset(ctx_obj) + 4096 + (j * 4),
> +						reg_state[j], reg_state[j + 1],
> +						reg_state[j + 2], reg_state[j + 3]);
> +					}
> +					kunmap_atomic(reg_state);
> +				} else if (dump_flag == FULL_CONTEXT_DUMP) {
> +					seq_printf(m, "Full Context Dump: %s %u\n", ring->name,
> +							intel_execlists_ctx_id(ctx_obj));
> +					dump_32_obj(m, ctx_obj);
>  				}
> -				kunmap_atomic(reg_state);
>  
>  				seq_putc(m, '\n');
>  			}
> @@ -4187,7 +4246,8 @@ static const struct drm_info_list i915_debugfs_list[] = {
>  	{"i915_opregion", i915_opregion, 0},
>  	{"i915_gem_framebuffer", i915_gem_framebuffer_info, 0},
>  	{"i915_context_status", i915_context_status, 0},
> -	{"i915_dump_lrc", i915_dump_lrc, 0},
> +	{"i915_context_dump", i915_dump_lrc, 0, (void *)FULL_CONTEXT_DUMP},
> +	{"i915_dump_lrc", i915_dump_lrc, 0, (void *)LRC_CONTEXT_DUMP},
>  	{"i915_execlists", i915_execlists, 0},
>  	{"i915_gen6_forcewake_count", i915_gen6_forcewake_count_info, 0},
>  	{"i915_swizzle_info", i915_swizzle_info, 0},
> -- 
> 1.9.1
> 
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/intel-gfx
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

[PATCH] drm/i915: Add debugfs file to dump entire logical context

[armin.c.reese]

From: Armin Reese <armin.c.reese@intel.com>

The new 'i915_context_dump' file generates a hex dump of the
entire logical context DRM object.  It is useful for
validating the contents of the default context set up by
the golden state batch buffer.

v1 - Reuse function i915_dump_lrc() instead of i915_context_status().
Separated display of HW Status Page from full Register State
Context

Signed-off-by: Armin Reese <armin.c.reese@intel.com>
---
 drivers/gpu/drm/i915/i915_debugfs.c | 97 +++++++++++++++++++++++++++++++------
 1 file changed, 81 insertions(+), 16 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index 0a69813..9bf519e 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -46,6 +46,11 @@ enum {
 	PINNED_LIST,
 };
 
+enum {
+	LRC_CONTEXT_DUMP,	/* First 1536 bytes of register state ctxt */
+	FULL_CONTEXT_DUMP,	/* Full context (HW status + reg state ctxt */
+};
+
 static const char *yesno(int v)
 {
 	return v ? "yes" : "no";
@@ -119,6 +124,48 @@ static inline const char *get_global_flag(struct drm_i915_gem_object *obj)
 	return i915_gem_obj_to_ggtt(obj) ? "g" : " ";
 }
 
+/*
+ *  Dump contents of GEM object to the screen. 8 DWORDs per line
+ */
+
+static void
+dump_32_obj(struct seq_file *m, struct drm_i915_gem_object *obj,
+		int start_pg, int num_pgs)
+{
+	struct page *page;
+	struct sg_page_iter sg_iter;
+	size_t pg_offset, obj_offset = 0;
+	uint32_t *pg_ptr, *curr_ptr;
+	int i, pg_cnt = 0;
+
+	for_each_sg_page(obj->pages->sgl, &sg_iter,
+			 obj->pages->nents, start_pg) {
+		page = sg_page_iter_page(&sg_iter);
+
+		pg_ptr = (uint32_t *)kmap_atomic(page);
+		curr_ptr = pg_ptr;
+		pg_offset = 0;
+
+		while (pg_offset < PAGE_SIZE) {
+			seq_printf(m, "0x%08lx:  ", obj_offset);
+			for (i = 0; i < 8; i++) {
+				seq_printf(m, "0x%08x ", *curr_ptr);
+				curr_ptr++;
+			}
+			seq_puts(m, "\n");
+
+			pg_offset += 8 * sizeof(uint32_t);
+			obj_offset += 8 * sizeof(uint32_t);
+		}
+
+		kunmap_atomic(pg_ptr);
+
+		if ((num_pgs != -1) &&
+		    (++pg_cnt >= num_pgs))
+			break;
+	}
+}
+
 static void
 describe_obj(struct seq_file *m, struct drm_i915_gem_object *obj)
 {
@@ -1773,9 +1820,10 @@ static int i915_context_status(struct seq_file *m, void *unused)
 	return 0;
 }
 
-static int i915_dump_lrc(struct seq_file *m, void *unused)
+static int i915_dump_lrc(struct seq_file *m, void *data)
 {
 	struct drm_info_node *node = (struct drm_info_node *) m->private;
+	uintptr_t dump_flag = (uintptr_t) node->info_ent->data;
 	struct drm_device *dev = node->minor->dev;
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	struct intel_engine_cs *ring;
@@ -1795,24 +1843,40 @@ static int i915_dump_lrc(struct seq_file *m, void *unused)
 		for_each_ring(ring, dev_priv, i) {
 			struct drm_i915_gem_object *ctx_obj = ctx->engine[i].state;
 
-			if (ring->default_context == ctx)
+			if ((ring->default_context == ctx) &&
+			    (dump_flag == LRC_CONTEXT_DUMP))
 				continue;
 
+			/*
+			 * Hardware Status Page is in first page of context
+			 * object.  Register state context begins on second
+			 * page.
+			 */
 			if (ctx_obj) {
-				struct page *page = i915_gem_object_get_page(ctx_obj, 1);
-				uint32_t *reg_state = kmap_atomic(page);
-				int j;
-
-				seq_printf(m, "CONTEXT: %s %u\n", ring->name,
-						intel_execlists_ctx_id(ctx_obj));
-
-				for (j = 0; j < 0x600 / sizeof(u32) / 4; j += 4) {
-					seq_printf(m, "\t[0x%08lx] 0x%08x 0x%08x 0x%08x 0x%08x\n",
-					i915_gem_obj_ggtt_offset(ctx_obj) + 4096 + (j * 4),
-					reg_state[j], reg_state[j + 1],
-					reg_state[j + 2], reg_state[j + 3]);
+				if (dump_flag == LRC_CONTEXT_DUMP) {
+					struct page *page = i915_gem_object_get_page(ctx_obj, 1);
+					uint32_t *reg_state = kmap_atomic(page);
+					int j;
+
+					seq_printf(m, "CONTEXT: %s %u\n", ring->name,
+							intel_execlists_ctx_id(ctx_obj));
+
+					for (j = 0; j < 0x600 / sizeof(u32) / 4; j += 4) {
+						seq_printf(m, "\t[0x%08lx] 0x%08x 0x%08x 0x%08x 0x%08x\n",
+						i915_gem_obj_ggtt_offset(ctx_obj) + 4096 + (j * 4),
+						reg_state[j], reg_state[j + 1],
+						reg_state[j + 2], reg_state[j + 3]);
+					}
+					kunmap_atomic(reg_state);
+				} else if (dump_flag == FULL_CONTEXT_DUMP) {
+					seq_printf(m, "Hardware Status Page: %s %u\n",
+							ring->name,
+							intel_execlists_ctx_id(ctx_obj));
+					dump_32_obj(m, ctx_obj, 0, 1);
+					seq_printf(m, "Register State Context: %s %u\n", ring->name,
+							intel_execlists_ctx_id(ctx_obj));
+					dump_32_obj(m, ctx_obj, 1, -1);
 				}
-				kunmap_atomic(reg_state);
 
 				seq_putc(m, '\n');
 			}
@@ -4188,7 +4252,8 @@ static const struct drm_info_list i915_debugfs_list[] = {
 	{"i915_opregion", i915_opregion, 0},
 	{"i915_gem_framebuffer", i915_gem_framebuffer_info, 0},
 	{"i915_context_status", i915_context_status, 0},
-	{"i915_dump_lrc", i915_dump_lrc, 0},
+	{"i915_context_dump", i915_dump_lrc, 0, (void *)FULL_CONTEXT_DUMP},
+	{"i915_dump_lrc", i915_dump_lrc, 0, (void *)LRC_CONTEXT_DUMP},
 	{"i915_execlists", i915_execlists, 0},
 	{"i915_gen6_forcewake_count", i915_gen6_forcewake_count_info, 0},
 	{"i915_swizzle_info", i915_swizzle_info, 0},
-- 
1.9.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH] drm/i915: Add debugfs file to dump entire logical context

[Daniel Vetter]

On Tue, Nov 04, 2014 at 09:19:36AM -0800, armin.c.reese@intel.com wrote:
> From: Armin Reese <armin.c.reese@intel.com>
> 
> The new 'i915_context_dump' file generates a hex dump of the
> entire logical context DRM object.  It is useful for
> validating the contents of the default context set up by
> the golden state batch buffer.
> 
> v1 - Reuse function i915_dump_lrc() instead of i915_context_status().

off-by-one in the patchlog ;-) Since this is v2 generally we use that tag
to list the changes since v1.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx