[PATCH 09/10] drm/i915: Remove the rounding down of the gen4+ fence region
Chris Wilson
chris at chris-wilson.co.uk
Fri Dec 23 18:56:15 UTC 2016
Restricting the fence to the end of the previous tile-row breaks access
to the final portion of the object. On gen2/3 we employed lazy fencing
to pad out the fence with scratch page to provide access to the tail,
and now we also pad out the object on gen4+ we can apply the same fix.
Fixes: af1a7301c7cf ("drm/i915: Only fence tiled region of object.")
Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
---
drivers/gpu/drm/i915/i915_gem_fence_reg.c | 51 ++++++++++++-------------------
1 file changed, 19 insertions(+), 32 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_gem_fence_reg.c b/drivers/gpu/drm/i915/i915_gem_fence_reg.c
index e03983973252..91da7888c8d4 100644
--- a/drivers/gpu/drm/i915/i915_gem_fence_reg.c
+++ b/drivers/gpu/drm/i915/i915_gem_fence_reg.c
@@ -78,15 +78,16 @@ static void i965_write_fence_reg(struct drm_i915_fence_reg *fence,
val = 0;
if (vma) {
unsigned int tiling = i915_gem_object_get_tiling(vma->obj);
- bool is_y_tiled = tiling == I915_TILING_Y;
unsigned int stride = i915_gem_object_get_stride(vma->obj);
- u32 row_size = stride * (is_y_tiled ? 32 : 8);
- u32 size = rounddown((u32)vma->node.size, row_size);
- val = ((vma->node.start + size - 4096) & 0xfffff000) << 32;
- val |= vma->node.start & 0xfffff000;
+ GEM_BUG_ON(vma->node.start & 4095);
+ GEM_BUG_ON(vma->node.size & 4095);
+ GEM_BUG_ON(stride & 127);
+
+ val = (vma->node.start + vma->node.size - 4096) << 32;
+ val |= vma->node.start;
val |= (u64)((stride / 128) - 1) << fence_pitch_shift;
- if (is_y_tiled)
+ if (tiling == I915_TILING_Y)
val |= BIT(I965_FENCE_TILING_Y_SHIFT);
val |= I965_FENCE_REG_VALID;
}
@@ -122,31 +123,22 @@ static void i915_write_fence_reg(struct drm_i915_fence_reg *fence,
unsigned int tiling = i915_gem_object_get_tiling(vma->obj);
bool is_y_tiled = tiling == I915_TILING_Y;
unsigned int stride = i915_gem_object_get_stride(vma->obj);
- int pitch_val;
- int tile_width;
- WARN((vma->node.start & ~I915_FENCE_START_MASK) ||
- !is_power_of_2(vma->node.size) ||
- (vma->node.start & (vma->node.size - 1)),
- "object 0x%08llx [fenceable? %d] not 1M or pot-size (0x%08llx) aligned\n",
- vma->node.start,
- i915_vma_is_map_and_fenceable(vma),
- vma->node.size);
+ GEM_BUG_ON(vma->node.start & ~I915_FENCE_START_MASK);
+ GEM_BUG_ON(!is_power_of_2(vma->node.size));
+ GEM_BUG_ON(!is_power_of_2(stride));
+ GEM_BUG_ON(vma->node.start & (vma->node.size - 1));
if (is_y_tiled && HAS_128_BYTE_Y_TILING(fence->i915))
- tile_width = 128;
+ stride /= 128;
else
- tile_width = 512;
-
- /* Note: pitch better be a power of two tile widths */
- pitch_val = stride / tile_width;
- pitch_val = ffs(pitch_val) - 1;
+ stride /= 512;
val = vma->node.start;
if (is_y_tiled)
val |= BIT(I830_FENCE_TILING_Y_SHIFT);
val |= I915_FENCE_SIZE_BITS(vma->node.size);
- val |= pitch_val << I830_FENCE_PITCH_SHIFT;
+ val |= ilog2(stride) << I830_FENCE_PITCH_SHIFT;
val |= I830_FENCE_REG_VALID;
}
@@ -169,22 +161,17 @@ static void i830_write_fence_reg(struct drm_i915_fence_reg *fence,
unsigned int tiling = i915_gem_object_get_tiling(vma->obj);
bool is_y_tiled = tiling == I915_TILING_Y;
unsigned int stride = i915_gem_object_get_stride(vma->obj);
- u32 pitch_val;
-
- WARN((vma->node.start & ~I830_FENCE_START_MASK) ||
- !is_power_of_2(vma->node.size) ||
- (vma->node.start & (vma->node.size - 1)),
- "object 0x%08llx not 512K or pot-size 0x%08llx aligned\n",
- vma->node.start, vma->node.size);
- pitch_val = stride / 128;
- pitch_val = ffs(pitch_val) - 1;
+ GEM_BUG_ON(vma->node.start & ~I830_FENCE_START_MASK);
+ GEM_BUG_ON(!is_power_of_2(vma->node.size));
+ GEM_BUG_ON(!is_power_of_2(stride));
+ GEM_BUG_ON(vma->node.start & (vma->node.size - 1));
val = vma->node.start;
if (is_y_tiled)
val |= BIT(I830_FENCE_TILING_Y_SHIFT);
val |= I830_FENCE_SIZE_BITS(vma->node.size);
- val |= pitch_val << I830_FENCE_PITCH_SHIFT;
+ val |= ilog2(stride / 128) << I830_FENCE_PITCH_SHIFT;
val |= I830_FENCE_REG_VALID;
}
--
2.11.0
More information about the Intel-gfx-trybot
mailing list