drm/i915: Add a debugfs file to dump complete context

Commit Message

[e-mail] Aug. 25, 2016, 2:19 p.m. UTC

From: Armin Reese <[e-mail]>

A 'cat' of the debugfs file i915_dump_lrc, dumps only the first 0x600(1536)
bytes of each ring's register state context. It does not provide
information about the remaining portion of the register state context.
This patch adds new file i915_dump_lrc_complete which displays full
register state context. Being able to view the latter in its entirety will
be very useful in debugging issues arising from incorrect context setup.

i915_dum_lrc still provides a snapshot of the context image.

v2: Since most of the content in some pages are mostly zeros, contiguous
sequence of zeros are skipped in the output.

Cc: Mika Kuoppala <[e-mail]>
Cc: Chris Wilson <[e-mail]>
Signed-off-by: Armin Reese <[e-mail]> (v1)
Signed-off-by: Arun Siluvery <[e-mail]>
---
 drivers/gpu/drm/i915/i915_debugfs.c | 70 ++++++++++++++++++++++++++++++-------
 1 file changed, 58 insertions(+), 12 deletions(-)

Comments

Chris Wilson Aug. 25, 2016, 3:13 p.m. UTC | #1

On Thu, Aug 25, 2016 at 03:19:15PM +0100, Arun Siluvery wrote:
> From: Armin Reese <[e-mail]>
> 
> A 'cat' of the debugfs file i915_dump_lrc, dumps only the first 0x600(1536)
> bytes of each ring's register state context. It does not provide
> information about the remaining portion of the register state context.
> This patch adds new file i915_dump_lrc_complete which displays full
> register state context. Being able to view the latter in its entirety will
> be very useful in debugging issues arising from incorrect context setup.

When I've been debugging context problems, I just dumped the image via
the CONTEXT_GETPARAM ioctl. Helped when looking at problems in situ and
I could then compare the images across the break point.

Dumping every page of every context object is not going to scale well,
my basic fear is that you will lose the information you need amongst the
noise. Especially as you will quickly run into the limits of seq_file...
-Chris

Patch

diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index 7d7b4d9..f4b5879 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -1969,13 +1969,20 @@  static int i915_context_status(struct seq_file *m, void *unused)
 	return 0;
 }
 
+enum {
+	LRC_CONTEXT_DUMP,  /* First PAGE size bytes of register state ctx */
+	FULL_CONTEXT_DUMP, /* Full context */
+};
+
 static void i915_dump_lrc_obj(struct seq_file *m,
 			      struct i915_gem_context *ctx,
-			      struct intel_engine_cs *engine)
+			      struct intel_engine_cs *engine,
+			      unsigned int flags)
 {
 	struct i915_vma *vma = ctx->engine[engine->id].state;
 	struct page *page;
-	int j;
+	struct sg_page_iter sg_iter;
+	int i;
 
 	seq_printf(m, "CONTEXT: %s %u\n", engine->name, ctx->hw_id);
 
@@ -1993,18 +2000,53 @@  static void i915_dump_lrc_obj(struct seq_file *m,
 		return;
 	}
 
+	i = 0;
 	page = i915_gem_object_get_page(vma->obj, LRC_STATE_PN);
-	if (page) {
-		u32 *reg_state = kmap_atomic(page);
+	for_each_sg_page(vma->pages->sgl, &sg_iter, vma->pages->nents, 0) {
+		int j;
+		u32 *reg_state;
+		int run_length = 0;
+		u32 data_size = (flags == LRC_CONTEXT_DUMP) ? 0x600 : PAGE_SIZE;
+		unsigned long ggtt_offset = i915_ggtt_offset(vma);
+		unsigned long page_offset = ggtt_offset + i*PAGE_SIZE;
+
+		if (flags == LRC_CONTEXT_DUMP && i > LRC_STATE_PN)
+			break;
 
-		for (j = 0; j < 0x600 / sizeof(u32) / 4; j += 4) {
-			seq_printf(m,
-				   "\t[0x%04x] 0x%08x 0x%08x 0x%08x 0x%08x\n",
-				   j * 4,
-				   reg_state[j], reg_state[j + 1],
+		page = sg_page_iter_page(&sg_iter);
+		if (WARN_ON(page == NULL))
+			break;
+
+		reg_state = kmap_atomic(page);
+
+		seq_printf(m, "Context object page: %d\n", i);
+		for (j = 0; j < data_size / sizeof(u32); j += 4) {
+			if (reg_state[j + 0] == 0 && reg_state[j + 1] == 0 &&
+			    reg_state[j + 2] == 0 && reg_state[j + 3] == 0) {
+				run_length += 4;
+				continue;
+			}
+
+			if (run_length) {
+				seq_printf(m, "\t[0x%08lx - 0x%08lx]: 0x00000000\n",
+					   page_offset + (j * 4) - (run_length * 4),
+					   page_offset + (j * 4) - 1);
+				run_length = 0;
+			}
+
+			seq_printf(m, "\t[0x%08lx] 0x%08x 0x%08x 0x%08x 0x%08x\n",
+				   page_offset + (j * 4),
+				   reg_state[j + 0], reg_state[j + 1],
 				   reg_state[j + 2], reg_state[j + 3]);
 		}
+		/* zeros in the whole page */
+		if (run_length) {
+			seq_printf(m, "\t[0x%08lx - 0x%08lx]: 0x00000000\n",
+				   page_offset, page_offset + data_size - 1);
+			run_length = 0;
+		}
 		kunmap_atomic(reg_state);
+		++i;
 	}
 
 	seq_putc(m, '\n');
@@ -2012,10 +2054,12 @@  static void i915_dump_lrc_obj(struct seq_file *m,
 
 static int i915_dump_lrc(struct seq_file *m, void *unused)
 {
+	struct drm_info_node *node = (struct drm_info_node *) m->private;
 	struct drm_i915_private *dev_priv = node_to_i915(m->private);
 	struct drm_device *dev = &dev_priv->drm;
 	struct intel_engine_cs *engine;
 	struct i915_gem_context *ctx;
+	uintptr_t flags = (uintptr_t) node->info_ent->data;
 	int ret;
 
 	if (!i915.enable_execlists) {
@@ -2027,9 +2071,10 @@  static int i915_dump_lrc(struct seq_file *m, void *unused)
 	if (ret)
 		return ret;
 
-	list_for_each_entry(ctx, &dev_priv->context_list, link)
+	list_for_each_entry(ctx, &dev_priv->context_list, link) {
 		for_each_engine(engine, dev_priv)
-			i915_dump_lrc_obj(m, ctx, engine);
+			i915_dump_lrc_obj(m, ctx, engine, flags);
+	}
 
 	mutex_unlock(&dev->struct_mutex);
 
@@ -5286,7 +5331,8 @@  static const struct drm_info_list i915_debugfs_list[] = {
 	{"i915_vbt", i915_vbt, 0},
 	{"i915_gem_framebuffer", i915_gem_framebuffer_info, 0},
 	{"i915_context_status", i915_context_status, 0},
-	{"i915_dump_lrc", i915_dump_lrc, 0},
+	{"i915_dump_lrc", i915_dump_lrc, 0, (void *)LRC_CONTEXT_DUMP},
+	{"i915_dump_lrc_complete", i915_dump_lrc, 0, (void *)FULL_CONTEXT_DUMP},
 	{"i915_execlists", i915_execlists, 0},
 	{"i915_forcewake_domains", i915_forcewake_domains, 0},
 	{"i915_swizzle_info", i915_swizzle_info, 0},