[Intel-gfx] [PATCH v2 26/37] drm/i915: error capture with no ggtt slot

[Matthew Auld]

From: Daniele Ceraolo Spurio <daniele.ceraolospurio at intel.com>

If the aperture is not available in HW we can't use a ggtt slot and wc
copy, so fall back to regular kmap.

Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio at intel.com>
Signed-off-by: Abdiel Janulgue <abdiel.janulgue at linux.intel.com>
---
 drivers/gpu/drm/i915/i915_gem_gtt.c   | 16 ++++---
 drivers/gpu/drm/i915/i915_gpu_error.c | 63 ++++++++++++++++++++-------
 2 files changed, 57 insertions(+), 22 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
index 43b99136a3ae..3a8965048a06 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -2850,13 +2850,15 @@ static int init_ggtt(struct i915_ggtt *ggtt)
 	if (ret)
 		return ret;
 
-	/* Reserve a mappable slot for our lockless error capture */
-	ret = drm_mm_insert_node_in_range(&ggtt->vm.mm, &ggtt->error_capture,
-					  PAGE_SIZE, 0, I915_COLOR_UNEVICTABLE,
-					  0, ggtt->mappable_end,
-					  DRM_MM_INSERT_LOW);
-	if (ret)
-		return ret;
+	if (HAS_MAPPABLE_APERTURE(ggtt->vm.i915)) {
+		/* Reserve a mappable slot for our lockless error capture */
+		ret = drm_mm_insert_node_in_range(&ggtt->vm.mm, &ggtt->error_capture,
+						  PAGE_SIZE, 0, I915_COLOR_UNEVICTABLE,
+						  0, ggtt->mappable_end,
+						  DRM_MM_INSERT_LOW);
+		if (ret)
+			return ret;
+	}
 
 	/*
 	 * The upper portion of the GuC address space has a sizeable hole
diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c
index 5489cd879315..be920deb7ed7 100644
--- a/drivers/gpu/drm/i915/i915_gpu_error.c
+++ b/drivers/gpu/drm/i915/i915_gpu_error.c
@@ -218,7 +218,7 @@ struct compress {
 	void *tmp;
 };
 
-static bool compress_init(struct compress *c)
+static bool compress_init(struct compress *c, bool wc)
 {
 	struct z_stream_s *zstream = memset(&c->zstream, 0, sizeof(c->zstream));
 
@@ -234,7 +234,7 @@ static bool compress_init(struct compress *c)
 	}
 
 	c->tmp = NULL;
-	if (i915_has_memcpy_from_wc())
+	if (wc && i915_has_memcpy_from_wc())
 		c->tmp = (void *)__get_free_page(GFP_ATOMIC | __GFP_NOWARN);
 
 	return true;
@@ -335,10 +335,12 @@ static void err_compression_marker(struct drm_i915_error_state_buf *m)
 #else
 
 struct compress {
+	bool wc;
 };
 
-static bool compress_init(struct compress *c)
+static bool compress_init(struct compress *c, bool wc)
 {
+	c->wc = wc;
 	return true;
 }
 
@@ -354,7 +356,7 @@ static int compress_page(struct compress *c,
 		return -ENOMEM;
 
 	ptr = (void *)page;
-	if (!i915_memcpy_from_wc(ptr, src, PAGE_SIZE))
+	if (!(c->wc && i915_memcpy_from_wc(ptr, src, PAGE_SIZE)))
 		memcpy(ptr, src, PAGE_SIZE);
 	dst->pages[dst->page_count++] = ptr;
 
@@ -998,7 +1000,6 @@ i915_error_object_create(struct drm_i915_private *i915,
 	struct compress compress;
 	unsigned long num_pages;
 	struct sgt_iter iter;
-	dma_addr_t dma;
 	int ret;
 
 	if (!vma || !vma->pages)
@@ -1017,22 +1018,52 @@ i915_error_object_create(struct drm_i915_private *i915,
 	dst->page_count = 0;
 	dst->unused = 0;
 
-	if (!compress_init(&compress)) {
+	if (!compress_init(&compress, drm_mm_node_allocated(&ggtt->error_capture))) {
 		kfree(dst);
 		return NULL;
 	}
 
 	ret = -EINVAL;
-	for_each_sgt_dma(dma, iter, vma->pages) {
+	if (drm_mm_node_allocated(&ggtt->error_capture)) {
 		void __iomem *s;
+		dma_addr_t dma;
 
-		ggtt->vm.insert_page(&ggtt->vm, dma, slot, I915_CACHE_NONE, 0);
+		for_each_sgt_dma(dma, iter, vma->pages) {
+			ggtt->vm.insert_page(&ggtt->vm, dma, slot,
+					     I915_CACHE_NONE, 0);
 
-		s = io_mapping_map_atomic_wc(&ggtt->iomap, slot);
-		ret = compress_page(&compress, (void  __force *)s, dst);
-		io_mapping_unmap_atomic(s);
-		if (ret)
-			break;
+			s = io_mapping_map_atomic_wc(&ggtt->iomap, slot);
+			ret = compress_page(&compress, (void  __force *)s, dst);
+			io_mapping_unmap_atomic(s);
+
+			if (ret)
+				break;
+		}
+	} else if (i915_gem_object_is_lmem(vma->obj)) {
+		void *s;
+		dma_addr_t dma;
+		struct intel_memory_region *mem = vma->obj->memory_region;
+
+		for_each_sgt_dma(dma, iter, vma->pages) {
+			s = io_mapping_map_atomic_wc(&mem->iomap, dma);
+			ret = compress_page(&compress, s, dst);
+			io_mapping_unmap_atomic(s);
+
+			if (ret)
+				break;
+		}
+	} else {
+		void *s;
+		struct page *page;
+
+		for_each_sgt_page(page, iter, vma->pages) {
+			s = kmap_atomic(page);
+			ret = compress_page(&compress, s, dst);
+			kunmap_atomic(s);
+
+			if (ret)
+				break;
+		}
 	}
 
 	if (ret || compress_flush(&compress, dst)) {
@@ -1745,9 +1776,11 @@ static unsigned long capture_find_epoch(const struct i915_gpu_state *error)
 static void capture_finish(struct i915_gpu_state *error)
 {
 	struct i915_ggtt *ggtt = &error->i915->ggtt;
-	const u64 slot = ggtt->error_capture.start;
 
-	ggtt->vm.clear_range(&ggtt->vm, slot, PAGE_SIZE);
+	if (drm_mm_node_allocated(&ggtt->error_capture)) {
+		const u64 slot = ggtt->error_capture.start;
+		ggtt->vm.clear_range(&ggtt->vm, slot, PAGE_SIZE);
+	}
 }
 
 static int capture(void *data)
-- 
2.20.1