[Intel-gfx] [PATCH] drm/i915: Workaround async flip + VT-d corruption on HSW/BDW
Ville Syrjala
ville.syrjala at linux.intel.com
Sat Feb 20 10:33:03 UTC 2021
From: Ville Syrjälä <ville.syrjala at linux.intel.com>
On HSW/BDW with VT-d active the first tile row scanned out
after the first async flip of the frame often ends up corrupted.
Whether the corruption happens or not depends on the scanline
on which the async flip happens, but the behaviour seems very
consistent. Ie. the same set of scanlines (which are most scanlines)
always show the corruption. And another set of scanlines (far less
of them) never shows the corruption.
I discovered that disabling the fetch-stride stretching
feature cures the corruption. This is some kind of TLB related
prefetch thing AFAIK. We already disable it on SNB primary
planes due to a documented workaround. The hardware folks
indicated that disabling this should be fine, so let's go
with that.
And while we're here, let's document the relevant bits on all
pre-skl platforms.
Fixes: 2a636e240c77 ("drm/i915: Implement async flip for ivb/hsw")
Fixes: cda195f13abd ("drm/i915: Implement async flips for bdw")
Signed-off-by: Ville Syrjälä <ville.syrjala at linux.intel.com>
---
drivers/gpu/drm/i915/i915_reg.h | 23 ++++++++++++++++++++++-
drivers/gpu/drm/i915/intel_pm.c | 16 +++++++++++++++-
2 files changed, 37 insertions(+), 2 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
index 224ad897af34..0bb0229a2e49 100644
--- a/drivers/gpu/drm/i915/i915_reg.h
+++ b/drivers/gpu/drm/i915/i915_reg.h
@@ -3319,7 +3319,18 @@ static inline bool i915_mmio_reg_valid(i915_reg_t reg)
#define ILK_DISPLAY_CHICKEN1 _MMIO(0x42000)
#define ILK_FBCQ_DIS (1 << 22)
-#define ILK_PABSTRETCH_DIS (1 << 21)
+#define ILK_PABSTRETCH_DIS REG_BIT(21)
+#define ILK_SABSTRETCH_DIS REG_BIT(20)
+#define IVB_PRI_STRETCH_MAX_MASK REG_GENMASK(21, 20)
+#define IVB_PRI_STRETCH_MAX_X8 REG_FIELD_PREP(IVB_PRI_STRETCH_MAX_MASK, 0)
+#define IVB_PRI_STRETCH_MAX_X4 REG_FIELD_PREP(IVB_PRI_STRETCH_MAX_MASK, 1)
+#define IVB_PRI_STRETCH_MAX_X2 REG_FIELD_PREP(IVB_PRI_STRETCH_MAX_MASK, 2)
+#define IVB_PRI_STRETCH_MAX_X1 REG_FIELD_PREP(IVB_PRI_STRETCH_MAX_MASK, 3)
+#define IVB_SPR_STRETCH_MAX_MASK REG_GENMASK(19, 18)
+#define IVB_SPR_STRETCH_MAX_X8 REG_FIELD_PREP(IVB_SPR_STRETCH_MAX_MASK, 0)
+#define IVB_SPR_STRETCH_MAX_X4 REG_FIELD_PREP(IVB_SPR_STRETCH_MAX_MASK, 1)
+#define IVB_SPR_STRETCH_MAX_X2 REG_FIELD_PREP(IVB_SPR_STRETCH_MAX_MASK, 2)
+#define IVB_SPR_STRETCH_MAX_X1 REG_FIELD_PREP(IVB_SPR_STRETCH_MAX_MASK, 3)
/*
@@ -8042,6 +8053,16 @@ enum {
#define _CHICKEN_PIPESL_1_A 0x420b0
#define _CHICKEN_PIPESL_1_B 0x420b4
+#define HSW_PRI_STRETCH_MAX_MASK REG_GENMASK(28, 27)
+#define HSW_PRI_STRETCH_MAX_X8 REG_FIELD_PREP(HSW_PRI_STRETCH_MAX_MASK, 0)
+#define HSW_PRI_STRETCH_MAX_X4 REG_FIELD_PREP(HSW_PRI_STRETCH_MAX_MASK, 1)
+#define HSW_PRI_STRETCH_MAX_X2 REG_FIELD_PREP(HSW_PRI_STRETCH_MAX_MASK, 2)
+#define HSW_PRI_STRETCH_MAX_X1 REG_FIELD_PREP(HSW_PRI_STRETCH_MAX_MASK, 3)
+#define HSW_SPR_STRETCH_MAX_MASK REG_GENMASK(26, 25)
+#define HSW_SPR_STRETCH_MAX_X8 REG_FIELD_PREP(HSW_SPR_STRETCH_MAX_MASK, 0)
+#define HSW_SPR_STRETCH_MAX_X4 REG_FIELD_PREP(HSW_SPR_STRETCH_MAX_MASK, 1)
+#define HSW_SPR_STRETCH_MAX_X2 REG_FIELD_PREP(HSW_SPR_STRETCH_MAX_MASK, 2)
+#define HSW_SPR_STRETCH_MAX_X1 REG_FIELD_PREP(HSW_SPR_STRETCH_MAX_MASK, 3)
#define HSW_FBCQ_DIS (1 << 22)
#define BDW_DPRS_MASK_VBLANK_SRD (1 << 0)
#define CHICKEN_PIPESL_1(pipe) _MMIO_PIPE(pipe, _CHICKEN_PIPESL_1_A, _CHICKEN_PIPESL_1_B)
diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c
index 8cc67f9c4e58..53b04800f9d0 100644
--- a/drivers/gpu/drm/i915/intel_pm.c
+++ b/drivers/gpu/drm/i915/intel_pm.c
@@ -7243,11 +7243,16 @@ static void bdw_init_clock_gating(struct drm_i915_private *dev_priv)
intel_uncore_write(&dev_priv->uncore, CHICKEN_PAR1_1,
intel_uncore_read(&dev_priv->uncore, CHICKEN_PAR1_1) | DPA_MASK_VBLANK_SRD);
- /* WaPsrDPRSUnmaskVBlankInSRD:bdw */
for_each_pipe(dev_priv, pipe) {
+ /* WaPsrDPRSUnmaskVBlankInSRD:bdw */
intel_uncore_write(&dev_priv->uncore, CHICKEN_PIPESL_1(pipe),
intel_uncore_read(&dev_priv->uncore, CHICKEN_PIPESL_1(pipe)) |
BDW_DPRS_MASK_VBLANK_SRD);
+
+ /* Undocumented but fixes async flip + VT-d corruption */
+ if (intel_vtd_active())
+ intel_uncore_rmw(&dev_priv->uncore, CHICKEN_PIPESL_1(pipe),
+ HSW_PRI_STRETCH_MAX_MASK, HSW_PRI_STRETCH_MAX_X1);
}
/* WaVSRefCountFullforceMissDisable:bdw */
@@ -7283,11 +7288,20 @@ static void bdw_init_clock_gating(struct drm_i915_private *dev_priv)
static void hsw_init_clock_gating(struct drm_i915_private *dev_priv)
{
+ enum pipe pipe;
+
/* WaFbcAsynchFlipDisableFbcQueue:hsw,bdw */
intel_uncore_write(&dev_priv->uncore, CHICKEN_PIPESL_1(PIPE_A),
intel_uncore_read(&dev_priv->uncore, CHICKEN_PIPESL_1(PIPE_A)) |
HSW_FBCQ_DIS);
+ for_each_pipe(dev_priv, pipe) {
+ /* Undocumented but fixes async flip + VT-d corruption */
+ if (intel_vtd_active())
+ intel_uncore_rmw(&dev_priv->uncore, CHICKEN_PIPESL_1(pipe),
+ HSW_PRI_STRETCH_MAX_MASK, HSW_PRI_STRETCH_MAX_X1);
+ }
+
/* This is required by WaCatErrorRejectionIssue:hsw */
intel_uncore_write(&dev_priv->uncore, GEN7_SQ_CHICKEN_MBCUNIT_CONFIG,
intel_uncore_read(&dev_priv->uncore, GEN7_SQ_CHICKEN_MBCUNIT_CONFIG) |
--
2.26.2
More information about the Intel-gfx
mailing list