[Intel-gfx] [PATCH] drm/i915: fixup in-line clflushing on bit17 swizzled bos

[Daniel Vetter]

The issue is that with inline clflushing the clflushing isn't properly
swizzled. Fix this by
- always clflushing entire 128 byte chunks and
- unconditionally flush before writes when swizzling a given page.
  We could be clever and check whether we pwrite a partial 128 byte
  chunk instead of a partial cacheline, but I've figured that's not
  worth it.

Now the usual approach is to fold this into the original patch series, but
I've opted against this because
- this fixes a corner case only very old userspace relies on and
- I'd like to not invalidate all the testing the pwrite rewrite has gotten.

This fixes the regression notice by tests/gem_tiled_partial_prite_pread
from i-g-t. Unfortunately it doesn't fix the issues with partial pwrites to
tiled buffers on bit17 swizzling machines. But that is also broken without
the pwrite patches, so likely a different issue (or a problem with the
testcase).

v2: Simplify the patch by dropping the overly clever partial write
logic for swizzled pages.

Signed-Off-by: Daniel Vetter <daniel.vetter at ffwll.ch>
---
 drivers/gpu/drm/i915/i915_gem.c |   39 ++++++++++++++++++++++++++++++++-------
 1 files changed, 32 insertions(+), 7 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 9b200f4e..c4f657c 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -333,6 +333,28 @@ shmem_pread_fast(struct page *page, int shmem_page_offset, int page_length,
 	return ret;
 }
 
+static void
+shmem_clflush_swizzled_range(char *addr, unsigned long length,
+			     bool swizzled)
+{
+	if (swizzled) {
+		unsigned long start = (unsigned long) addr;
+		unsigned long end = (unsigned long) addr + length;
+
+		/* For swizzling simply ensure that we always flush both
+		 * channels. Lame, but simple and it works. Swizzled
+		 * pwrite/pread is far from a hotpath - current userspace
+		 * doesn't use it at all. */
+		start = round_down(start, 128);
+		end = round_up(end, 128);
+
+		drm_clflush_virt_range((void *)start, end - start);
+	} else {
+		drm_clflush_virt_range(addr, length);
+	}
+
+}
+
 /* Only difference to the fast-path function is that this can handle bit17
  * and uses non-atomic copy and kmap functions. */
 static int
@@ -345,8 +367,9 @@ shmem_pread_slow(struct page *page, int shmem_page_offset, int page_length,
 
 	vaddr = kmap(page);
 	if (needs_clflush)
-		drm_clflush_virt_range(vaddr + shmem_page_offset,
-				       page_length);
+		shmem_clflush_swizzled_range(vaddr + shmem_page_offset,
+					     page_length,
+					     page_do_bit17_swizzling);
 
 	if (page_do_bit17_swizzling)
 		ret = __copy_to_user_swizzled(user_data,
@@ -655,9 +678,10 @@ shmem_pwrite_slow(struct page *page, int shmem_page_offset, int page_length,
 	int ret;
 
 	vaddr = kmap(page);
-	if (needs_clflush_before)
-		drm_clflush_virt_range(vaddr + shmem_page_offset,
-				       page_length);
+	if (needs_clflush_before || page_do_bit17_swizzling)
+		shmem_clflush_swizzled_range(vaddr + shmem_page_offset,
+					     page_length,
+					     page_do_bit17_swizzling);
 	if (page_do_bit17_swizzling)
 		ret = __copy_from_user_swizzled(vaddr, shmem_page_offset,
 						user_data,
@@ -667,8 +691,9 @@ shmem_pwrite_slow(struct page *page, int shmem_page_offset, int page_length,
 				       user_data,
 				       page_length);
 	if (needs_clflush)
-		drm_clflush_virt_range(vaddr + shmem_page_offset,
-				       page_length);
+		shmem_clflush_swizzled_range(vaddr + shmem_page_offset,
+					     page_length,
+					     page_do_bit17_swizzling);
 	kunmap(page);
 
 	return ret;
-- 
1.7.8.3