* [PATCH 0/5] Assorted engine tidy
@ 2017-02-16 12:23 Tvrtko Ursulin
2017-02-16 12:23 ` [PATCH 1/5] drm/i915: Make int __intel_ring_space static Tvrtko Ursulin
` (6 more replies)
0 siblings, 7 replies; 21+ messages in thread
From: Tvrtko Ursulin @ 2017-02-16 12:23 UTC (permalink / raw)
To: Intel-gfx
From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
A few previously posted patches to tidy the code, extract some commonality,
and make workaround batch buffer emission more compact and readable.
text data bss dec hex filename
1080898 24080 2852 1107830 10e776 i915.ko.0
1080070 24080 2852 1107002 10e43a i915.ko.1
Tvrtko Ursulin (5):
drm/i915: Make int __intel_ring_space static
drm/i915: Simplify cleanup path in intel_engines_init
drm/i915: Move common workaround code to intel_engine_cs
drm/i915: Tidy workaround batch buffer emission
drm/i915: Consolidate gen8_emit_pipe_control
drivers/gpu/drm/i915/intel_engine_cs.c | 554 ++++++++++++++++++++++++++++-
drivers/gpu/drm/i915/intel_lrc.c | 351 +++++++------------
drivers/gpu/drm/i915/intel_ringbuffer.c | 597 +-------------------------------
drivers/gpu/drm/i915/intel_ringbuffer.h | 12 +-
4 files changed, 696 insertions(+), 818 deletions(-)
--
2.9.3
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 21+ messages in thread
* [PATCH 1/5] drm/i915: Make int __intel_ring_space static
2017-02-16 12:23 [PATCH 0/5] Assorted engine tidy Tvrtko Ursulin
@ 2017-02-16 12:23 ` Tvrtko Ursulin
2017-02-16 13:12 ` Chris Wilson
2017-02-16 12:23 ` [PATCH 2/5] drm/i915: Simplify cleanup path in intel_engines_init Tvrtko Ursulin
` (5 subsequent siblings)
6 siblings, 1 reply; 21+ messages in thread
From: Tvrtko Ursulin @ 2017-02-16 12:23 UTC (permalink / raw)
To: Intel-gfx
From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
It is only used within intel_ringbuffer.c
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
---
drivers/gpu/drm/i915/intel_ringbuffer.c | 2 +-
drivers/gpu/drm/i915/intel_ringbuffer.h | 1 -
2 files changed, 1 insertion(+), 2 deletions(-)
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c
index f7499829ecc2..8c17db72489f 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.c
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.c
@@ -39,7 +39,7 @@
*/
#define LEGACY_REQUEST_SIZE 200
-int __intel_ring_space(int head, int tail, int size)
+static int __intel_ring_space(int head, int tail, int size)
{
int space = head - tail;
if (space <= 0)
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h
index 16714096810d..1d9fd4e42622 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.h
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.h
@@ -523,7 +523,6 @@ intel_ring_offset(struct drm_i915_gem_request *req, void *addr)
return offset & (req->ring->size - 1);
}
-int __intel_ring_space(int head, int tail, int size);
void intel_ring_update_space(struct intel_ring *ring);
void intel_engine_init_global_seqno(struct intel_engine_cs *engine, u32 seqno);
--
2.9.3
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 21+ messages in thread
* [PATCH 2/5] drm/i915: Simplify cleanup path in intel_engines_init
2017-02-16 12:23 [PATCH 0/5] Assorted engine tidy Tvrtko Ursulin
2017-02-16 12:23 ` [PATCH 1/5] drm/i915: Make int __intel_ring_space static Tvrtko Ursulin
@ 2017-02-16 12:23 ` Tvrtko Ursulin
2017-02-16 13:15 ` Chris Wilson
2017-02-16 12:23 ` [PATCH 3/5] drm/i915: Move common workaround code to intel_engine_cs Tvrtko Ursulin
` (4 subsequent siblings)
6 siblings, 1 reply; 21+ messages in thread
From: Tvrtko Ursulin @ 2017-02-16 12:23 UTC (permalink / raw)
To: Intel-gfx
From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
We can call the engine cleanup vfunc instead of duplicating the
decision making here.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
---
drivers/gpu/drm/i915/intel_engine_cs.c | 4 +---
1 file changed, 1 insertion(+), 3 deletions(-)
diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c
index 538d845d7251..012ac9483491 100644
--- a/drivers/gpu/drm/i915/intel_engine_cs.c
+++ b/drivers/gpu/drm/i915/intel_engine_cs.c
@@ -210,10 +210,8 @@ int intel_engines_init(struct drm_i915_private *dev_priv)
for_each_engine(engine, dev_priv, id) {
if (id >= err_id)
kfree(engine);
- else if (i915.enable_execlists)
- intel_logical_ring_cleanup(engine);
else
- intel_engine_cleanup(engine);
+ dev_priv->gt.cleanup_engine(engine);
}
return err;
}
--
2.9.3
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 21+ messages in thread
* [PATCH 3/5] drm/i915: Move common workaround code to intel_engine_cs
2017-02-16 12:23 [PATCH 0/5] Assorted engine tidy Tvrtko Ursulin
2017-02-16 12:23 ` [PATCH 1/5] drm/i915: Make int __intel_ring_space static Tvrtko Ursulin
2017-02-16 12:23 ` [PATCH 2/5] drm/i915: Simplify cleanup path in intel_engines_init Tvrtko Ursulin
@ 2017-02-16 12:23 ` Tvrtko Ursulin
2017-02-16 13:18 ` Chris Wilson
2017-02-16 12:23 ` [PATCH 4/5] drm/i915: Tidy workaround batch buffer emission Tvrtko Ursulin
` (3 subsequent siblings)
6 siblings, 1 reply; 21+ messages in thread
From: Tvrtko Ursulin @ 2017-02-16 12:23 UTC (permalink / raw)
To: Intel-gfx
From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
It is used by all submission backends.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
---
drivers/gpu/drm/i915/intel_engine_cs.c | 550 ++++++++++++++++++++++++++++++++
drivers/gpu/drm/i915/intel_ringbuffer.c | 550 --------------------------------
2 files changed, 550 insertions(+), 550 deletions(-)
diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c
index 012ac9483491..46c94740be53 100644
--- a/drivers/gpu/drm/i915/intel_engine_cs.c
+++ b/drivers/gpu/drm/i915/intel_engine_cs.c
@@ -523,6 +523,556 @@ void intel_engine_get_instdone(struct intel_engine_cs *engine,
}
}
+static int wa_add(struct drm_i915_private *dev_priv,
+ i915_reg_t addr,
+ const u32 mask, const u32 val)
+{
+ const u32 idx = dev_priv->workarounds.count;
+
+ if (WARN_ON(idx >= I915_MAX_WA_REGS))
+ return -ENOSPC;
+
+ dev_priv->workarounds.reg[idx].addr = addr;
+ dev_priv->workarounds.reg[idx].value = val;
+ dev_priv->workarounds.reg[idx].mask = mask;
+
+ dev_priv->workarounds.count++;
+
+ return 0;
+}
+
+#define WA_REG(addr, mask, val) do { \
+ const int r = wa_add(dev_priv, (addr), (mask), (val)); \
+ if (r) \
+ return r; \
+ } while (0)
+
+#define WA_SET_BIT_MASKED(addr, mask) \
+ WA_REG(addr, (mask), _MASKED_BIT_ENABLE(mask))
+
+#define WA_CLR_BIT_MASKED(addr, mask) \
+ WA_REG(addr, (mask), _MASKED_BIT_DISABLE(mask))
+
+#define WA_SET_FIELD_MASKED(addr, mask, value) \
+ WA_REG(addr, mask, _MASKED_FIELD(mask, value))
+
+#define WA_SET_BIT(addr, mask) WA_REG(addr, mask, I915_READ(addr) | (mask))
+#define WA_CLR_BIT(addr, mask) WA_REG(addr, mask, I915_READ(addr) & ~(mask))
+
+#define WA_WRITE(addr, val) WA_REG(addr, 0xffffffff, val)
+
+static int wa_ring_whitelist_reg(struct intel_engine_cs *engine,
+ i915_reg_t reg)
+{
+ struct drm_i915_private *dev_priv = engine->i915;
+ struct i915_workarounds *wa = &dev_priv->workarounds;
+ const uint32_t index = wa->hw_whitelist_count[engine->id];
+
+ if (WARN_ON(index >= RING_MAX_NONPRIV_SLOTS))
+ return -EINVAL;
+
+ WA_WRITE(RING_FORCE_TO_NONPRIV(engine->mmio_base, index),
+ i915_mmio_reg_offset(reg));
+ wa->hw_whitelist_count[engine->id]++;
+
+ return 0;
+}
+
+static int gen8_init_workarounds(struct intel_engine_cs *engine)
+{
+ struct drm_i915_private *dev_priv = engine->i915;
+
+ WA_SET_BIT_MASKED(INSTPM, INSTPM_FORCE_ORDERING);
+
+ /* WaDisableAsyncFlipPerfMode:bdw,chv */
+ WA_SET_BIT_MASKED(MI_MODE, ASYNC_FLIP_PERF_DISABLE);
+
+ /* WaDisablePartialInstShootdown:bdw,chv */
+ WA_SET_BIT_MASKED(GEN8_ROW_CHICKEN,
+ PARTIAL_INSTRUCTION_SHOOTDOWN_DISABLE);
+
+ /* Use Force Non-Coherent whenever executing a 3D context. This is a
+ * workaround for for a possible hang in the unlikely event a TLB
+ * invalidation occurs during a PSD flush.
+ */
+ /* WaForceEnableNonCoherent:bdw,chv */
+ /* WaHdcDisableFetchWhenMasked:bdw,chv */
+ WA_SET_BIT_MASKED(HDC_CHICKEN0,
+ HDC_DONOT_FETCH_MEM_WHEN_MASKED |
+ HDC_FORCE_NON_COHERENT);
+
+ /* From the Haswell PRM, Command Reference: Registers, CACHE_MODE_0:
+ * "The Hierarchical Z RAW Stall Optimization allows non-overlapping
+ * polygons in the same 8x4 pixel/sample area to be processed without
+ * stalling waiting for the earlier ones to write to Hierarchical Z
+ * buffer."
+ *
+ * This optimization is off by default for BDW and CHV; turn it on.
+ */
+ WA_CLR_BIT_MASKED(CACHE_MODE_0_GEN7, HIZ_RAW_STALL_OPT_DISABLE);
+
+ /* Wa4x4STCOptimizationDisable:bdw,chv */
+ WA_SET_BIT_MASKED(CACHE_MODE_1, GEN8_4x4_STC_OPTIMIZATION_DISABLE);
+
+ /*
+ * BSpec recommends 8x4 when MSAA is used,
+ * however in practice 16x4 seems fastest.
+ *
+ * Note that PS/WM thread counts depend on the WIZ hashing
+ * disable bit, which we don't touch here, but it's good
+ * to keep in mind (see 3DSTATE_PS and 3DSTATE_WM).
+ */
+ WA_SET_FIELD_MASKED(GEN7_GT_MODE,
+ GEN6_WIZ_HASHING_MASK,
+ GEN6_WIZ_HASHING_16x4);
+
+ return 0;
+}
+
+static int bdw_init_workarounds(struct intel_engine_cs *engine)
+{
+ struct drm_i915_private *dev_priv = engine->i915;
+ int ret;
+
+ ret = gen8_init_workarounds(engine);
+ if (ret)
+ return ret;
+
+ /* WaDisableThreadStallDopClockGating:bdw (pre-production) */
+ WA_SET_BIT_MASKED(GEN8_ROW_CHICKEN, STALL_DOP_GATING_DISABLE);
+
+ /* WaDisableDopClockGating:bdw
+ *
+ * Also see the related UCGTCL1 write in broadwell_init_clock_gating()
+ * to disable EUTC clock gating.
+ */
+ WA_SET_BIT_MASKED(GEN7_ROW_CHICKEN2,
+ DOP_CLOCK_GATING_DISABLE);
+
+ WA_SET_BIT_MASKED(HALF_SLICE_CHICKEN3,
+ GEN8_SAMPLER_POWER_BYPASS_DIS);
+
+ WA_SET_BIT_MASKED(HDC_CHICKEN0,
+ /* WaForceContextSaveRestoreNonCoherent:bdw */
+ HDC_FORCE_CONTEXT_SAVE_RESTORE_NON_COHERENT |
+ /* WaDisableFenceDestinationToSLM:bdw (pre-prod) */
+ (IS_BDW_GT3(dev_priv) ? HDC_FENCE_DEST_SLM_DISABLE : 0));
+
+ return 0;
+}
+
+static int chv_init_workarounds(struct intel_engine_cs *engine)
+{
+ struct drm_i915_private *dev_priv = engine->i915;
+ int ret;
+
+ ret = gen8_init_workarounds(engine);
+ if (ret)
+ return ret;
+
+ /* WaDisableThreadStallDopClockGating:chv */
+ WA_SET_BIT_MASKED(GEN8_ROW_CHICKEN, STALL_DOP_GATING_DISABLE);
+
+ /* Improve HiZ throughput on CHV. */
+ WA_SET_BIT_MASKED(HIZ_CHICKEN, CHV_HZ_8X8_MODE_IN_1X);
+
+ return 0;
+}
+
+static int gen9_init_workarounds(struct intel_engine_cs *engine)
+{
+ struct drm_i915_private *dev_priv = engine->i915;
+ int ret;
+
+ /* WaConextSwitchWithConcurrentTLBInvalidate:skl,bxt,kbl,glk */
+ I915_WRITE(GEN9_CSFE_CHICKEN1_RCS, _MASKED_BIT_ENABLE(GEN9_PREEMPT_GPGPU_SYNC_SWITCH_DISABLE));
+
+ /* WaEnableLbsSlaRetryTimerDecrement:skl,bxt,kbl,glk */
+ I915_WRITE(BDW_SCRATCH1, I915_READ(BDW_SCRATCH1) |
+ GEN9_LBS_SLA_RETRY_TIMER_DECREMENT_ENABLE);
+
+ /* WaDisableKillLogic:bxt,skl,kbl */
+ I915_WRITE(GAM_ECOCHK, I915_READ(GAM_ECOCHK) |
+ ECOCHK_DIS_TLB);
+
+ /* WaClearFlowControlGpgpuContextSave:skl,bxt,kbl,glk */
+ /* WaDisablePartialInstShootdown:skl,bxt,kbl,glk */
+ WA_SET_BIT_MASKED(GEN8_ROW_CHICKEN,
+ FLOW_CONTROL_ENABLE |
+ PARTIAL_INSTRUCTION_SHOOTDOWN_DISABLE);
+
+ /* Syncing dependencies between camera and graphics:skl,bxt,kbl */
+ WA_SET_BIT_MASKED(HALF_SLICE_CHICKEN3,
+ GEN9_DISABLE_OCL_OOB_SUPPRESS_LOGIC);
+
+ /* WaDisableDgMirrorFixInHalfSliceChicken5:bxt */
+ if (IS_BXT_REVID(dev_priv, 0, BXT_REVID_A1))
+ WA_CLR_BIT_MASKED(GEN9_HALF_SLICE_CHICKEN5,
+ GEN9_DG_MIRROR_FIX_ENABLE);
+
+ /* WaSetDisablePixMaskCammingAndRhwoInCommonSliceChicken:bxt */
+ if (IS_BXT_REVID(dev_priv, 0, BXT_REVID_A1)) {
+ WA_SET_BIT_MASKED(GEN7_COMMON_SLICE_CHICKEN1,
+ GEN9_RHWO_OPTIMIZATION_DISABLE);
+ /*
+ * WA also requires GEN9_SLICE_COMMON_ECO_CHICKEN0[14:14] to be set
+ * but we do that in per ctx batchbuffer as there is an issue
+ * with this register not getting restored on ctx restore
+ */
+ }
+
+ /* WaEnableSamplerGPGPUPreemptionSupport:skl,bxt,kbl */
+ WA_SET_BIT_MASKED(GEN9_HALF_SLICE_CHICKEN7,
+ GEN9_ENABLE_GPGPU_PREEMPTION);
+
+ /* Wa4x4STCOptimizationDisable:skl,bxt,kbl,glk */
+ /* WaDisablePartialResolveInVc:skl,bxt,kbl */
+ WA_SET_BIT_MASKED(CACHE_MODE_1, (GEN8_4x4_STC_OPTIMIZATION_DISABLE |
+ GEN9_PARTIAL_RESOLVE_IN_VC_DISABLE));
+
+ /* WaCcsTlbPrefetchDisable:skl,bxt,kbl,glk */
+ WA_CLR_BIT_MASKED(GEN9_HALF_SLICE_CHICKEN5,
+ GEN9_CCS_TLB_PREFETCH_ENABLE);
+
+ /* WaDisableMaskBasedCammingInRCC:bxt */
+ if (IS_BXT_REVID(dev_priv, 0, BXT_REVID_A1))
+ WA_SET_BIT_MASKED(SLICE_ECO_CHICKEN0,
+ PIXEL_MASK_CAMMING_DISABLE);
+
+ /* WaForceContextSaveRestoreNonCoherent:skl,bxt,kbl */
+ WA_SET_BIT_MASKED(HDC_CHICKEN0,
+ HDC_FORCE_CONTEXT_SAVE_RESTORE_NON_COHERENT |
+ HDC_FORCE_CSR_NON_COHERENT_OVR_DISABLE);
+
+ /* WaForceEnableNonCoherent and WaDisableHDCInvalidation are
+ * both tied to WaForceContextSaveRestoreNonCoherent
+ * in some hsds for skl. We keep the tie for all gen9. The
+ * documentation is a bit hazy and so we want to get common behaviour,
+ * even though there is no clear evidence we would need both on kbl/bxt.
+ * This area has been source of system hangs so we play it safe
+ * and mimic the skl regardless of what bspec says.
+ *
+ * Use Force Non-Coherent whenever executing a 3D context. This
+ * is a workaround for a possible hang in the unlikely event
+ * a TLB invalidation occurs during a PSD flush.
+ */
+
+ /* WaForceEnableNonCoherent:skl,bxt,kbl */
+ WA_SET_BIT_MASKED(HDC_CHICKEN0,
+ HDC_FORCE_NON_COHERENT);
+
+ /* WaDisableHDCInvalidation:skl,bxt,kbl */
+ I915_WRITE(GAM_ECOCHK, I915_READ(GAM_ECOCHK) |
+ BDW_DISABLE_HDC_INVALIDATION);
+
+ /* WaDisableSamplerPowerBypassForSOPingPong:skl,bxt,kbl */
+ if (IS_SKYLAKE(dev_priv) ||
+ IS_KABYLAKE(dev_priv) ||
+ IS_BXT_REVID(dev_priv, 0, BXT_REVID_B0))
+ WA_SET_BIT_MASKED(HALF_SLICE_CHICKEN3,
+ GEN8_SAMPLER_POWER_BYPASS_DIS);
+
+ /* WaDisableSTUnitPowerOptimization:skl,bxt,kbl,glk */
+ WA_SET_BIT_MASKED(HALF_SLICE_CHICKEN2, GEN8_ST_PO_DISABLE);
+
+ /* WaOCLCoherentLineFlush:skl,bxt,kbl */
+ I915_WRITE(GEN8_L3SQCREG4, (I915_READ(GEN8_L3SQCREG4) |
+ GEN8_LQSC_FLUSH_COHERENT_LINES));
+
+ /* WaVFEStateAfterPipeControlwithMediaStateClear:skl,bxt,glk */
+ ret = wa_ring_whitelist_reg(engine, GEN9_CTX_PREEMPT_REG);
+ if (ret)
+ return ret;
+
+ /* WaEnablePreemptionGranularityControlByUMD:skl,bxt,kbl */
+ ret= wa_ring_whitelist_reg(engine, GEN8_CS_CHICKEN1);
+ if (ret)
+ return ret;
+
+ /* WaAllowUMDToModifyHDCChicken1:skl,bxt,kbl,glk */
+ ret = wa_ring_whitelist_reg(engine, GEN8_HDC_CHICKEN1);
+ if (ret)
+ return ret;
+
+ return 0;
+}
+
+static int skl_tune_iz_hashing(struct intel_engine_cs *engine)
+{
+ struct drm_i915_private *dev_priv = engine->i915;
+ u8 vals[3] = { 0, 0, 0 };
+ unsigned int i;
+
+ for (i = 0; i < 3; i++) {
+ u8 ss;
+
+ /*
+ * Only consider slices where one, and only one, subslice has 7
+ * EUs
+ */
+ if (!is_power_of_2(INTEL_INFO(dev_priv)->sseu.subslice_7eu[i]))
+ continue;
+
+ /*
+ * subslice_7eu[i] != 0 (because of the check above) and
+ * ss_max == 4 (maximum number of subslices possible per slice)
+ *
+ * -> 0 <= ss <= 3;
+ */
+ ss = ffs(INTEL_INFO(dev_priv)->sseu.subslice_7eu[i]) - 1;
+ vals[i] = 3 - ss;
+ }
+
+ if (vals[0] == 0 && vals[1] == 0 && vals[2] == 0)
+ return 0;
+
+ /* Tune IZ hashing. See intel_device_info_runtime_init() */
+ WA_SET_FIELD_MASKED(GEN7_GT_MODE,
+ GEN9_IZ_HASHING_MASK(2) |
+ GEN9_IZ_HASHING_MASK(1) |
+ GEN9_IZ_HASHING_MASK(0),
+ GEN9_IZ_HASHING(2, vals[2]) |
+ GEN9_IZ_HASHING(1, vals[1]) |
+ GEN9_IZ_HASHING(0, vals[0]));
+
+ return 0;
+}
+
+static int skl_init_workarounds(struct intel_engine_cs *engine)
+{
+ struct drm_i915_private *dev_priv = engine->i915;
+ int ret;
+
+ ret = gen9_init_workarounds(engine);
+ if (ret)
+ return ret;
+
+ /*
+ * Actual WA is to disable percontext preemption granularity control
+ * until D0 which is the default case so this is equivalent to
+ * !WaDisablePerCtxtPreemptionGranularityControl:skl
+ */
+ I915_WRITE(GEN7_FF_SLICE_CS_CHICKEN1,
+ _MASKED_BIT_ENABLE(GEN9_FFSC_PERCTX_PREEMPT_CTRL));
+
+ /* WaEnableGapsTsvCreditFix:skl */
+ I915_WRITE(GEN8_GARBCNTL, (I915_READ(GEN8_GARBCNTL) |
+ GEN9_GAPS_TSV_CREDIT_DISABLE));
+
+ /* WaDisableGafsUnitClkGating:skl */
+ WA_SET_BIT(GEN7_UCGCTL4, GEN8_EU_GAUNIT_CLOCK_GATE_DISABLE);
+
+ /* WaInPlaceDecompressionHang:skl */
+ if (IS_SKL_REVID(dev_priv, SKL_REVID_H0, REVID_FOREVER))
+ WA_SET_BIT(GEN9_GAMT_ECO_REG_RW_IA,
+ GAMT_ECO_ENABLE_IN_PLACE_DECOMPRESS);
+
+ /* WaDisableLSQCROPERFforOCL:skl */
+ ret = wa_ring_whitelist_reg(engine, GEN8_L3SQCREG4);
+ if (ret)
+ return ret;
+
+ return skl_tune_iz_hashing(engine);
+}
+
+static int bxt_init_workarounds(struct intel_engine_cs *engine)
+{
+ struct drm_i915_private *dev_priv = engine->i915;
+ int ret;
+
+ ret = gen9_init_workarounds(engine);
+ if (ret)
+ return ret;
+
+ /* WaStoreMultiplePTEenable:bxt */
+ /* This is a requirement according to Hardware specification */
+ if (IS_BXT_REVID(dev_priv, 0, BXT_REVID_A1))
+ I915_WRITE(TILECTL, I915_READ(TILECTL) | TILECTL_TLBPF);
+
+ /* WaSetClckGatingDisableMedia:bxt */
+ if (IS_BXT_REVID(dev_priv, 0, BXT_REVID_A1)) {
+ I915_WRITE(GEN7_MISCCPCTL, (I915_READ(GEN7_MISCCPCTL) &
+ ~GEN8_DOP_CLOCK_GATE_MEDIA_ENABLE));
+ }
+
+ /* WaDisableThreadStallDopClockGating:bxt */
+ WA_SET_BIT_MASKED(GEN8_ROW_CHICKEN,
+ STALL_DOP_GATING_DISABLE);
+
+ /* WaDisablePooledEuLoadBalancingFix:bxt */
+ if (IS_BXT_REVID(dev_priv, BXT_REVID_B0, REVID_FOREVER)) {
+ WA_SET_BIT_MASKED(FF_SLICE_CS_CHICKEN2,
+ GEN9_POOLED_EU_LOAD_BALANCING_FIX_DISABLE);
+ }
+
+ /* WaDisableSbeCacheDispatchPortSharing:bxt */
+ if (IS_BXT_REVID(dev_priv, 0, BXT_REVID_B0)) {
+ WA_SET_BIT_MASKED(
+ GEN7_HALF_SLICE_CHICKEN1,
+ GEN7_SBE_SS_CACHE_DISPATCH_PORT_SHARING_DISABLE);
+ }
+
+ /* WaDisableObjectLevelPreemptionForTrifanOrPolygon:bxt */
+ /* WaDisableObjectLevelPreemptionForInstancedDraw:bxt */
+ /* WaDisableObjectLevelPreemtionForInstanceId:bxt */
+ /* WaDisableLSQCROPERFforOCL:bxt */
+ if (IS_BXT_REVID(dev_priv, 0, BXT_REVID_A1)) {
+ ret = wa_ring_whitelist_reg(engine, GEN9_CS_DEBUG_MODE1);
+ if (ret)
+ return ret;
+
+ ret = wa_ring_whitelist_reg(engine, GEN8_L3SQCREG4);
+ if (ret)
+ return ret;
+ }
+
+ /* WaProgramL3SqcReg1DefaultForPerf:bxt */
+ if (IS_BXT_REVID(dev_priv, BXT_REVID_B0, REVID_FOREVER))
+ I915_WRITE(GEN8_L3SQCREG1, L3_GENERAL_PRIO_CREDITS(62) |
+ L3_HIGH_PRIO_CREDITS(2));
+
+ /* WaToEnableHwFixForPushConstHWBug:bxt */
+ if (IS_BXT_REVID(dev_priv, BXT_REVID_C0, REVID_FOREVER))
+ WA_SET_BIT_MASKED(COMMON_SLICE_CHICKEN2,
+ GEN8_SBE_DISABLE_REPLAY_BUF_OPTIMIZATION);
+
+ /* WaInPlaceDecompressionHang:bxt */
+ if (IS_BXT_REVID(dev_priv, BXT_REVID_C0, REVID_FOREVER))
+ WA_SET_BIT(GEN9_GAMT_ECO_REG_RW_IA,
+ GAMT_ECO_ENABLE_IN_PLACE_DECOMPRESS);
+
+ return 0;
+}
+
+static int kbl_init_workarounds(struct intel_engine_cs *engine)
+{
+ struct drm_i915_private *dev_priv = engine->i915;
+ int ret;
+
+ ret = gen9_init_workarounds(engine);
+ if (ret)
+ return ret;
+
+ /* WaEnableGapsTsvCreditFix:kbl */
+ I915_WRITE(GEN8_GARBCNTL, (I915_READ(GEN8_GARBCNTL) |
+ GEN9_GAPS_TSV_CREDIT_DISABLE));
+
+ /* WaDisableDynamicCreditSharing:kbl */
+ if (IS_KBL_REVID(dev_priv, 0, KBL_REVID_B0))
+ WA_SET_BIT(GAMT_CHKN_BIT_REG,
+ GAMT_CHKN_DISABLE_DYNAMIC_CREDIT_SHARING);
+
+ /* WaDisableFenceDestinationToSLM:kbl (pre-prod) */
+ if (IS_KBL_REVID(dev_priv, KBL_REVID_A0, KBL_REVID_A0))
+ WA_SET_BIT_MASKED(HDC_CHICKEN0,
+ HDC_FENCE_DEST_SLM_DISABLE);
+
+ /* WaToEnableHwFixForPushConstHWBug:kbl */
+ if (IS_KBL_REVID(dev_priv, KBL_REVID_C0, REVID_FOREVER))
+ WA_SET_BIT_MASKED(COMMON_SLICE_CHICKEN2,
+ GEN8_SBE_DISABLE_REPLAY_BUF_OPTIMIZATION);
+
+ /* WaDisableGafsUnitClkGating:kbl */
+ WA_SET_BIT(GEN7_UCGCTL4, GEN8_EU_GAUNIT_CLOCK_GATE_DISABLE);
+
+ /* WaDisableSbeCacheDispatchPortSharing:kbl */
+ WA_SET_BIT_MASKED(
+ GEN7_HALF_SLICE_CHICKEN1,
+ GEN7_SBE_SS_CACHE_DISPATCH_PORT_SHARING_DISABLE);
+
+ /* WaInPlaceDecompressionHang:kbl */
+ WA_SET_BIT(GEN9_GAMT_ECO_REG_RW_IA,
+ GAMT_ECO_ENABLE_IN_PLACE_DECOMPRESS);
+
+ /* WaDisableLSQCROPERFforOCL:kbl */
+ ret = wa_ring_whitelist_reg(engine, GEN8_L3SQCREG4);
+ if (ret)
+ return ret;
+
+ return 0;
+}
+
+static int glk_init_workarounds(struct intel_engine_cs *engine)
+{
+ struct drm_i915_private *dev_priv = engine->i915;
+ int ret;
+
+ ret = gen9_init_workarounds(engine);
+ if (ret)
+ return ret;
+
+ /* WaToEnableHwFixForPushConstHWBug:glk */
+ WA_SET_BIT_MASKED(COMMON_SLICE_CHICKEN2,
+ GEN8_SBE_DISABLE_REPLAY_BUF_OPTIMIZATION);
+
+ return 0;
+}
+
+int init_workarounds_ring(struct intel_engine_cs *engine)
+{
+ struct drm_i915_private *dev_priv = engine->i915;
+
+ WARN_ON(engine->id != RCS);
+
+ dev_priv->workarounds.count = 0;
+ dev_priv->workarounds.hw_whitelist_count[RCS] = 0;
+
+ if (IS_BROADWELL(dev_priv))
+ return bdw_init_workarounds(engine);
+
+ if (IS_CHERRYVIEW(dev_priv))
+ return chv_init_workarounds(engine);
+
+ if (IS_SKYLAKE(dev_priv))
+ return skl_init_workarounds(engine);
+
+ if (IS_BROXTON(dev_priv))
+ return bxt_init_workarounds(engine);
+
+ if (IS_KABYLAKE(dev_priv))
+ return kbl_init_workarounds(engine);
+
+ if (IS_GEMINILAKE(dev_priv))
+ return glk_init_workarounds(engine);
+
+ return 0;
+}
+
+int intel_ring_workarounds_emit(struct drm_i915_gem_request *req)
+{
+ struct i915_workarounds *w = &req->i915->workarounds;
+ u32 *cs;
+ int ret, i;
+
+ if (w->count == 0)
+ return 0;
+
+ ret = req->engine->emit_flush(req, EMIT_BARRIER);
+ if (ret)
+ return ret;
+
+ cs = intel_ring_begin(req, (w->count * 2 + 2));
+ if (IS_ERR(cs))
+ return PTR_ERR(cs);
+
+ *cs++ = MI_LOAD_REGISTER_IMM(w->count);
+ for (i = 0; i < w->count; i++) {
+ *cs++ = i915_mmio_reg_offset(w->reg[i].addr);
+ *cs++ = w->reg[i].value;
+ }
+ *cs++ = MI_NOOP;
+
+ intel_ring_advance(req, cs);
+
+ ret = req->engine->emit_flush(req, EMIT_BARRIER);
+ if (ret)
+ return ret;
+
+ DRM_DEBUG_DRIVER("Number of Workarounds emitted: %d\n", w->count);
+
+ return 0;
+}
+
#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
#include "selftests/mock_engine.c"
#endif
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c
index 8c17db72489f..629fe6584e61 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.c
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.c
@@ -644,41 +644,6 @@ static void reset_ring_common(struct intel_engine_cs *engine,
}
}
-int intel_ring_workarounds_emit(struct drm_i915_gem_request *req)
-{
- struct i915_workarounds *w = &req->i915->workarounds;
- u32 *cs;
- int ret, i;
-
- if (w->count == 0)
- return 0;
-
- ret = req->engine->emit_flush(req, EMIT_BARRIER);
- if (ret)
- return ret;
-
- cs = intel_ring_begin(req, (w->count * 2 + 2));
- if (IS_ERR(cs))
- return PTR_ERR(cs);
-
- *cs++ = MI_LOAD_REGISTER_IMM(w->count);
- for (i = 0; i < w->count; i++) {
- *cs++ = i915_mmio_reg_offset(w->reg[i].addr);
- *cs++ = w->reg[i].value;
- }
- *cs++ = MI_NOOP;
-
- intel_ring_advance(req, cs);
-
- ret = req->engine->emit_flush(req, EMIT_BARRIER);
- if (ret)
- return ret;
-
- DRM_DEBUG_DRIVER("Number of Workarounds emitted: %d\n", w->count);
-
- return 0;
-}
-
static int intel_rcs_ctx_init(struct drm_i915_gem_request *req)
{
int ret;
@@ -694,521 +659,6 @@ static int intel_rcs_ctx_init(struct drm_i915_gem_request *req)
return 0;
}
-static int wa_add(struct drm_i915_private *dev_priv,
- i915_reg_t addr,
- const u32 mask, const u32 val)
-{
- const u32 idx = dev_priv->workarounds.count;
-
- if (WARN_ON(idx >= I915_MAX_WA_REGS))
- return -ENOSPC;
-
- dev_priv->workarounds.reg[idx].addr = addr;
- dev_priv->workarounds.reg[idx].value = val;
- dev_priv->workarounds.reg[idx].mask = mask;
-
- dev_priv->workarounds.count++;
-
- return 0;
-}
-
-#define WA_REG(addr, mask, val) do { \
- const int r = wa_add(dev_priv, (addr), (mask), (val)); \
- if (r) \
- return r; \
- } while (0)
-
-#define WA_SET_BIT_MASKED(addr, mask) \
- WA_REG(addr, (mask), _MASKED_BIT_ENABLE(mask))
-
-#define WA_CLR_BIT_MASKED(addr, mask) \
- WA_REG(addr, (mask), _MASKED_BIT_DISABLE(mask))
-
-#define WA_SET_FIELD_MASKED(addr, mask, value) \
- WA_REG(addr, mask, _MASKED_FIELD(mask, value))
-
-#define WA_SET_BIT(addr, mask) WA_REG(addr, mask, I915_READ(addr) | (mask))
-#define WA_CLR_BIT(addr, mask) WA_REG(addr, mask, I915_READ(addr) & ~(mask))
-
-#define WA_WRITE(addr, val) WA_REG(addr, 0xffffffff, val)
-
-static int wa_ring_whitelist_reg(struct intel_engine_cs *engine,
- i915_reg_t reg)
-{
- struct drm_i915_private *dev_priv = engine->i915;
- struct i915_workarounds *wa = &dev_priv->workarounds;
- const uint32_t index = wa->hw_whitelist_count[engine->id];
-
- if (WARN_ON(index >= RING_MAX_NONPRIV_SLOTS))
- return -EINVAL;
-
- WA_WRITE(RING_FORCE_TO_NONPRIV(engine->mmio_base, index),
- i915_mmio_reg_offset(reg));
- wa->hw_whitelist_count[engine->id]++;
-
- return 0;
-}
-
-static int gen8_init_workarounds(struct intel_engine_cs *engine)
-{
- struct drm_i915_private *dev_priv = engine->i915;
-
- WA_SET_BIT_MASKED(INSTPM, INSTPM_FORCE_ORDERING);
-
- /* WaDisableAsyncFlipPerfMode:bdw,chv */
- WA_SET_BIT_MASKED(MI_MODE, ASYNC_FLIP_PERF_DISABLE);
-
- /* WaDisablePartialInstShootdown:bdw,chv */
- WA_SET_BIT_MASKED(GEN8_ROW_CHICKEN,
- PARTIAL_INSTRUCTION_SHOOTDOWN_DISABLE);
-
- /* Use Force Non-Coherent whenever executing a 3D context. This is a
- * workaround for for a possible hang in the unlikely event a TLB
- * invalidation occurs during a PSD flush.
- */
- /* WaForceEnableNonCoherent:bdw,chv */
- /* WaHdcDisableFetchWhenMasked:bdw,chv */
- WA_SET_BIT_MASKED(HDC_CHICKEN0,
- HDC_DONOT_FETCH_MEM_WHEN_MASKED |
- HDC_FORCE_NON_COHERENT);
-
- /* From the Haswell PRM, Command Reference: Registers, CACHE_MODE_0:
- * "The Hierarchical Z RAW Stall Optimization allows non-overlapping
- * polygons in the same 8x4 pixel/sample area to be processed without
- * stalling waiting for the earlier ones to write to Hierarchical Z
- * buffer."
- *
- * This optimization is off by default for BDW and CHV; turn it on.
- */
- WA_CLR_BIT_MASKED(CACHE_MODE_0_GEN7, HIZ_RAW_STALL_OPT_DISABLE);
-
- /* Wa4x4STCOptimizationDisable:bdw,chv */
- WA_SET_BIT_MASKED(CACHE_MODE_1, GEN8_4x4_STC_OPTIMIZATION_DISABLE);
-
- /*
- * BSpec recommends 8x4 when MSAA is used,
- * however in practice 16x4 seems fastest.
- *
- * Note that PS/WM thread counts depend on the WIZ hashing
- * disable bit, which we don't touch here, but it's good
- * to keep in mind (see 3DSTATE_PS and 3DSTATE_WM).
- */
- WA_SET_FIELD_MASKED(GEN7_GT_MODE,
- GEN6_WIZ_HASHING_MASK,
- GEN6_WIZ_HASHING_16x4);
-
- return 0;
-}
-
-static int bdw_init_workarounds(struct intel_engine_cs *engine)
-{
- struct drm_i915_private *dev_priv = engine->i915;
- int ret;
-
- ret = gen8_init_workarounds(engine);
- if (ret)
- return ret;
-
- /* WaDisableThreadStallDopClockGating:bdw (pre-production) */
- WA_SET_BIT_MASKED(GEN8_ROW_CHICKEN, STALL_DOP_GATING_DISABLE);
-
- /* WaDisableDopClockGating:bdw
- *
- * Also see the related UCGTCL1 write in broadwell_init_clock_gating()
- * to disable EUTC clock gating.
- */
- WA_SET_BIT_MASKED(GEN7_ROW_CHICKEN2,
- DOP_CLOCK_GATING_DISABLE);
-
- WA_SET_BIT_MASKED(HALF_SLICE_CHICKEN3,
- GEN8_SAMPLER_POWER_BYPASS_DIS);
-
- WA_SET_BIT_MASKED(HDC_CHICKEN0,
- /* WaForceContextSaveRestoreNonCoherent:bdw */
- HDC_FORCE_CONTEXT_SAVE_RESTORE_NON_COHERENT |
- /* WaDisableFenceDestinationToSLM:bdw (pre-prod) */
- (IS_BDW_GT3(dev_priv) ? HDC_FENCE_DEST_SLM_DISABLE : 0));
-
- return 0;
-}
-
-static int chv_init_workarounds(struct intel_engine_cs *engine)
-{
- struct drm_i915_private *dev_priv = engine->i915;
- int ret;
-
- ret = gen8_init_workarounds(engine);
- if (ret)
- return ret;
-
- /* WaDisableThreadStallDopClockGating:chv */
- WA_SET_BIT_MASKED(GEN8_ROW_CHICKEN, STALL_DOP_GATING_DISABLE);
-
- /* Improve HiZ throughput on CHV. */
- WA_SET_BIT_MASKED(HIZ_CHICKEN, CHV_HZ_8X8_MODE_IN_1X);
-
- return 0;
-}
-
-static int gen9_init_workarounds(struct intel_engine_cs *engine)
-{
- struct drm_i915_private *dev_priv = engine->i915;
- int ret;
-
- /* WaConextSwitchWithConcurrentTLBInvalidate:skl,bxt,kbl,glk */
- I915_WRITE(GEN9_CSFE_CHICKEN1_RCS, _MASKED_BIT_ENABLE(GEN9_PREEMPT_GPGPU_SYNC_SWITCH_DISABLE));
-
- /* WaEnableLbsSlaRetryTimerDecrement:skl,bxt,kbl,glk */
- I915_WRITE(BDW_SCRATCH1, I915_READ(BDW_SCRATCH1) |
- GEN9_LBS_SLA_RETRY_TIMER_DECREMENT_ENABLE);
-
- /* WaDisableKillLogic:bxt,skl,kbl */
- I915_WRITE(GAM_ECOCHK, I915_READ(GAM_ECOCHK) |
- ECOCHK_DIS_TLB);
-
- /* WaClearFlowControlGpgpuContextSave:skl,bxt,kbl,glk */
- /* WaDisablePartialInstShootdown:skl,bxt,kbl,glk */
- WA_SET_BIT_MASKED(GEN8_ROW_CHICKEN,
- FLOW_CONTROL_ENABLE |
- PARTIAL_INSTRUCTION_SHOOTDOWN_DISABLE);
-
- /* Syncing dependencies between camera and graphics:skl,bxt,kbl */
- WA_SET_BIT_MASKED(HALF_SLICE_CHICKEN3,
- GEN9_DISABLE_OCL_OOB_SUPPRESS_LOGIC);
-
- /* WaDisableDgMirrorFixInHalfSliceChicken5:bxt */
- if (IS_BXT_REVID(dev_priv, 0, BXT_REVID_A1))
- WA_CLR_BIT_MASKED(GEN9_HALF_SLICE_CHICKEN5,
- GEN9_DG_MIRROR_FIX_ENABLE);
-
- /* WaSetDisablePixMaskCammingAndRhwoInCommonSliceChicken:bxt */
- if (IS_BXT_REVID(dev_priv, 0, BXT_REVID_A1)) {
- WA_SET_BIT_MASKED(GEN7_COMMON_SLICE_CHICKEN1,
- GEN9_RHWO_OPTIMIZATION_DISABLE);
- /*
- * WA also requires GEN9_SLICE_COMMON_ECO_CHICKEN0[14:14] to be set
- * but we do that in per ctx batchbuffer as there is an issue
- * with this register not getting restored on ctx restore
- */
- }
-
- /* WaEnableSamplerGPGPUPreemptionSupport:skl,bxt,kbl */
- WA_SET_BIT_MASKED(GEN9_HALF_SLICE_CHICKEN7,
- GEN9_ENABLE_GPGPU_PREEMPTION);
-
- /* Wa4x4STCOptimizationDisable:skl,bxt,kbl,glk */
- /* WaDisablePartialResolveInVc:skl,bxt,kbl */
- WA_SET_BIT_MASKED(CACHE_MODE_1, (GEN8_4x4_STC_OPTIMIZATION_DISABLE |
- GEN9_PARTIAL_RESOLVE_IN_VC_DISABLE));
-
- /* WaCcsTlbPrefetchDisable:skl,bxt,kbl,glk */
- WA_CLR_BIT_MASKED(GEN9_HALF_SLICE_CHICKEN5,
- GEN9_CCS_TLB_PREFETCH_ENABLE);
-
- /* WaDisableMaskBasedCammingInRCC:bxt */
- if (IS_BXT_REVID(dev_priv, 0, BXT_REVID_A1))
- WA_SET_BIT_MASKED(SLICE_ECO_CHICKEN0,
- PIXEL_MASK_CAMMING_DISABLE);
-
- /* WaForceContextSaveRestoreNonCoherent:skl,bxt,kbl */
- WA_SET_BIT_MASKED(HDC_CHICKEN0,
- HDC_FORCE_CONTEXT_SAVE_RESTORE_NON_COHERENT |
- HDC_FORCE_CSR_NON_COHERENT_OVR_DISABLE);
-
- /* WaForceEnableNonCoherent and WaDisableHDCInvalidation are
- * both tied to WaForceContextSaveRestoreNonCoherent
- * in some hsds for skl. We keep the tie for all gen9. The
- * documentation is a bit hazy and so we want to get common behaviour,
- * even though there is no clear evidence we would need both on kbl/bxt.
- * This area has been source of system hangs so we play it safe
- * and mimic the skl regardless of what bspec says.
- *
- * Use Force Non-Coherent whenever executing a 3D context. This
- * is a workaround for a possible hang in the unlikely event
- * a TLB invalidation occurs during a PSD flush.
- */
-
- /* WaForceEnableNonCoherent:skl,bxt,kbl */
- WA_SET_BIT_MASKED(HDC_CHICKEN0,
- HDC_FORCE_NON_COHERENT);
-
- /* WaDisableHDCInvalidation:skl,bxt,kbl */
- I915_WRITE(GAM_ECOCHK, I915_READ(GAM_ECOCHK) |
- BDW_DISABLE_HDC_INVALIDATION);
-
- /* WaDisableSamplerPowerBypassForSOPingPong:skl,bxt,kbl */
- if (IS_SKYLAKE(dev_priv) ||
- IS_KABYLAKE(dev_priv) ||
- IS_BXT_REVID(dev_priv, 0, BXT_REVID_B0))
- WA_SET_BIT_MASKED(HALF_SLICE_CHICKEN3,
- GEN8_SAMPLER_POWER_BYPASS_DIS);
-
- /* WaDisableSTUnitPowerOptimization:skl,bxt,kbl,glk */
- WA_SET_BIT_MASKED(HALF_SLICE_CHICKEN2, GEN8_ST_PO_DISABLE);
-
- /* WaOCLCoherentLineFlush:skl,bxt,kbl */
- I915_WRITE(GEN8_L3SQCREG4, (I915_READ(GEN8_L3SQCREG4) |
- GEN8_LQSC_FLUSH_COHERENT_LINES));
-
- /* WaVFEStateAfterPipeControlwithMediaStateClear:skl,bxt,glk */
- ret = wa_ring_whitelist_reg(engine, GEN9_CTX_PREEMPT_REG);
- if (ret)
- return ret;
-
- /* WaEnablePreemptionGranularityControlByUMD:skl,bxt,kbl */
- ret= wa_ring_whitelist_reg(engine, GEN8_CS_CHICKEN1);
- if (ret)
- return ret;
-
- /* WaAllowUMDToModifyHDCChicken1:skl,bxt,kbl,glk */
- ret = wa_ring_whitelist_reg(engine, GEN8_HDC_CHICKEN1);
- if (ret)
- return ret;
-
- return 0;
-}
-
-static int skl_tune_iz_hashing(struct intel_engine_cs *engine)
-{
- struct drm_i915_private *dev_priv = engine->i915;
- u8 vals[3] = { 0, 0, 0 };
- unsigned int i;
-
- for (i = 0; i < 3; i++) {
- u8 ss;
-
- /*
- * Only consider slices where one, and only one, subslice has 7
- * EUs
- */
- if (!is_power_of_2(INTEL_INFO(dev_priv)->sseu.subslice_7eu[i]))
- continue;
-
- /*
- * subslice_7eu[i] != 0 (because of the check above) and
- * ss_max == 4 (maximum number of subslices possible per slice)
- *
- * -> 0 <= ss <= 3;
- */
- ss = ffs(INTEL_INFO(dev_priv)->sseu.subslice_7eu[i]) - 1;
- vals[i] = 3 - ss;
- }
-
- if (vals[0] == 0 && vals[1] == 0 && vals[2] == 0)
- return 0;
-
- /* Tune IZ hashing. See intel_device_info_runtime_init() */
- WA_SET_FIELD_MASKED(GEN7_GT_MODE,
- GEN9_IZ_HASHING_MASK(2) |
- GEN9_IZ_HASHING_MASK(1) |
- GEN9_IZ_HASHING_MASK(0),
- GEN9_IZ_HASHING(2, vals[2]) |
- GEN9_IZ_HASHING(1, vals[1]) |
- GEN9_IZ_HASHING(0, vals[0]));
-
- return 0;
-}
-
-static int skl_init_workarounds(struct intel_engine_cs *engine)
-{
- struct drm_i915_private *dev_priv = engine->i915;
- int ret;
-
- ret = gen9_init_workarounds(engine);
- if (ret)
- return ret;
-
- /*
- * Actual WA is to disable percontext preemption granularity control
- * until D0 which is the default case so this is equivalent to
- * !WaDisablePerCtxtPreemptionGranularityControl:skl
- */
- I915_WRITE(GEN7_FF_SLICE_CS_CHICKEN1,
- _MASKED_BIT_ENABLE(GEN9_FFSC_PERCTX_PREEMPT_CTRL));
-
- /* WaEnableGapsTsvCreditFix:skl */
- I915_WRITE(GEN8_GARBCNTL, (I915_READ(GEN8_GARBCNTL) |
- GEN9_GAPS_TSV_CREDIT_DISABLE));
-
- /* WaDisableGafsUnitClkGating:skl */
- WA_SET_BIT(GEN7_UCGCTL4, GEN8_EU_GAUNIT_CLOCK_GATE_DISABLE);
-
- /* WaInPlaceDecompressionHang:skl */
- if (IS_SKL_REVID(dev_priv, SKL_REVID_H0, REVID_FOREVER))
- WA_SET_BIT(GEN9_GAMT_ECO_REG_RW_IA,
- GAMT_ECO_ENABLE_IN_PLACE_DECOMPRESS);
-
- /* WaDisableLSQCROPERFforOCL:skl */
- ret = wa_ring_whitelist_reg(engine, GEN8_L3SQCREG4);
- if (ret)
- return ret;
-
- return skl_tune_iz_hashing(engine);
-}
-
-static int bxt_init_workarounds(struct intel_engine_cs *engine)
-{
- struct drm_i915_private *dev_priv = engine->i915;
- int ret;
-
- ret = gen9_init_workarounds(engine);
- if (ret)
- return ret;
-
- /* WaStoreMultiplePTEenable:bxt */
- /* This is a requirement according to Hardware specification */
- if (IS_BXT_REVID(dev_priv, 0, BXT_REVID_A1))
- I915_WRITE(TILECTL, I915_READ(TILECTL) | TILECTL_TLBPF);
-
- /* WaSetClckGatingDisableMedia:bxt */
- if (IS_BXT_REVID(dev_priv, 0, BXT_REVID_A1)) {
- I915_WRITE(GEN7_MISCCPCTL, (I915_READ(GEN7_MISCCPCTL) &
- ~GEN8_DOP_CLOCK_GATE_MEDIA_ENABLE));
- }
-
- /* WaDisableThreadStallDopClockGating:bxt */
- WA_SET_BIT_MASKED(GEN8_ROW_CHICKEN,
- STALL_DOP_GATING_DISABLE);
-
- /* WaDisablePooledEuLoadBalancingFix:bxt */
- if (IS_BXT_REVID(dev_priv, BXT_REVID_B0, REVID_FOREVER)) {
- WA_SET_BIT_MASKED(FF_SLICE_CS_CHICKEN2,
- GEN9_POOLED_EU_LOAD_BALANCING_FIX_DISABLE);
- }
-
- /* WaDisableSbeCacheDispatchPortSharing:bxt */
- if (IS_BXT_REVID(dev_priv, 0, BXT_REVID_B0)) {
- WA_SET_BIT_MASKED(
- GEN7_HALF_SLICE_CHICKEN1,
- GEN7_SBE_SS_CACHE_DISPATCH_PORT_SHARING_DISABLE);
- }
-
- /* WaDisableObjectLevelPreemptionForTrifanOrPolygon:bxt */
- /* WaDisableObjectLevelPreemptionForInstancedDraw:bxt */
- /* WaDisableObjectLevelPreemtionForInstanceId:bxt */
- /* WaDisableLSQCROPERFforOCL:bxt */
- if (IS_BXT_REVID(dev_priv, 0, BXT_REVID_A1)) {
- ret = wa_ring_whitelist_reg(engine, GEN9_CS_DEBUG_MODE1);
- if (ret)
- return ret;
-
- ret = wa_ring_whitelist_reg(engine, GEN8_L3SQCREG4);
- if (ret)
- return ret;
- }
-
- /* WaProgramL3SqcReg1DefaultForPerf:bxt */
- if (IS_BXT_REVID(dev_priv, BXT_REVID_B0, REVID_FOREVER))
- I915_WRITE(GEN8_L3SQCREG1, L3_GENERAL_PRIO_CREDITS(62) |
- L3_HIGH_PRIO_CREDITS(2));
-
- /* WaToEnableHwFixForPushConstHWBug:bxt */
- if (IS_BXT_REVID(dev_priv, BXT_REVID_C0, REVID_FOREVER))
- WA_SET_BIT_MASKED(COMMON_SLICE_CHICKEN2,
- GEN8_SBE_DISABLE_REPLAY_BUF_OPTIMIZATION);
-
- /* WaInPlaceDecompressionHang:bxt */
- if (IS_BXT_REVID(dev_priv, BXT_REVID_C0, REVID_FOREVER))
- WA_SET_BIT(GEN9_GAMT_ECO_REG_RW_IA,
- GAMT_ECO_ENABLE_IN_PLACE_DECOMPRESS);
-
- return 0;
-}
-
-static int kbl_init_workarounds(struct intel_engine_cs *engine)
-{
- struct drm_i915_private *dev_priv = engine->i915;
- int ret;
-
- ret = gen9_init_workarounds(engine);
- if (ret)
- return ret;
-
- /* WaEnableGapsTsvCreditFix:kbl */
- I915_WRITE(GEN8_GARBCNTL, (I915_READ(GEN8_GARBCNTL) |
- GEN9_GAPS_TSV_CREDIT_DISABLE));
-
- /* WaDisableDynamicCreditSharing:kbl */
- if (IS_KBL_REVID(dev_priv, 0, KBL_REVID_B0))
- WA_SET_BIT(GAMT_CHKN_BIT_REG,
- GAMT_CHKN_DISABLE_DYNAMIC_CREDIT_SHARING);
-
- /* WaDisableFenceDestinationToSLM:kbl (pre-prod) */
- if (IS_KBL_REVID(dev_priv, KBL_REVID_A0, KBL_REVID_A0))
- WA_SET_BIT_MASKED(HDC_CHICKEN0,
- HDC_FENCE_DEST_SLM_DISABLE);
-
- /* WaToEnableHwFixForPushConstHWBug:kbl */
- if (IS_KBL_REVID(dev_priv, KBL_REVID_C0, REVID_FOREVER))
- WA_SET_BIT_MASKED(COMMON_SLICE_CHICKEN2,
- GEN8_SBE_DISABLE_REPLAY_BUF_OPTIMIZATION);
-
- /* WaDisableGafsUnitClkGating:kbl */
- WA_SET_BIT(GEN7_UCGCTL4, GEN8_EU_GAUNIT_CLOCK_GATE_DISABLE);
-
- /* WaDisableSbeCacheDispatchPortSharing:kbl */
- WA_SET_BIT_MASKED(
- GEN7_HALF_SLICE_CHICKEN1,
- GEN7_SBE_SS_CACHE_DISPATCH_PORT_SHARING_DISABLE);
-
- /* WaInPlaceDecompressionHang:kbl */
- WA_SET_BIT(GEN9_GAMT_ECO_REG_RW_IA,
- GAMT_ECO_ENABLE_IN_PLACE_DECOMPRESS);
-
- /* WaDisableLSQCROPERFforOCL:kbl */
- ret = wa_ring_whitelist_reg(engine, GEN8_L3SQCREG4);
- if (ret)
- return ret;
-
- return 0;
-}
-
-static int glk_init_workarounds(struct intel_engine_cs *engine)
-{
- struct drm_i915_private *dev_priv = engine->i915;
- int ret;
-
- ret = gen9_init_workarounds(engine);
- if (ret)
- return ret;
-
- /* WaToEnableHwFixForPushConstHWBug:glk */
- WA_SET_BIT_MASKED(COMMON_SLICE_CHICKEN2,
- GEN8_SBE_DISABLE_REPLAY_BUF_OPTIMIZATION);
-
- return 0;
-}
-
-int init_workarounds_ring(struct intel_engine_cs *engine)
-{
- struct drm_i915_private *dev_priv = engine->i915;
-
- WARN_ON(engine->id != RCS);
-
- dev_priv->workarounds.count = 0;
- dev_priv->workarounds.hw_whitelist_count[RCS] = 0;
-
- if (IS_BROADWELL(dev_priv))
- return bdw_init_workarounds(engine);
-
- if (IS_CHERRYVIEW(dev_priv))
- return chv_init_workarounds(engine);
-
- if (IS_SKYLAKE(dev_priv))
- return skl_init_workarounds(engine);
-
- if (IS_BROXTON(dev_priv))
- return bxt_init_workarounds(engine);
-
- if (IS_KABYLAKE(dev_priv))
- return kbl_init_workarounds(engine);
-
- if (IS_GEMINILAKE(dev_priv))
- return glk_init_workarounds(engine);
-
- return 0;
-}
-
static int init_render_ring(struct intel_engine_cs *engine)
{
struct drm_i915_private *dev_priv = engine->i915;
--
2.9.3
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 21+ messages in thread
* [PATCH 4/5] drm/i915: Tidy workaround batch buffer emission
2017-02-16 12:23 [PATCH 0/5] Assorted engine tidy Tvrtko Ursulin
` (2 preceding siblings ...)
2017-02-16 12:23 ` [PATCH 3/5] drm/i915: Move common workaround code to intel_engine_cs Tvrtko Ursulin
@ 2017-02-16 12:23 ` Tvrtko Ursulin
2017-02-16 13:20 ` Chris Wilson
2017-02-16 12:23 ` [PATCH 5/5] drm/i915: Consolidate gen8_emit_pipe_control Tvrtko Ursulin
` (2 subsequent siblings)
6 siblings, 1 reply; 21+ messages in thread
From: Tvrtko Ursulin @ 2017-02-16 12:23 UTC (permalink / raw)
To: Intel-gfx
From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Use the "*batch++ = " style as in the ring emission for better
readability and also simplify the logic a bit by consolidating
the offset and size calculations and overflow checking. The
latter is a programming error so it is not required to check
for it after each write to the object, but rather do it once the
whole state has been written and fail the driver if something
went wrong.
v2: Rebase.
v3: Keep track of offsets and sizes in bytes for simplicity
and rename function pointer variable to _fn suffix.
(Chris Wilson)
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> (v2)
---
drivers/gpu/drm/i915/intel_lrc.c | 316 ++++++++++++++-------------------------
1 file changed, 113 insertions(+), 203 deletions(-)
diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c
index ee431d39ce06..ec39e3e422ca 100644
--- a/drivers/gpu/drm/i915/intel_lrc.c
+++ b/drivers/gpu/drm/i915/intel_lrc.c
@@ -890,18 +890,6 @@ static int execlists_request_alloc(struct drm_i915_gem_request *request)
return ret;
}
-#define wa_ctx_emit(batch, index, cmd) \
- do { \
- int __index = (index)++; \
- if (WARN_ON(__index >= (PAGE_SIZE / sizeof(uint32_t)))) { \
- return -ENOSPC; \
- } \
- batch[__index] = (cmd); \
- } while (0)
-
-#define wa_ctx_emit_reg(batch, index, reg) \
- wa_ctx_emit((batch), (index), i915_mmio_reg_offset(reg))
-
/*
* In this WA we need to set GEN8_L3SQCREG4[21:21] and reset it after
* PIPE_CONTROL instruction. This is required for the flush to happen correctly
@@ -918,56 +906,31 @@ static int execlists_request_alloc(struct drm_i915_gem_request *request)
* This WA is also required for Gen9 so extracting as a function avoids
* code duplication.
*/
-static inline int gen8_emit_flush_coherentl3_wa(struct intel_engine_cs *engine,
- uint32_t *batch,
- uint32_t index)
-{
- uint32_t l3sqc4_flush = (0x40400000 | GEN8_LQSC_FLUSH_COHERENT_LINES);
-
- wa_ctx_emit(batch, index, (MI_STORE_REGISTER_MEM_GEN8 |
- MI_SRM_LRM_GLOBAL_GTT));
- wa_ctx_emit_reg(batch, index, GEN8_L3SQCREG4);
- wa_ctx_emit(batch, index, i915_ggtt_offset(engine->scratch) + 256);
- wa_ctx_emit(batch, index, 0);
-
- wa_ctx_emit(batch, index, MI_LOAD_REGISTER_IMM(1));
- wa_ctx_emit_reg(batch, index, GEN8_L3SQCREG4);
- wa_ctx_emit(batch, index, l3sqc4_flush);
-
- wa_ctx_emit(batch, index, GFX_OP_PIPE_CONTROL(6));
- wa_ctx_emit(batch, index, (PIPE_CONTROL_CS_STALL |
- PIPE_CONTROL_DC_FLUSH_ENABLE));
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
-
- wa_ctx_emit(batch, index, (MI_LOAD_REGISTER_MEM_GEN8 |
- MI_SRM_LRM_GLOBAL_GTT));
- wa_ctx_emit_reg(batch, index, GEN8_L3SQCREG4);
- wa_ctx_emit(batch, index, i915_ggtt_offset(engine->scratch) + 256);
- wa_ctx_emit(batch, index, 0);
-
- return index;
-}
-
-static inline uint32_t wa_ctx_start(struct i915_wa_ctx_bb *wa_ctx,
- uint32_t offset,
- uint32_t start_alignment)
+static u32 *
+gen8_emit_flush_coherentl3_wa(struct intel_engine_cs *engine, u32 *batch)
{
- return wa_ctx->offset = ALIGN(offset, start_alignment);
-}
-
-static inline int wa_ctx_end(struct i915_wa_ctx_bb *wa_ctx,
- uint32_t offset,
- uint32_t size_alignment)
-{
- wa_ctx->size = offset - wa_ctx->offset;
-
- WARN(wa_ctx->size % size_alignment,
- "wa_ctx_bb failed sanity checks: size %d is not aligned to %d\n",
- wa_ctx->size, size_alignment);
- return 0;
+ *batch++ = MI_STORE_REGISTER_MEM_GEN8 | MI_SRM_LRM_GLOBAL_GTT;
+ *batch++ = i915_mmio_reg_offset(GEN8_L3SQCREG4);
+ *batch++ = i915_ggtt_offset(engine->scratch) + 256;
+ *batch++ = 0;
+
+ *batch++ = MI_LOAD_REGISTER_IMM(1);
+ *batch++ = i915_mmio_reg_offset(GEN8_L3SQCREG4);
+ *batch++ = 0x40400000 | GEN8_LQSC_FLUSH_COHERENT_LINES;
+
+ *batch++ = GFX_OP_PIPE_CONTROL(6);
+ *batch++ = PIPE_CONTROL_CS_STALL | PIPE_CONTROL_DC_FLUSH_ENABLE;
+ *batch++ = 0;
+ *batch++ = 0;
+ *batch++ = 0;
+ *batch++ = 0;
+
+ *batch++ = MI_LOAD_REGISTER_MEM_GEN8 | MI_SRM_LRM_GLOBAL_GTT;
+ *batch++ = i915_mmio_reg_offset(GEN8_L3SQCREG4);
+ *batch++ = i915_ggtt_offset(engine->scratch) + 256;
+ *batch++ = 0;
+
+ return batch;
}
/*
@@ -985,42 +948,28 @@ static inline int wa_ctx_end(struct i915_wa_ctx_bb *wa_ctx,
* MI_BATCH_BUFFER_END will be added to perctx batch and both of them together
* makes a complete batch buffer.
*/
-static int gen8_init_indirectctx_bb(struct intel_engine_cs *engine,
- struct i915_wa_ctx_bb *wa_ctx,
- uint32_t *batch,
- uint32_t *offset)
+static u32 *gen8_init_indirectctx_bb(struct intel_engine_cs *engine, u32 *batch)
{
- uint32_t scratch_addr;
- uint32_t index = wa_ctx_start(wa_ctx, *offset, CACHELINE_DWORDS);
-
/* WaDisableCtxRestoreArbitration:bdw,chv */
- wa_ctx_emit(batch, index, MI_ARB_ON_OFF | MI_ARB_DISABLE);
+ *batch++ = MI_ARB_ON_OFF | MI_ARB_DISABLE;
/* WaFlushCoherentL3CacheLinesAtContextSwitch:bdw */
- if (IS_BROADWELL(engine->i915)) {
- int rc = gen8_emit_flush_coherentl3_wa(engine, batch, index);
- if (rc < 0)
- return rc;
- index = rc;
- }
+ if (IS_BROADWELL(engine->i915))
+ batch = gen8_emit_flush_coherentl3_wa(engine, batch);
+ *batch++ = GFX_OP_PIPE_CONTROL(6);
+ *batch++ = PIPE_CONTROL_FLUSH_L3 | PIPE_CONTROL_GLOBAL_GTT_IVB |
+ PIPE_CONTROL_CS_STALL | PIPE_CONTROL_QW_WRITE;
/* WaClearSlmSpaceAtContextSwitch:bdw,chv */
/* Actual scratch location is at 128 bytes offset */
- scratch_addr = i915_ggtt_offset(engine->scratch) + 2 * CACHELINE_BYTES;
-
- wa_ctx_emit(batch, index, GFX_OP_PIPE_CONTROL(6));
- wa_ctx_emit(batch, index, (PIPE_CONTROL_FLUSH_L3 |
- PIPE_CONTROL_GLOBAL_GTT_IVB |
- PIPE_CONTROL_CS_STALL |
- PIPE_CONTROL_QW_WRITE));
- wa_ctx_emit(batch, index, scratch_addr);
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
+ *batch++ = i915_ggtt_offset(engine->scratch) + 2 * CACHELINE_BYTES;
+ *batch++ = 0;
+ *batch++ = 0;
+ *batch++ = 0;
/* Pad to end of cacheline */
- while (index % CACHELINE_DWORDS)
- wa_ctx_emit(batch, index, MI_NOOP);
+ while ((unsigned long)batch % CACHELINE_BYTES)
+ *batch++ = MI_NOOP;
/*
* MI_BATCH_BUFFER_END is not required in Indirect ctx BB because
@@ -1028,7 +977,7 @@ static int gen8_init_indirectctx_bb(struct intel_engine_cs *engine,
* in the register CTX_RCS_INDIRECT_CTX
*/
- return wa_ctx_end(wa_ctx, *offset = index, CACHELINE_DWORDS);
+ return batch;
}
/*
@@ -1040,58 +989,38 @@ static int gen8_init_indirectctx_bb(struct intel_engine_cs *engine,
* This batch is terminated with MI_BATCH_BUFFER_END and so we need not add padding
* to align it with cacheline as padding after MI_BATCH_BUFFER_END is redundant.
*/
-static int gen8_init_perctx_bb(struct intel_engine_cs *engine,
- struct i915_wa_ctx_bb *wa_ctx,
- uint32_t *batch,
- uint32_t *offset)
+static u32 *gen8_init_perctx_bb(struct intel_engine_cs *engine, u32 *batch)
{
- uint32_t index = wa_ctx_start(wa_ctx, *offset, CACHELINE_DWORDS);
-
/* WaDisableCtxRestoreArbitration:bdw,chv */
- wa_ctx_emit(batch, index, MI_ARB_ON_OFF | MI_ARB_ENABLE);
-
- wa_ctx_emit(batch, index, MI_BATCH_BUFFER_END);
+ *batch++ = MI_ARB_ON_OFF | MI_ARB_ENABLE;
+ *batch++ = MI_BATCH_BUFFER_END;
- return wa_ctx_end(wa_ctx, *offset = index, 1);
+ return batch;
}
-static int gen9_init_indirectctx_bb(struct intel_engine_cs *engine,
- struct i915_wa_ctx_bb *wa_ctx,
- uint32_t *batch,
- uint32_t *offset)
+static u32 *gen9_init_indirectctx_bb(struct intel_engine_cs *engine, u32 *batch)
{
- int ret;
- struct drm_i915_private *dev_priv = engine->i915;
- uint32_t index = wa_ctx_start(wa_ctx, *offset, CACHELINE_DWORDS);
-
/* WaFlushCoherentL3CacheLinesAtContextSwitch:skl,bxt,glk */
- ret = gen8_emit_flush_coherentl3_wa(engine, batch, index);
- if (ret < 0)
- return ret;
- index = ret;
+ batch = gen8_emit_flush_coherentl3_wa(engine, batch);
/* WaDisableGatherAtSetShaderCommonSlice:skl,bxt,kbl,glk */
- wa_ctx_emit(batch, index, MI_LOAD_REGISTER_IMM(1));
- wa_ctx_emit_reg(batch, index, COMMON_SLICE_CHICKEN2);
- wa_ctx_emit(batch, index, _MASKED_BIT_DISABLE(
- GEN9_DISABLE_GATHER_AT_SET_SHADER_COMMON_SLICE));
- wa_ctx_emit(batch, index, MI_NOOP);
+ *batch++ = MI_LOAD_REGISTER_IMM(1);
+ *batch++ = i915_mmio_reg_offset(COMMON_SLICE_CHICKEN2);
+ *batch++ = _MASKED_BIT_DISABLE(
+ GEN9_DISABLE_GATHER_AT_SET_SHADER_COMMON_SLICE);
+ *batch++ = MI_NOOP;
/* WaClearSlmSpaceAtContextSwitch:kbl */
/* Actual scratch location is at 128 bytes offset */
- if (IS_KBL_REVID(dev_priv, 0, KBL_REVID_A0)) {
- u32 scratch_addr =
- i915_ggtt_offset(engine->scratch) + 2 * CACHELINE_BYTES;
-
- wa_ctx_emit(batch, index, GFX_OP_PIPE_CONTROL(6));
- wa_ctx_emit(batch, index, (PIPE_CONTROL_FLUSH_L3 |
- PIPE_CONTROL_GLOBAL_GTT_IVB |
- PIPE_CONTROL_CS_STALL |
- PIPE_CONTROL_QW_WRITE));
- wa_ctx_emit(batch, index, scratch_addr);
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
+ if (IS_KBL_REVID(engine->i915, 0, KBL_REVID_A0)) {
+ *batch++ = GFX_OP_PIPE_CONTROL(6);
+ *batch++ = PIPE_CONTROL_FLUSH_L3 | PIPE_CONTROL_GLOBAL_GTT_IVB |
+ PIPE_CONTROL_CS_STALL | PIPE_CONTROL_QW_WRITE;
+ *batch++ = i915_ggtt_offset(engine->scratch) +
+ 2 * CACHELINE_BYTES;
+ *batch++ = 0;
+ *batch++ = 0;
+ *batch++ = 0;
}
/* WaMediaPoolStateCmdInWABB:bxt,glk */
@@ -1109,41 +1038,37 @@ static int gen9_init_indirectctx_bb(struct intel_engine_cs *engine,
* possible configurations, to avoid duplication they are
* not shown here again.
*/
- u32 eu_pool_config = 0x00777000;
- wa_ctx_emit(batch, index, GEN9_MEDIA_POOL_STATE);
- wa_ctx_emit(batch, index, GEN9_MEDIA_POOL_ENABLE);
- wa_ctx_emit(batch, index, eu_pool_config);
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
+ *batch++ = GEN9_MEDIA_POOL_STATE;
+ *batch++ = GEN9_MEDIA_POOL_ENABLE;
+ *batch++ = 0x00777000;
+ *batch++ = 0;
+ *batch++ = 0;
+ *batch++ = 0;
}
/* Pad to end of cacheline */
- while (index % CACHELINE_DWORDS)
- wa_ctx_emit(batch, index, MI_NOOP);
+ while ((unsigned long)batch % CACHELINE_BYTES)
+ *batch++ = MI_NOOP;
- return wa_ctx_end(wa_ctx, *offset = index, CACHELINE_DWORDS);
+ return batch;
}
-static int gen9_init_perctx_bb(struct intel_engine_cs *engine,
- struct i915_wa_ctx_bb *wa_ctx,
- uint32_t *batch,
- uint32_t *offset)
+static u32 *gen9_init_perctx_bb(struct intel_engine_cs *engine, u32 *batch)
{
- uint32_t index = wa_ctx_start(wa_ctx, *offset, CACHELINE_DWORDS);
-
- wa_ctx_emit(batch, index, MI_BATCH_BUFFER_END);
+ *batch++ = MI_BATCH_BUFFER_END;
- return wa_ctx_end(wa_ctx, *offset = index, 1);
+ return batch;
}
-static int lrc_setup_wa_ctx_obj(struct intel_engine_cs *engine, u32 size)
+#define CTX_WA_BB_OBJ_SIZE (PAGE_SIZE)
+
+static int lrc_setup_wa_ctx(struct intel_engine_cs *engine)
{
struct drm_i915_gem_object *obj;
struct i915_vma *vma;
int err;
- obj = i915_gem_object_create(engine->i915, PAGE_ALIGN(size));
+ obj = i915_gem_object_create(engine->i915, CTX_WA_BB_OBJ_SIZE);
if (IS_ERR(obj))
return PTR_ERR(obj);
@@ -1165,80 +1090,66 @@ static int lrc_setup_wa_ctx_obj(struct intel_engine_cs *engine, u32 size)
return err;
}
-static void lrc_destroy_wa_ctx_obj(struct intel_engine_cs *engine)
+static void lrc_destroy_wa_ctx(struct intel_engine_cs *engine)
{
i915_vma_unpin_and_release(&engine->wa_ctx.vma);
}
+typedef u32 *(*wa_bb_func_t)(struct intel_engine_cs *engine, u32 *batch);
+
static int intel_init_workaround_bb(struct intel_engine_cs *engine)
{
struct i915_ctx_workarounds *wa_ctx = &engine->wa_ctx;
- uint32_t *batch;
- uint32_t offset;
+ struct i915_wa_ctx_bb *wa_bb[2] = { &wa_ctx->indirect_ctx,
+ &wa_ctx->per_ctx };
+ wa_bb_func_t wa_bb_fn[2];
struct page *page;
+ void *batch, *batch_ptr;
+ unsigned int i;
int ret;
- WARN_ON(engine->id != RCS);
+ if (WARN_ON(engine->id != RCS || !engine->scratch))
+ return -EINVAL;
- /* update this when WA for higher Gen are added */
- if (INTEL_GEN(engine->i915) > 9) {
- DRM_ERROR("WA batch buffer is not initialized for Gen%d\n",
- INTEL_GEN(engine->i915));
+ switch (INTEL_GEN(engine->i915)) {
+ case 9:
+ wa_bb_fn[0] = gen9_init_indirectctx_bb;
+ wa_bb_fn[1] = gen9_init_perctx_bb;
+ break;
+ case 8:
+ wa_bb_fn[0] = gen8_init_indirectctx_bb;
+ wa_bb_fn[1] = gen8_init_perctx_bb;
+ break;
+ default:
+ MISSING_CASE(INTEL_GEN(engine->i915));
return 0;
}
- /* some WA perform writes to scratch page, ensure it is valid */
- if (!engine->scratch) {
- DRM_ERROR("scratch page not allocated for %s\n", engine->name);
- return -EINVAL;
- }
-
- ret = lrc_setup_wa_ctx_obj(engine, PAGE_SIZE);
+ ret = lrc_setup_wa_ctx(engine);
if (ret) {
DRM_DEBUG_DRIVER("Failed to setup context WA page: %d\n", ret);
return ret;
}
page = i915_gem_object_get_dirty_page(wa_ctx->vma->obj, 0);
- batch = kmap_atomic(page);
- offset = 0;
-
- if (IS_GEN8(engine->i915)) {
- ret = gen8_init_indirectctx_bb(engine,
- &wa_ctx->indirect_ctx,
- batch,
- &offset);
- if (ret)
- goto out;
+ batch = batch_ptr = kmap_atomic(page);
- ret = gen8_init_perctx_bb(engine,
- &wa_ctx->per_ctx,
- batch,
- &offset);
- if (ret)
- goto out;
- } else if (IS_GEN9(engine->i915)) {
- ret = gen9_init_indirectctx_bb(engine,
- &wa_ctx->indirect_ctx,
- batch,
- &offset);
- if (ret)
- goto out;
-
- ret = gen9_init_perctx_bb(engine,
- &wa_ctx->per_ctx,
- batch,
- &offset);
- if (ret)
- goto out;
+ /*
+ * Emit the two workaround batch buffers, recording the offset from the
+ * start of the workaround batch buffer object for each and their
+ * respective sizes.
+ */
+ for (i = 0; i < ARRAY_SIZE(wa_bb_fn); i++) {
+ wa_bb[i]->offset = ALIGN(batch_ptr - batch, CACHELINE_BYTES);
+ batch_ptr = wa_bb_fn[i](engine, batch_ptr);
+ wa_bb[i]->size = batch_ptr - batch + wa_bb[i]->offset;
}
-out:
+ BUG_ON(batch_ptr - batch > CTX_WA_BB_OBJ_SIZE);
+
kunmap_atomic(batch);
- if (ret)
- lrc_destroy_wa_ctx_obj(engine);
- return ret;
+ return 0;
}
static u32 port_seqno(struct execlist_port *port)
@@ -1685,7 +1596,7 @@ void intel_logical_ring_cleanup(struct intel_engine_cs *engine)
intel_engine_cleanup_common(engine);
- lrc_destroy_wa_ctx_obj(engine);
+ lrc_destroy_wa_ctx(engine);
engine->i915 = NULL;
dev_priv->engine[engine->id] = NULL;
kfree(engine);
@@ -1971,15 +1882,14 @@ static void execlists_init_reg_state(u32 *reg_state,
u32 ggtt_offset = i915_ggtt_offset(wa_ctx->vma);
reg_state[CTX_RCS_INDIRECT_CTX+1] =
- (ggtt_offset + wa_ctx->indirect_ctx.offset * sizeof(uint32_t)) |
- (wa_ctx->indirect_ctx.size / CACHELINE_DWORDS);
+ (ggtt_offset + wa_ctx->indirect_ctx.offset) |
+ (wa_ctx->indirect_ctx.size / CACHELINE_BYTES);
reg_state[CTX_RCS_INDIRECT_CTX_OFFSET+1] =
intel_lr_indirect_ctx_offset(engine) << 6;
reg_state[CTX_BB_PER_CTX_PTR+1] =
- (ggtt_offset + wa_ctx->per_ctx.offset * sizeof(uint32_t)) |
- 0x01;
+ (ggtt_offset + wa_ctx->per_ctx.offset) | 0x01;
}
}
reg_state[CTX_LRI_HEADER_1] = MI_LOAD_REGISTER_IMM(9) | MI_LRI_FORCE_POSTED;
--
2.9.3
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 21+ messages in thread
* [PATCH 5/5] drm/i915: Consolidate gen8_emit_pipe_control
2017-02-16 12:23 [PATCH 0/5] Assorted engine tidy Tvrtko Ursulin
` (3 preceding siblings ...)
2017-02-16 12:23 ` [PATCH 4/5] drm/i915: Tidy workaround batch buffer emission Tvrtko Ursulin
@ 2017-02-16 12:23 ` Tvrtko Ursulin
2017-02-16 18:22 ` ✓ Fi.CI.BAT: success for Assorted engine tidy (rev3) Patchwork
2017-02-17 8:22 ` ✗ Fi.CI.BAT: failure for Assorted engine tidy (rev4) Patchwork
6 siblings, 0 replies; 21+ messages in thread
From: Tvrtko Ursulin @ 2017-02-16 12:23 UTC (permalink / raw)
To: Intel-gfx
From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
We have a few open coded instances in the execlists code and an
almost suitable helper in intel_ringbuf.c
We can consolidate to a single helper if we change the existing
helper to emit directly to ring buffer memory and move the space
reservation outside it.
v2: Drop memcpy for memset. (Chris Wilson)
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/intel_lrc.c | 77 +++++++++++----------------------
drivers/gpu/drm/i915/intel_ringbuffer.c | 45 +++++++------------
drivers/gpu/drm/i915/intel_ringbuffer.h | 11 +++++
3 files changed, 52 insertions(+), 81 deletions(-)
diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c
index ec39e3e422ca..231b011ee1f3 100644
--- a/drivers/gpu/drm/i915/intel_lrc.c
+++ b/drivers/gpu/drm/i915/intel_lrc.c
@@ -918,12 +918,10 @@ gen8_emit_flush_coherentl3_wa(struct intel_engine_cs *engine, u32 *batch)
*batch++ = i915_mmio_reg_offset(GEN8_L3SQCREG4);
*batch++ = 0x40400000 | GEN8_LQSC_FLUSH_COHERENT_LINES;
- *batch++ = GFX_OP_PIPE_CONTROL(6);
- *batch++ = PIPE_CONTROL_CS_STALL | PIPE_CONTROL_DC_FLUSH_ENABLE;
- *batch++ = 0;
- *batch++ = 0;
- *batch++ = 0;
- *batch++ = 0;
+ batch = gen8_emit_pipe_control(batch,
+ PIPE_CONTROL_CS_STALL |
+ PIPE_CONTROL_DC_FLUSH_ENABLE,
+ 0);
*batch++ = MI_LOAD_REGISTER_MEM_GEN8 | MI_SRM_LRM_GLOBAL_GTT;
*batch++ = i915_mmio_reg_offset(GEN8_L3SQCREG4);
@@ -957,15 +955,15 @@ static u32 *gen8_init_indirectctx_bb(struct intel_engine_cs *engine, u32 *batch)
if (IS_BROADWELL(engine->i915))
batch = gen8_emit_flush_coherentl3_wa(engine, batch);
- *batch++ = GFX_OP_PIPE_CONTROL(6);
- *batch++ = PIPE_CONTROL_FLUSH_L3 | PIPE_CONTROL_GLOBAL_GTT_IVB |
- PIPE_CONTROL_CS_STALL | PIPE_CONTROL_QW_WRITE;
/* WaClearSlmSpaceAtContextSwitch:bdw,chv */
/* Actual scratch location is at 128 bytes offset */
- *batch++ = i915_ggtt_offset(engine->scratch) + 2 * CACHELINE_BYTES;
- *batch++ = 0;
- *batch++ = 0;
- *batch++ = 0;
+ batch = gen8_emit_pipe_control(batch,
+ PIPE_CONTROL_FLUSH_L3 |
+ PIPE_CONTROL_GLOBAL_GTT_IVB |
+ PIPE_CONTROL_CS_STALL |
+ PIPE_CONTROL_QW_WRITE,
+ i915_ggtt_offset(engine->scratch) +
+ 2 * CACHELINE_BYTES);
/* Pad to end of cacheline */
while ((unsigned long)batch % CACHELINE_BYTES)
@@ -1013,14 +1011,13 @@ static u32 *gen9_init_indirectctx_bb(struct intel_engine_cs *engine, u32 *batch)
/* WaClearSlmSpaceAtContextSwitch:kbl */
/* Actual scratch location is at 128 bytes offset */
if (IS_KBL_REVID(engine->i915, 0, KBL_REVID_A0)) {
- *batch++ = GFX_OP_PIPE_CONTROL(6);
- *batch++ = PIPE_CONTROL_FLUSH_L3 | PIPE_CONTROL_GLOBAL_GTT_IVB |
- PIPE_CONTROL_CS_STALL | PIPE_CONTROL_QW_WRITE;
- *batch++ = i915_ggtt_offset(engine->scratch) +
- 2 * CACHELINE_BYTES;
- *batch++ = 0;
- *batch++ = 0;
- *batch++ = 0;
+ batch = gen8_emit_pipe_control(batch,
+ PIPE_CONTROL_FLUSH_L3 |
+ PIPE_CONTROL_GLOBAL_GTT_IVB |
+ PIPE_CONTROL_CS_STALL |
+ PIPE_CONTROL_QW_WRITE,
+ i915_ggtt_offset(engine->scratch)
+ + 2 * CACHELINE_BYTES);
}
/* WaMediaPoolStateCmdInWABB:bxt,glk */
@@ -1450,39 +1447,17 @@ static int gen8_emit_flush_render(struct drm_i915_gem_request *request,
if (IS_ERR(cs))
return PTR_ERR(cs);
- if (vf_flush_wa) {
- *cs++ = GFX_OP_PIPE_CONTROL(6);
- *cs++ = 0;
- *cs++ = 0;
- *cs++ = 0;
- *cs++ = 0;
- *cs++ = 0;
- }
+ if (vf_flush_wa)
+ cs = gen8_emit_pipe_control(cs, 0, 0);
- if (dc_flush_wa) {
- *cs++ = GFX_OP_PIPE_CONTROL(6);
- *cs++ = PIPE_CONTROL_DC_FLUSH_ENABLE;
- *cs++ = 0;
- *cs++ = 0;
- *cs++ = 0;
- *cs++ = 0;
- }
+ if (dc_flush_wa)
+ cs = gen8_emit_pipe_control(cs, PIPE_CONTROL_DC_FLUSH_ENABLE,
+ 0);
- *cs++ = GFX_OP_PIPE_CONTROL(6);
- *cs++ = flags;
- *cs++ = scratch_addr;
- *cs++ = 0;
- *cs++ = 0;
- *cs++ = 0;
+ cs = gen8_emit_pipe_control(cs, flags, scratch_addr);
- if (dc_flush_wa) {
- *cs++ = GFX_OP_PIPE_CONTROL(6);
- *cs++ = PIPE_CONTROL_CS_STALL;
- *cs++ = 0;
- *cs++ = 0;
- *cs++ = 0;
- *cs++ = 0;
- }
+ if (dc_flush_wa)
+ cs = gen8_emit_pipe_control(cs, PIPE_CONTROL_CS_STALL, 0);
intel_ring_advance(request, cs);
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c
index 629fe6584e61..d56f384938f7 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.c
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.c
@@ -334,35 +334,16 @@ gen7_render_ring_flush(struct drm_i915_gem_request *req, u32 mode)
}
static int
-gen8_emit_pipe_control(struct drm_i915_gem_request *req,
- u32 flags, u32 scratch_addr)
+gen8_render_ring_flush(struct drm_i915_gem_request *req, u32 mode)
{
+ u32 flags;
u32 *cs;
- cs = intel_ring_begin(req, 6);
+ cs = intel_ring_begin(req, mode & EMIT_INVALIDATE ? 12 : 6);
if (IS_ERR(cs))
return PTR_ERR(cs);
- *cs++ = GFX_OP_PIPE_CONTROL(6);
- *cs++ = flags;
- *cs++ = scratch_addr;
- *cs++ = 0;
- *cs++ = 0;
- *cs++ = 0;
- intel_ring_advance(req, cs);
-
- return 0;
-}
-
-static int
-gen8_render_ring_flush(struct drm_i915_gem_request *req, u32 mode)
-{
- u32 scratch_addr =
- i915_ggtt_offset(req->engine->scratch) + 2 * CACHELINE_BYTES;
- u32 flags = 0;
- int ret;
-
- flags |= PIPE_CONTROL_CS_STALL;
+ flags = PIPE_CONTROL_CS_STALL;
if (mode & EMIT_FLUSH) {
flags |= PIPE_CONTROL_RENDER_TARGET_CACHE_FLUSH;
@@ -381,15 +362,19 @@ gen8_render_ring_flush(struct drm_i915_gem_request *req, u32 mode)
flags |= PIPE_CONTROL_GLOBAL_GTT_IVB;
/* WaCsStallBeforeStateCacheInvalidate:bdw,chv */
- ret = gen8_emit_pipe_control(req,
- PIPE_CONTROL_CS_STALL |
- PIPE_CONTROL_STALL_AT_SCOREBOARD,
- 0);
- if (ret)
- return ret;
+ cs = gen8_emit_pipe_control(cs,
+ PIPE_CONTROL_CS_STALL |
+ PIPE_CONTROL_STALL_AT_SCOREBOARD,
+ 0);
}
- return gen8_emit_pipe_control(req, flags, scratch_addr);
+ cs = gen8_emit_pipe_control(cs, flags,
+ i915_ggtt_offset(req->engine->scratch) +
+ 2 * CACHELINE_BYTES);
+
+ intel_ring_advance(req, cs);
+
+ return 0;
}
static void ring_setup_phys_status_page(struct intel_engine_cs *engine)
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h
index 1d9fd4e42622..4ca5308e12bb 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.h
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.h
@@ -631,4 +631,15 @@ void intel_engine_reset_breadcrumbs(struct intel_engine_cs *engine);
void intel_engine_fini_breadcrumbs(struct intel_engine_cs *engine);
unsigned int intel_breadcrumbs_busy(struct drm_i915_private *i915);
+static inline u32 *gen8_emit_pipe_control(u32 *batch, u32 flags, u32 offset)
+{
+ memset(batch, 0, 6 * sizeof(u32));
+
+ batch[0] = GFX_OP_PIPE_CONTROL(6);
+ batch[1] = flags;
+ batch[2] = offset;
+
+ return batch + 6;
+}
+
#endif /* _INTEL_RINGBUFFER_H_ */
--
2.9.3
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 21+ messages in thread
* Re: [PATCH 1/5] drm/i915: Make int __intel_ring_space static
2017-02-16 12:23 ` [PATCH 1/5] drm/i915: Make int __intel_ring_space static Tvrtko Ursulin
@ 2017-02-16 13:12 ` Chris Wilson
0 siblings, 0 replies; 21+ messages in thread
From: Chris Wilson @ 2017-02-16 13:12 UTC (permalink / raw)
To: Tvrtko Ursulin; +Cc: Intel-gfx
On Thu, Feb 16, 2017 at 12:23:21PM +0000, Tvrtko Ursulin wrote:
> From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
>
> It is only used within intel_ringbuffer.c
>
> Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.oc.uk>
-Chris
--
Chris Wilson, Intel Open Source Technology Centre
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 21+ messages in thread
* Re: [PATCH 2/5] drm/i915: Simplify cleanup path in intel_engines_init
2017-02-16 12:23 ` [PATCH 2/5] drm/i915: Simplify cleanup path in intel_engines_init Tvrtko Ursulin
@ 2017-02-16 13:15 ` Chris Wilson
0 siblings, 0 replies; 21+ messages in thread
From: Chris Wilson @ 2017-02-16 13:15 UTC (permalink / raw)
To: Tvrtko Ursulin; +Cc: Intel-gfx
On Thu, Feb 16, 2017 at 12:23:22PM +0000, Tvrtko Ursulin wrote:
> From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
>
> We can call the engine cleanup vfunc instead of duplicating the
> decision making here.
>
> Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Ok, but dev_priv->gt.cleanup_engine() I still want to see removed at
some point. The paths are nearly identical that most can be done in
common, with the existing per-hw engine->cleanup() to cater for the
rest. Or everything is just done from engine->cleanup() as the entry
point.
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
-Chris
--
Chris Wilson, Intel Open Source Technology Centre
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 21+ messages in thread
* Re: [PATCH 3/5] drm/i915: Move common workaround code to intel_engine_cs
2017-02-16 12:23 ` [PATCH 3/5] drm/i915: Move common workaround code to intel_engine_cs Tvrtko Ursulin
@ 2017-02-16 13:18 ` Chris Wilson
0 siblings, 0 replies; 21+ messages in thread
From: Chris Wilson @ 2017-02-16 13:18 UTC (permalink / raw)
To: Tvrtko Ursulin; +Cc: Intel-gfx
On Thu, Feb 16, 2017 at 12:23:23PM +0000, Tvrtko Ursulin wrote:
> From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
>
> It is used by all submission backends.
>
> Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
> +int init_workarounds_ring(struct intel_engine_cs *engine)
Tomorrow: s/init_workarounds_ring/intel_engine_init_wa/ !
-Chris
--
Chris Wilson, Intel Open Source Technology Centre
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 21+ messages in thread
* Re: [PATCH 4/5] drm/i915: Tidy workaround batch buffer emission
2017-02-16 12:23 ` [PATCH 4/5] drm/i915: Tidy workaround batch buffer emission Tvrtko Ursulin
@ 2017-02-16 13:20 ` Chris Wilson
2017-02-16 13:23 ` Tvrtko Ursulin
0 siblings, 1 reply; 21+ messages in thread
From: Chris Wilson @ 2017-02-16 13:20 UTC (permalink / raw)
To: Tvrtko Ursulin; +Cc: Intel-gfx
On Thu, Feb 16, 2017 at 12:23:24PM +0000, Tvrtko Ursulin wrote:
> + /*
> + * Emit the two workaround batch buffers, recording the offset from the
> + * start of the workaround batch buffer object for each and their
> + * respective sizes.
> + */
> + for (i = 0; i < ARRAY_SIZE(wa_bb_fn); i++) {
> + wa_bb[i]->offset = ALIGN(batch_ptr - batch, CACHELINE_BYTES);
> + batch_ptr = wa_bb_fn[i](engine, batch_ptr);
> + wa_bb[i]->size = batch_ptr - batch + wa_bb[i]->offset;
wa_bb[i]->size = batch_ptr - (batch + wa_bb[i]->offset);
?
--
Chris Wilson, Intel Open Source Technology Centre
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 21+ messages in thread
* Re: [PATCH 4/5] drm/i915: Tidy workaround batch buffer emission
2017-02-16 13:20 ` Chris Wilson
@ 2017-02-16 13:23 ` Tvrtko Ursulin
2017-02-16 13:28 ` Chris Wilson
0 siblings, 1 reply; 21+ messages in thread
From: Tvrtko Ursulin @ 2017-02-16 13:23 UTC (permalink / raw)
To: Chris Wilson, Tvrtko Ursulin, Intel-gfx
On 16/02/2017 13:20, Chris Wilson wrote:
> On Thu, Feb 16, 2017 at 12:23:24PM +0000, Tvrtko Ursulin wrote:
>> + /*
>> + * Emit the two workaround batch buffers, recording the offset from the
>> + * start of the workaround batch buffer object for each and their
>> + * respective sizes.
>> + */
>> + for (i = 0; i < ARRAY_SIZE(wa_bb_fn); i++) {
>> + wa_bb[i]->offset = ALIGN(batch_ptr - batch, CACHELINE_BYTES);
>> + batch_ptr = wa_bb_fn[i](engine, batch_ptr);
>> + wa_bb[i]->size = batch_ptr - batch + wa_bb[i]->offset;
>
> wa_bb[i]->size = batch_ptr - (batch + wa_bb[i]->offset);
> ?
Shock horror.. Thanks.
Regards,
Tvrtko
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 21+ messages in thread
* Re: [PATCH 4/5] drm/i915: Tidy workaround batch buffer emission
2017-02-16 13:23 ` Tvrtko Ursulin
@ 2017-02-16 13:28 ` Chris Wilson
2017-02-16 13:47 ` Tvrtko Ursulin
0 siblings, 1 reply; 21+ messages in thread
From: Chris Wilson @ 2017-02-16 13:28 UTC (permalink / raw)
To: Tvrtko Ursulin; +Cc: Intel-gfx
On Thu, Feb 16, 2017 at 01:23:51PM +0000, Tvrtko Ursulin wrote:
>
> On 16/02/2017 13:20, Chris Wilson wrote:
> >On Thu, Feb 16, 2017 at 12:23:24PM +0000, Tvrtko Ursulin wrote:
> >>+ /*
> >>+ * Emit the two workaround batch buffers, recording the offset from the
> >>+ * start of the workaround batch buffer object for each and their
> >>+ * respective sizes.
> >>+ */
> >>+ for (i = 0; i < ARRAY_SIZE(wa_bb_fn); i++) {
> >>+ wa_bb[i]->offset = ALIGN(batch_ptr - batch, CACHELINE_BYTES);
> >>+ batch_ptr = wa_bb_fn[i](engine, batch_ptr);
> >>+ wa_bb[i]->size = batch_ptr - batch + wa_bb[i]->offset;
> >
> >wa_bb[i]->size = batch_ptr - (batch + wa_bb[i]->offset);
> >?
Also should it be
batch_ptr = wa_bb_fn[i](engine, batch + wa_bb[i]->offset);
?
--
Chris Wilson, Intel Open Source Technology Centre
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 21+ messages in thread
* Re: [PATCH 4/5] drm/i915: Tidy workaround batch buffer emission
2017-02-16 13:28 ` Chris Wilson
@ 2017-02-16 13:47 ` Tvrtko Ursulin
2017-02-16 13:51 ` [PATCH v4 1/2] " Tvrtko Ursulin
0 siblings, 1 reply; 21+ messages in thread
From: Tvrtko Ursulin @ 2017-02-16 13:47 UTC (permalink / raw)
To: Chris Wilson, Tvrtko Ursulin, Intel-gfx
On 16/02/2017 13:28, Chris Wilson wrote:
> On Thu, Feb 16, 2017 at 01:23:51PM +0000, Tvrtko Ursulin wrote:
>>
>> On 16/02/2017 13:20, Chris Wilson wrote:
>>> On Thu, Feb 16, 2017 at 12:23:24PM +0000, Tvrtko Ursulin wrote:
>>>> + /*
>>>> + * Emit the two workaround batch buffers, recording the offset from the
>>>> + * start of the workaround batch buffer object for each and their
>>>> + * respective sizes.
>>>> + */
>>>> + for (i = 0; i < ARRAY_SIZE(wa_bb_fn); i++) {
>>>> + wa_bb[i]->offset = ALIGN(batch_ptr - batch, CACHELINE_BYTES);
>>>> + batch_ptr = wa_bb_fn[i](engine, batch_ptr);
>>>> + wa_bb[i]->size = batch_ptr - batch + wa_bb[i]->offset;
>>>
>>> wa_bb[i]->size = batch_ptr - (batch + wa_bb[i]->offset);
>>> ?
>
> Also should it be
>
> batch_ptr = wa_bb_fn[i](engine, batch + wa_bb[i]->offset);
No but you made me realize it should already be aligned by contract
expected from the vfunc which emits the indirect bb. I'll add a WARN and
abort if it is not.
Regards,
Tvrtko
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 21+ messages in thread
* [PATCH v4 1/2] drm/i915: Tidy workaround batch buffer emission
2017-02-16 13:47 ` Tvrtko Ursulin
@ 2017-02-16 13:51 ` Tvrtko Ursulin
2017-02-16 13:54 ` [PATCH v5 " Tvrtko Ursulin
0 siblings, 1 reply; 21+ messages in thread
From: Tvrtko Ursulin @ 2017-02-16 13:51 UTC (permalink / raw)
To: Intel-gfx
From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Use the "*batch++ = " style as in the ring emission for better
readability and also simplify the logic a bit by consolidating
the offset and size calculations and overflow checking. The
latter is a programming error so it is not required to check
for it after each write to the object, but rather do it once the
whole state has been written and fail the driver if something
went wrong.
v2: Rebase.
v3: Keep track of offsets and sizes in bytes for simplicity
and rename function pointer variable to _fn suffix.
(Chris Wilson)
v4: Fix size calc broken in v3 and add alignment warning. (Chris Wilson)
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> (v2)
Cc: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/intel_lrc.c | 316 ++++++++++++++-------------------------
1 file changed, 115 insertions(+), 201 deletions(-)
diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c
index ee431d39ce06..2d839f364eb4 100644
--- a/drivers/gpu/drm/i915/intel_lrc.c
+++ b/drivers/gpu/drm/i915/intel_lrc.c
@@ -890,18 +890,6 @@ static int execlists_request_alloc(struct drm_i915_gem_request *request)
return ret;
}
-#define wa_ctx_emit(batch, index, cmd) \
- do { \
- int __index = (index)++; \
- if (WARN_ON(__index >= (PAGE_SIZE / sizeof(uint32_t)))) { \
- return -ENOSPC; \
- } \
- batch[__index] = (cmd); \
- } while (0)
-
-#define wa_ctx_emit_reg(batch, index, reg) \
- wa_ctx_emit((batch), (index), i915_mmio_reg_offset(reg))
-
/*
* In this WA we need to set GEN8_L3SQCREG4[21:21] and reset it after
* PIPE_CONTROL instruction. This is required for the flush to happen correctly
@@ -918,56 +906,31 @@ static int execlists_request_alloc(struct drm_i915_gem_request *request)
* This WA is also required for Gen9 so extracting as a function avoids
* code duplication.
*/
-static inline int gen8_emit_flush_coherentl3_wa(struct intel_engine_cs *engine,
- uint32_t *batch,
- uint32_t index)
-{
- uint32_t l3sqc4_flush = (0x40400000 | GEN8_LQSC_FLUSH_COHERENT_LINES);
-
- wa_ctx_emit(batch, index, (MI_STORE_REGISTER_MEM_GEN8 |
- MI_SRM_LRM_GLOBAL_GTT));
- wa_ctx_emit_reg(batch, index, GEN8_L3SQCREG4);
- wa_ctx_emit(batch, index, i915_ggtt_offset(engine->scratch) + 256);
- wa_ctx_emit(batch, index, 0);
-
- wa_ctx_emit(batch, index, MI_LOAD_REGISTER_IMM(1));
- wa_ctx_emit_reg(batch, index, GEN8_L3SQCREG4);
- wa_ctx_emit(batch, index, l3sqc4_flush);
-
- wa_ctx_emit(batch, index, GFX_OP_PIPE_CONTROL(6));
- wa_ctx_emit(batch, index, (PIPE_CONTROL_CS_STALL |
- PIPE_CONTROL_DC_FLUSH_ENABLE));
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
-
- wa_ctx_emit(batch, index, (MI_LOAD_REGISTER_MEM_GEN8 |
- MI_SRM_LRM_GLOBAL_GTT));
- wa_ctx_emit_reg(batch, index, GEN8_L3SQCREG4);
- wa_ctx_emit(batch, index, i915_ggtt_offset(engine->scratch) + 256);
- wa_ctx_emit(batch, index, 0);
-
- return index;
-}
-
-static inline uint32_t wa_ctx_start(struct i915_wa_ctx_bb *wa_ctx,
- uint32_t offset,
- uint32_t start_alignment)
-{
- return wa_ctx->offset = ALIGN(offset, start_alignment);
-}
-
-static inline int wa_ctx_end(struct i915_wa_ctx_bb *wa_ctx,
- uint32_t offset,
- uint32_t size_alignment)
+static u32 *
+gen8_emit_flush_coherentl3_wa(struct intel_engine_cs *engine, u32 *batch)
{
- wa_ctx->size = offset - wa_ctx->offset;
-
- WARN(wa_ctx->size % size_alignment,
- "wa_ctx_bb failed sanity checks: size %d is not aligned to %d\n",
- wa_ctx->size, size_alignment);
- return 0;
+ *batch++ = MI_STORE_REGISTER_MEM_GEN8 | MI_SRM_LRM_GLOBAL_GTT;
+ *batch++ = i915_mmio_reg_offset(GEN8_L3SQCREG4);
+ *batch++ = i915_ggtt_offset(engine->scratch) + 256;
+ *batch++ = 0;
+
+ *batch++ = MI_LOAD_REGISTER_IMM(1);
+ *batch++ = i915_mmio_reg_offset(GEN8_L3SQCREG4);
+ *batch++ = 0x40400000 | GEN8_LQSC_FLUSH_COHERENT_LINES;
+
+ *batch++ = GFX_OP_PIPE_CONTROL(6);
+ *batch++ = PIPE_CONTROL_CS_STALL | PIPE_CONTROL_DC_FLUSH_ENABLE;
+ *batch++ = 0;
+ *batch++ = 0;
+ *batch++ = 0;
+ *batch++ = 0;
+
+ *batch++ = MI_LOAD_REGISTER_MEM_GEN8 | MI_SRM_LRM_GLOBAL_GTT;
+ *batch++ = i915_mmio_reg_offset(GEN8_L3SQCREG4);
+ *batch++ = i915_ggtt_offset(engine->scratch) + 256;
+ *batch++ = 0;
+
+ return batch;
}
/*
@@ -985,42 +948,28 @@ static inline int wa_ctx_end(struct i915_wa_ctx_bb *wa_ctx,
* MI_BATCH_BUFFER_END will be added to perctx batch and both of them together
* makes a complete batch buffer.
*/
-static int gen8_init_indirectctx_bb(struct intel_engine_cs *engine,
- struct i915_wa_ctx_bb *wa_ctx,
- uint32_t *batch,
- uint32_t *offset)
+static u32 *gen8_init_indirectctx_bb(struct intel_engine_cs *engine, u32 *batch)
{
- uint32_t scratch_addr;
- uint32_t index = wa_ctx_start(wa_ctx, *offset, CACHELINE_DWORDS);
-
/* WaDisableCtxRestoreArbitration:bdw,chv */
- wa_ctx_emit(batch, index, MI_ARB_ON_OFF | MI_ARB_DISABLE);
+ *batch++ = MI_ARB_ON_OFF | MI_ARB_DISABLE;
/* WaFlushCoherentL3CacheLinesAtContextSwitch:bdw */
- if (IS_BROADWELL(engine->i915)) {
- int rc = gen8_emit_flush_coherentl3_wa(engine, batch, index);
- if (rc < 0)
- return rc;
- index = rc;
- }
+ if (IS_BROADWELL(engine->i915))
+ batch = gen8_emit_flush_coherentl3_wa(engine, batch);
+ *batch++ = GFX_OP_PIPE_CONTROL(6);
+ *batch++ = PIPE_CONTROL_FLUSH_L3 | PIPE_CONTROL_GLOBAL_GTT_IVB |
+ PIPE_CONTROL_CS_STALL | PIPE_CONTROL_QW_WRITE;
/* WaClearSlmSpaceAtContextSwitch:bdw,chv */
/* Actual scratch location is at 128 bytes offset */
- scratch_addr = i915_ggtt_offset(engine->scratch) + 2 * CACHELINE_BYTES;
-
- wa_ctx_emit(batch, index, GFX_OP_PIPE_CONTROL(6));
- wa_ctx_emit(batch, index, (PIPE_CONTROL_FLUSH_L3 |
- PIPE_CONTROL_GLOBAL_GTT_IVB |
- PIPE_CONTROL_CS_STALL |
- PIPE_CONTROL_QW_WRITE));
- wa_ctx_emit(batch, index, scratch_addr);
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
+ *batch++ = i915_ggtt_offset(engine->scratch) + 2 * CACHELINE_BYTES;
+ *batch++ = 0;
+ *batch++ = 0;
+ *batch++ = 0;
/* Pad to end of cacheline */
- while (index % CACHELINE_DWORDS)
- wa_ctx_emit(batch, index, MI_NOOP);
+ while ((unsigned long)batch % CACHELINE_BYTES)
+ *batch++ = MI_NOOP;
/*
* MI_BATCH_BUFFER_END is not required in Indirect ctx BB because
@@ -1028,7 +977,7 @@ static int gen8_init_indirectctx_bb(struct intel_engine_cs *engine,
* in the register CTX_RCS_INDIRECT_CTX
*/
- return wa_ctx_end(wa_ctx, *offset = index, CACHELINE_DWORDS);
+ return batch;
}
/*
@@ -1040,58 +989,38 @@ static int gen8_init_indirectctx_bb(struct intel_engine_cs *engine,
* This batch is terminated with MI_BATCH_BUFFER_END and so we need not add padding
* to align it with cacheline as padding after MI_BATCH_BUFFER_END is redundant.
*/
-static int gen8_init_perctx_bb(struct intel_engine_cs *engine,
- struct i915_wa_ctx_bb *wa_ctx,
- uint32_t *batch,
- uint32_t *offset)
+static u32 *gen8_init_perctx_bb(struct intel_engine_cs *engine, u32 *batch)
{
- uint32_t index = wa_ctx_start(wa_ctx, *offset, CACHELINE_DWORDS);
-
/* WaDisableCtxRestoreArbitration:bdw,chv */
- wa_ctx_emit(batch, index, MI_ARB_ON_OFF | MI_ARB_ENABLE);
+ *batch++ = MI_ARB_ON_OFF | MI_ARB_ENABLE;
+ *batch++ = MI_BATCH_BUFFER_END;
- wa_ctx_emit(batch, index, MI_BATCH_BUFFER_END);
-
- return wa_ctx_end(wa_ctx, *offset = index, 1);
+ return batch;
}
-static int gen9_init_indirectctx_bb(struct intel_engine_cs *engine,
- struct i915_wa_ctx_bb *wa_ctx,
- uint32_t *batch,
- uint32_t *offset)
+static u32 *gen9_init_indirectctx_bb(struct intel_engine_cs *engine, u32 *batch)
{
- int ret;
- struct drm_i915_private *dev_priv = engine->i915;
- uint32_t index = wa_ctx_start(wa_ctx, *offset, CACHELINE_DWORDS);
-
/* WaFlushCoherentL3CacheLinesAtContextSwitch:skl,bxt,glk */
- ret = gen8_emit_flush_coherentl3_wa(engine, batch, index);
- if (ret < 0)
- return ret;
- index = ret;
+ batch = gen8_emit_flush_coherentl3_wa(engine, batch);
/* WaDisableGatherAtSetShaderCommonSlice:skl,bxt,kbl,glk */
- wa_ctx_emit(batch, index, MI_LOAD_REGISTER_IMM(1));
- wa_ctx_emit_reg(batch, index, COMMON_SLICE_CHICKEN2);
- wa_ctx_emit(batch, index, _MASKED_BIT_DISABLE(
- GEN9_DISABLE_GATHER_AT_SET_SHADER_COMMON_SLICE));
- wa_ctx_emit(batch, index, MI_NOOP);
+ *batch++ = MI_LOAD_REGISTER_IMM(1);
+ *batch++ = i915_mmio_reg_offset(COMMON_SLICE_CHICKEN2);
+ *batch++ = _MASKED_BIT_DISABLE(
+ GEN9_DISABLE_GATHER_AT_SET_SHADER_COMMON_SLICE);
+ *batch++ = MI_NOOP;
/* WaClearSlmSpaceAtContextSwitch:kbl */
/* Actual scratch location is at 128 bytes offset */
- if (IS_KBL_REVID(dev_priv, 0, KBL_REVID_A0)) {
- u32 scratch_addr =
- i915_ggtt_offset(engine->scratch) + 2 * CACHELINE_BYTES;
-
- wa_ctx_emit(batch, index, GFX_OP_PIPE_CONTROL(6));
- wa_ctx_emit(batch, index, (PIPE_CONTROL_FLUSH_L3 |
- PIPE_CONTROL_GLOBAL_GTT_IVB |
- PIPE_CONTROL_CS_STALL |
- PIPE_CONTROL_QW_WRITE));
- wa_ctx_emit(batch, index, scratch_addr);
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
+ if (IS_KBL_REVID(engine->i915, 0, KBL_REVID_A0)) {
+ *batch++ = GFX_OP_PIPE_CONTROL(6);
+ *batch++ = PIPE_CONTROL_FLUSH_L3 | PIPE_CONTROL_GLOBAL_GTT_IVB |
+ PIPE_CONTROL_CS_STALL | PIPE_CONTROL_QW_WRITE;
+ *batch++ = i915_ggtt_offset(engine->scratch) +
+ 2 * CACHELINE_BYTES;
+ *batch++ = 0;
+ *batch++ = 0;
+ *batch++ = 0;
}
/* WaMediaPoolStateCmdInWABB:bxt,glk */
@@ -1109,41 +1038,37 @@ static int gen9_init_indirectctx_bb(struct intel_engine_cs *engine,
* possible configurations, to avoid duplication they are
* not shown here again.
*/
- u32 eu_pool_config = 0x00777000;
- wa_ctx_emit(batch, index, GEN9_MEDIA_POOL_STATE);
- wa_ctx_emit(batch, index, GEN9_MEDIA_POOL_ENABLE);
- wa_ctx_emit(batch, index, eu_pool_config);
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
+ *batch++ = GEN9_MEDIA_POOL_STATE;
+ *batch++ = GEN9_MEDIA_POOL_ENABLE;
+ *batch++ = 0x00777000;
+ *batch++ = 0;
+ *batch++ = 0;
+ *batch++ = 0;
}
/* Pad to end of cacheline */
- while (index % CACHELINE_DWORDS)
- wa_ctx_emit(batch, index, MI_NOOP);
+ while ((unsigned long)batch % CACHELINE_BYTES)
+ *batch++ = MI_NOOP;
- return wa_ctx_end(wa_ctx, *offset = index, CACHELINE_DWORDS);
+ return batch;
}
-static int gen9_init_perctx_bb(struct intel_engine_cs *engine,
- struct i915_wa_ctx_bb *wa_ctx,
- uint32_t *batch,
- uint32_t *offset)
+static u32 *gen9_init_perctx_bb(struct intel_engine_cs *engine, u32 *batch)
{
- uint32_t index = wa_ctx_start(wa_ctx, *offset, CACHELINE_DWORDS);
-
- wa_ctx_emit(batch, index, MI_BATCH_BUFFER_END);
+ *batch++ = MI_BATCH_BUFFER_END;
- return wa_ctx_end(wa_ctx, *offset = index, 1);
+ return batch;
}
-static int lrc_setup_wa_ctx_obj(struct intel_engine_cs *engine, u32 size)
+#define CTX_WA_BB_OBJ_SIZE (PAGE_SIZE)
+
+static int lrc_setup_wa_ctx(struct intel_engine_cs *engine)
{
struct drm_i915_gem_object *obj;
struct i915_vma *vma;
int err;
- obj = i915_gem_object_create(engine->i915, PAGE_ALIGN(size));
+ obj = i915_gem_object_create(engine->i915, CTX_WA_BB_OBJ_SIZE);
if (IS_ERR(obj))
return PTR_ERR(obj);
@@ -1165,80 +1090,70 @@ static int lrc_setup_wa_ctx_obj(struct intel_engine_cs *engine, u32 size)
return err;
}
-static void lrc_destroy_wa_ctx_obj(struct intel_engine_cs *engine)
+static void lrc_destroy_wa_ctx(struct intel_engine_cs *engine)
{
i915_vma_unpin_and_release(&engine->wa_ctx.vma);
}
+typedef u32 *(*wa_bb_func_t)(struct intel_engine_cs *engine, u32 *batch);
+
static int intel_init_workaround_bb(struct intel_engine_cs *engine)
{
struct i915_ctx_workarounds *wa_ctx = &engine->wa_ctx;
- uint32_t *batch;
- uint32_t offset;
+ struct i915_wa_ctx_bb *wa_bb[2] = { &wa_ctx->indirect_ctx,
+ &wa_ctx->per_ctx };
+ wa_bb_func_t wa_bb_fn[2];
struct page *page;
+ void *batch, *batch_ptr;
+ unsigned int i;
int ret;
- WARN_ON(engine->id != RCS);
+ if (WARN_ON(engine->id != RCS || !engine->scratch))
+ return -EINVAL;
- /* update this when WA for higher Gen are added */
- if (INTEL_GEN(engine->i915) > 9) {
- DRM_ERROR("WA batch buffer is not initialized for Gen%d\n",
- INTEL_GEN(engine->i915));
+ switch (INTEL_GEN(engine->i915)) {
+ case 9:
+ wa_bb_fn[0] = gen9_init_indirectctx_bb;
+ wa_bb_fn[1] = gen9_init_perctx_bb;
+ break;
+ case 8:
+ wa_bb_fn[0] = gen8_init_indirectctx_bb;
+ wa_bb_fn[1] = gen8_init_perctx_bb;
+ break;
+ default:
+ MISSING_CASE(INTEL_GEN(engine->i915));
return 0;
}
- /* some WA perform writes to scratch page, ensure it is valid */
- if (!engine->scratch) {
- DRM_ERROR("scratch page not allocated for %s\n", engine->name);
- return -EINVAL;
- }
-
- ret = lrc_setup_wa_ctx_obj(engine, PAGE_SIZE);
+ ret = lrc_setup_wa_ctx(engine);
if (ret) {
DRM_DEBUG_DRIVER("Failed to setup context WA page: %d\n", ret);
return ret;
}
page = i915_gem_object_get_dirty_page(wa_ctx->vma->obj, 0);
- batch = kmap_atomic(page);
- offset = 0;
-
- if (IS_GEN8(engine->i915)) {
- ret = gen8_init_indirectctx_bb(engine,
- &wa_ctx->indirect_ctx,
- batch,
- &offset);
- if (ret)
- goto out;
-
- ret = gen8_init_perctx_bb(engine,
- &wa_ctx->per_ctx,
- batch,
- &offset);
- if (ret)
- goto out;
- } else if (IS_GEN9(engine->i915)) {
- ret = gen9_init_indirectctx_bb(engine,
- &wa_ctx->indirect_ctx,
- batch,
- &offset);
- if (ret)
- goto out;
+ batch = batch_ptr = kmap_atomic(page);
- ret = gen9_init_perctx_bb(engine,
- &wa_ctx->per_ctx,
- batch,
- &offset);
- if (ret)
+ /*
+ * Emit the two workaround batch buffers, recording the offset from the
+ * start of the workaround batch buffer object for each and their
+ * respective sizes.
+ */
+ for (i = 0; i < ARRAY_SIZE(wa_bb_fn); i++) {
+ wa_bb[i]->offset = batch_ptr - batch;
+ if (WARN_ON(!IS_ALIGNED(wa_bb[i]->offset, CACHELINE_BYTES))) {
+ ret = -EINVAL;
goto out;
+ }
+ batch_ptr = wa_bb_fn[i](engine, batch_ptr);
+ wa_bb[i]->size = batch_ptr - (batch + wa_bb[i]->offset);
}
+ BUG_ON(batch_ptr - batch > CTX_WA_BB_OBJ_SIZE);
out:
kunmap_atomic(batch);
- if (ret)
- lrc_destroy_wa_ctx_obj(engine);
- return ret;
+ return 0;
}
static u32 port_seqno(struct execlist_port *port)
@@ -1685,7 +1600,7 @@ void intel_logical_ring_cleanup(struct intel_engine_cs *engine)
intel_engine_cleanup_common(engine);
- lrc_destroy_wa_ctx_obj(engine);
+ lrc_destroy_wa_ctx(engine);
engine->i915 = NULL;
dev_priv->engine[engine->id] = NULL;
kfree(engine);
@@ -1971,15 +1886,14 @@ static void execlists_init_reg_state(u32 *reg_state,
u32 ggtt_offset = i915_ggtt_offset(wa_ctx->vma);
reg_state[CTX_RCS_INDIRECT_CTX+1] =
- (ggtt_offset + wa_ctx->indirect_ctx.offset * sizeof(uint32_t)) |
- (wa_ctx->indirect_ctx.size / CACHELINE_DWORDS);
+ (ggtt_offset + wa_ctx->indirect_ctx.offset) |
+ (wa_ctx->indirect_ctx.size / CACHELINE_BYTES);
reg_state[CTX_RCS_INDIRECT_CTX_OFFSET+1] =
intel_lr_indirect_ctx_offset(engine) << 6;
reg_state[CTX_BB_PER_CTX_PTR+1] =
- (ggtt_offset + wa_ctx->per_ctx.offset * sizeof(uint32_t)) |
- 0x01;
+ (ggtt_offset + wa_ctx->per_ctx.offset) | 0x01;
}
}
reg_state[CTX_LRI_HEADER_1] = MI_LOAD_REGISTER_IMM(9) | MI_LRI_FORCE_POSTED;
--
2.9.3
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 21+ messages in thread
* [PATCH v5 1/2] drm/i915: Tidy workaround batch buffer emission
2017-02-16 13:51 ` [PATCH v4 1/2] " Tvrtko Ursulin
@ 2017-02-16 13:54 ` Tvrtko Ursulin
2017-02-16 14:03 ` Chris Wilson
2017-02-17 7:58 ` [PATCH v6] " Tvrtko Ursulin
0 siblings, 2 replies; 21+ messages in thread
From: Tvrtko Ursulin @ 2017-02-16 13:54 UTC (permalink / raw)
To: Intel-gfx
From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Use the "*batch++ = " style as in the ring emission for better
readability and also simplify the logic a bit by consolidating
the offset and size calculations and overflow checking. The
latter is a programming error so it is not required to check
for it after each write to the object, but rather do it once the
whole state has been written and fail the driver if something
went wrong.
v2: Rebase.
v3: Keep track of offsets and sizes in bytes for simplicity
and rename function pointer variable to _fn suffix.
(Chris Wilson)
v4: Fix size calc broken in v3 and add alignment warning. (Chris Wilson)
v5: Fix return code.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> (v2)
Cc: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/intel_lrc.c | 314 ++++++++++++++-------------------------
1 file changed, 114 insertions(+), 200 deletions(-)
diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c
index ee431d39ce06..4c2dfadb9883 100644
--- a/drivers/gpu/drm/i915/intel_lrc.c
+++ b/drivers/gpu/drm/i915/intel_lrc.c
@@ -890,18 +890,6 @@ static int execlists_request_alloc(struct drm_i915_gem_request *request)
return ret;
}
-#define wa_ctx_emit(batch, index, cmd) \
- do { \
- int __index = (index)++; \
- if (WARN_ON(__index >= (PAGE_SIZE / sizeof(uint32_t)))) { \
- return -ENOSPC; \
- } \
- batch[__index] = (cmd); \
- } while (0)
-
-#define wa_ctx_emit_reg(batch, index, reg) \
- wa_ctx_emit((batch), (index), i915_mmio_reg_offset(reg))
-
/*
* In this WA we need to set GEN8_L3SQCREG4[21:21] and reset it after
* PIPE_CONTROL instruction. This is required for the flush to happen correctly
@@ -918,56 +906,31 @@ static int execlists_request_alloc(struct drm_i915_gem_request *request)
* This WA is also required for Gen9 so extracting as a function avoids
* code duplication.
*/
-static inline int gen8_emit_flush_coherentl3_wa(struct intel_engine_cs *engine,
- uint32_t *batch,
- uint32_t index)
-{
- uint32_t l3sqc4_flush = (0x40400000 | GEN8_LQSC_FLUSH_COHERENT_LINES);
-
- wa_ctx_emit(batch, index, (MI_STORE_REGISTER_MEM_GEN8 |
- MI_SRM_LRM_GLOBAL_GTT));
- wa_ctx_emit_reg(batch, index, GEN8_L3SQCREG4);
- wa_ctx_emit(batch, index, i915_ggtt_offset(engine->scratch) + 256);
- wa_ctx_emit(batch, index, 0);
-
- wa_ctx_emit(batch, index, MI_LOAD_REGISTER_IMM(1));
- wa_ctx_emit_reg(batch, index, GEN8_L3SQCREG4);
- wa_ctx_emit(batch, index, l3sqc4_flush);
-
- wa_ctx_emit(batch, index, GFX_OP_PIPE_CONTROL(6));
- wa_ctx_emit(batch, index, (PIPE_CONTROL_CS_STALL |
- PIPE_CONTROL_DC_FLUSH_ENABLE));
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
-
- wa_ctx_emit(batch, index, (MI_LOAD_REGISTER_MEM_GEN8 |
- MI_SRM_LRM_GLOBAL_GTT));
- wa_ctx_emit_reg(batch, index, GEN8_L3SQCREG4);
- wa_ctx_emit(batch, index, i915_ggtt_offset(engine->scratch) + 256);
- wa_ctx_emit(batch, index, 0);
-
- return index;
-}
-
-static inline uint32_t wa_ctx_start(struct i915_wa_ctx_bb *wa_ctx,
- uint32_t offset,
- uint32_t start_alignment)
+static u32 *
+gen8_emit_flush_coherentl3_wa(struct intel_engine_cs *engine, u32 *batch)
{
- return wa_ctx->offset = ALIGN(offset, start_alignment);
-}
-
-static inline int wa_ctx_end(struct i915_wa_ctx_bb *wa_ctx,
- uint32_t offset,
- uint32_t size_alignment)
-{
- wa_ctx->size = offset - wa_ctx->offset;
-
- WARN(wa_ctx->size % size_alignment,
- "wa_ctx_bb failed sanity checks: size %d is not aligned to %d\n",
- wa_ctx->size, size_alignment);
- return 0;
+ *batch++ = MI_STORE_REGISTER_MEM_GEN8 | MI_SRM_LRM_GLOBAL_GTT;
+ *batch++ = i915_mmio_reg_offset(GEN8_L3SQCREG4);
+ *batch++ = i915_ggtt_offset(engine->scratch) + 256;
+ *batch++ = 0;
+
+ *batch++ = MI_LOAD_REGISTER_IMM(1);
+ *batch++ = i915_mmio_reg_offset(GEN8_L3SQCREG4);
+ *batch++ = 0x40400000 | GEN8_LQSC_FLUSH_COHERENT_LINES;
+
+ *batch++ = GFX_OP_PIPE_CONTROL(6);
+ *batch++ = PIPE_CONTROL_CS_STALL | PIPE_CONTROL_DC_FLUSH_ENABLE;
+ *batch++ = 0;
+ *batch++ = 0;
+ *batch++ = 0;
+ *batch++ = 0;
+
+ *batch++ = MI_LOAD_REGISTER_MEM_GEN8 | MI_SRM_LRM_GLOBAL_GTT;
+ *batch++ = i915_mmio_reg_offset(GEN8_L3SQCREG4);
+ *batch++ = i915_ggtt_offset(engine->scratch) + 256;
+ *batch++ = 0;
+
+ return batch;
}
/*
@@ -985,42 +948,28 @@ static inline int wa_ctx_end(struct i915_wa_ctx_bb *wa_ctx,
* MI_BATCH_BUFFER_END will be added to perctx batch and both of them together
* makes a complete batch buffer.
*/
-static int gen8_init_indirectctx_bb(struct intel_engine_cs *engine,
- struct i915_wa_ctx_bb *wa_ctx,
- uint32_t *batch,
- uint32_t *offset)
+static u32 *gen8_init_indirectctx_bb(struct intel_engine_cs *engine, u32 *batch)
{
- uint32_t scratch_addr;
- uint32_t index = wa_ctx_start(wa_ctx, *offset, CACHELINE_DWORDS);
-
/* WaDisableCtxRestoreArbitration:bdw,chv */
- wa_ctx_emit(batch, index, MI_ARB_ON_OFF | MI_ARB_DISABLE);
+ *batch++ = MI_ARB_ON_OFF | MI_ARB_DISABLE;
/* WaFlushCoherentL3CacheLinesAtContextSwitch:bdw */
- if (IS_BROADWELL(engine->i915)) {
- int rc = gen8_emit_flush_coherentl3_wa(engine, batch, index);
- if (rc < 0)
- return rc;
- index = rc;
- }
+ if (IS_BROADWELL(engine->i915))
+ batch = gen8_emit_flush_coherentl3_wa(engine, batch);
+ *batch++ = GFX_OP_PIPE_CONTROL(6);
+ *batch++ = PIPE_CONTROL_FLUSH_L3 | PIPE_CONTROL_GLOBAL_GTT_IVB |
+ PIPE_CONTROL_CS_STALL | PIPE_CONTROL_QW_WRITE;
/* WaClearSlmSpaceAtContextSwitch:bdw,chv */
/* Actual scratch location is at 128 bytes offset */
- scratch_addr = i915_ggtt_offset(engine->scratch) + 2 * CACHELINE_BYTES;
-
- wa_ctx_emit(batch, index, GFX_OP_PIPE_CONTROL(6));
- wa_ctx_emit(batch, index, (PIPE_CONTROL_FLUSH_L3 |
- PIPE_CONTROL_GLOBAL_GTT_IVB |
- PIPE_CONTROL_CS_STALL |
- PIPE_CONTROL_QW_WRITE));
- wa_ctx_emit(batch, index, scratch_addr);
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
+ *batch++ = i915_ggtt_offset(engine->scratch) + 2 * CACHELINE_BYTES;
+ *batch++ = 0;
+ *batch++ = 0;
+ *batch++ = 0;
/* Pad to end of cacheline */
- while (index % CACHELINE_DWORDS)
- wa_ctx_emit(batch, index, MI_NOOP);
+ while ((unsigned long)batch % CACHELINE_BYTES)
+ *batch++ = MI_NOOP;
/*
* MI_BATCH_BUFFER_END is not required in Indirect ctx BB because
@@ -1028,7 +977,7 @@ static int gen8_init_indirectctx_bb(struct intel_engine_cs *engine,
* in the register CTX_RCS_INDIRECT_CTX
*/
- return wa_ctx_end(wa_ctx, *offset = index, CACHELINE_DWORDS);
+ return batch;
}
/*
@@ -1040,58 +989,38 @@ static int gen8_init_indirectctx_bb(struct intel_engine_cs *engine,
* This batch is terminated with MI_BATCH_BUFFER_END and so we need not add padding
* to align it with cacheline as padding after MI_BATCH_BUFFER_END is redundant.
*/
-static int gen8_init_perctx_bb(struct intel_engine_cs *engine,
- struct i915_wa_ctx_bb *wa_ctx,
- uint32_t *batch,
- uint32_t *offset)
+static u32 *gen8_init_perctx_bb(struct intel_engine_cs *engine, u32 *batch)
{
- uint32_t index = wa_ctx_start(wa_ctx, *offset, CACHELINE_DWORDS);
-
/* WaDisableCtxRestoreArbitration:bdw,chv */
- wa_ctx_emit(batch, index, MI_ARB_ON_OFF | MI_ARB_ENABLE);
-
- wa_ctx_emit(batch, index, MI_BATCH_BUFFER_END);
+ *batch++ = MI_ARB_ON_OFF | MI_ARB_ENABLE;
+ *batch++ = MI_BATCH_BUFFER_END;
- return wa_ctx_end(wa_ctx, *offset = index, 1);
+ return batch;
}
-static int gen9_init_indirectctx_bb(struct intel_engine_cs *engine,
- struct i915_wa_ctx_bb *wa_ctx,
- uint32_t *batch,
- uint32_t *offset)
+static u32 *gen9_init_indirectctx_bb(struct intel_engine_cs *engine, u32 *batch)
{
- int ret;
- struct drm_i915_private *dev_priv = engine->i915;
- uint32_t index = wa_ctx_start(wa_ctx, *offset, CACHELINE_DWORDS);
-
/* WaFlushCoherentL3CacheLinesAtContextSwitch:skl,bxt,glk */
- ret = gen8_emit_flush_coherentl3_wa(engine, batch, index);
- if (ret < 0)
- return ret;
- index = ret;
+ batch = gen8_emit_flush_coherentl3_wa(engine, batch);
/* WaDisableGatherAtSetShaderCommonSlice:skl,bxt,kbl,glk */
- wa_ctx_emit(batch, index, MI_LOAD_REGISTER_IMM(1));
- wa_ctx_emit_reg(batch, index, COMMON_SLICE_CHICKEN2);
- wa_ctx_emit(batch, index, _MASKED_BIT_DISABLE(
- GEN9_DISABLE_GATHER_AT_SET_SHADER_COMMON_SLICE));
- wa_ctx_emit(batch, index, MI_NOOP);
+ *batch++ = MI_LOAD_REGISTER_IMM(1);
+ *batch++ = i915_mmio_reg_offset(COMMON_SLICE_CHICKEN2);
+ *batch++ = _MASKED_BIT_DISABLE(
+ GEN9_DISABLE_GATHER_AT_SET_SHADER_COMMON_SLICE);
+ *batch++ = MI_NOOP;
/* WaClearSlmSpaceAtContextSwitch:kbl */
/* Actual scratch location is at 128 bytes offset */
- if (IS_KBL_REVID(dev_priv, 0, KBL_REVID_A0)) {
- u32 scratch_addr =
- i915_ggtt_offset(engine->scratch) + 2 * CACHELINE_BYTES;
-
- wa_ctx_emit(batch, index, GFX_OP_PIPE_CONTROL(6));
- wa_ctx_emit(batch, index, (PIPE_CONTROL_FLUSH_L3 |
- PIPE_CONTROL_GLOBAL_GTT_IVB |
- PIPE_CONTROL_CS_STALL |
- PIPE_CONTROL_QW_WRITE));
- wa_ctx_emit(batch, index, scratch_addr);
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
+ if (IS_KBL_REVID(engine->i915, 0, KBL_REVID_A0)) {
+ *batch++ = GFX_OP_PIPE_CONTROL(6);
+ *batch++ = PIPE_CONTROL_FLUSH_L3 | PIPE_CONTROL_GLOBAL_GTT_IVB |
+ PIPE_CONTROL_CS_STALL | PIPE_CONTROL_QW_WRITE;
+ *batch++ = i915_ggtt_offset(engine->scratch) +
+ 2 * CACHELINE_BYTES;
+ *batch++ = 0;
+ *batch++ = 0;
+ *batch++ = 0;
}
/* WaMediaPoolStateCmdInWABB:bxt,glk */
@@ -1109,41 +1038,37 @@ static int gen9_init_indirectctx_bb(struct intel_engine_cs *engine,
* possible configurations, to avoid duplication they are
* not shown here again.
*/
- u32 eu_pool_config = 0x00777000;
- wa_ctx_emit(batch, index, GEN9_MEDIA_POOL_STATE);
- wa_ctx_emit(batch, index, GEN9_MEDIA_POOL_ENABLE);
- wa_ctx_emit(batch, index, eu_pool_config);
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
+ *batch++ = GEN9_MEDIA_POOL_STATE;
+ *batch++ = GEN9_MEDIA_POOL_ENABLE;
+ *batch++ = 0x00777000;
+ *batch++ = 0;
+ *batch++ = 0;
+ *batch++ = 0;
}
/* Pad to end of cacheline */
- while (index % CACHELINE_DWORDS)
- wa_ctx_emit(batch, index, MI_NOOP);
+ while ((unsigned long)batch % CACHELINE_BYTES)
+ *batch++ = MI_NOOP;
- return wa_ctx_end(wa_ctx, *offset = index, CACHELINE_DWORDS);
+ return batch;
}
-static int gen9_init_perctx_bb(struct intel_engine_cs *engine,
- struct i915_wa_ctx_bb *wa_ctx,
- uint32_t *batch,
- uint32_t *offset)
+static u32 *gen9_init_perctx_bb(struct intel_engine_cs *engine, u32 *batch)
{
- uint32_t index = wa_ctx_start(wa_ctx, *offset, CACHELINE_DWORDS);
+ *batch++ = MI_BATCH_BUFFER_END;
- wa_ctx_emit(batch, index, MI_BATCH_BUFFER_END);
-
- return wa_ctx_end(wa_ctx, *offset = index, 1);
+ return batch;
}
-static int lrc_setup_wa_ctx_obj(struct intel_engine_cs *engine, u32 size)
+#define CTX_WA_BB_OBJ_SIZE (PAGE_SIZE)
+
+static int lrc_setup_wa_ctx(struct intel_engine_cs *engine)
{
struct drm_i915_gem_object *obj;
struct i915_vma *vma;
int err;
- obj = i915_gem_object_create(engine->i915, PAGE_ALIGN(size));
+ obj = i915_gem_object_create(engine->i915, CTX_WA_BB_OBJ_SIZE);
if (IS_ERR(obj))
return PTR_ERR(obj);
@@ -1165,78 +1090,68 @@ static int lrc_setup_wa_ctx_obj(struct intel_engine_cs *engine, u32 size)
return err;
}
-static void lrc_destroy_wa_ctx_obj(struct intel_engine_cs *engine)
+static void lrc_destroy_wa_ctx(struct intel_engine_cs *engine)
{
i915_vma_unpin_and_release(&engine->wa_ctx.vma);
}
+typedef u32 *(*wa_bb_func_t)(struct intel_engine_cs *engine, u32 *batch);
+
static int intel_init_workaround_bb(struct intel_engine_cs *engine)
{
struct i915_ctx_workarounds *wa_ctx = &engine->wa_ctx;
- uint32_t *batch;
- uint32_t offset;
+ struct i915_wa_ctx_bb *wa_bb[2] = { &wa_ctx->indirect_ctx,
+ &wa_ctx->per_ctx };
+ wa_bb_func_t wa_bb_fn[2];
struct page *page;
+ void *batch, *batch_ptr;
+ unsigned int i;
int ret;
- WARN_ON(engine->id != RCS);
+ if (WARN_ON(engine->id != RCS || !engine->scratch))
+ return -EINVAL;
- /* update this when WA for higher Gen are added */
- if (INTEL_GEN(engine->i915) > 9) {
- DRM_ERROR("WA batch buffer is not initialized for Gen%d\n",
- INTEL_GEN(engine->i915));
+ switch (INTEL_GEN(engine->i915)) {
+ case 9:
+ wa_bb_fn[0] = gen9_init_indirectctx_bb;
+ wa_bb_fn[1] = gen9_init_perctx_bb;
+ break;
+ case 8:
+ wa_bb_fn[0] = gen8_init_indirectctx_bb;
+ wa_bb_fn[1] = gen8_init_perctx_bb;
+ break;
+ default:
+ MISSING_CASE(INTEL_GEN(engine->i915));
return 0;
}
- /* some WA perform writes to scratch page, ensure it is valid */
- if (!engine->scratch) {
- DRM_ERROR("scratch page not allocated for %s\n", engine->name);
- return -EINVAL;
- }
-
- ret = lrc_setup_wa_ctx_obj(engine, PAGE_SIZE);
+ ret = lrc_setup_wa_ctx(engine);
if (ret) {
DRM_DEBUG_DRIVER("Failed to setup context WA page: %d\n", ret);
return ret;
}
page = i915_gem_object_get_dirty_page(wa_ctx->vma->obj, 0);
- batch = kmap_atomic(page);
- offset = 0;
-
- if (IS_GEN8(engine->i915)) {
- ret = gen8_init_indirectctx_bb(engine,
- &wa_ctx->indirect_ctx,
- batch,
- &offset);
- if (ret)
- goto out;
+ batch = batch_ptr = kmap_atomic(page);
- ret = gen8_init_perctx_bb(engine,
- &wa_ctx->per_ctx,
- batch,
- &offset);
- if (ret)
- goto out;
- } else if (IS_GEN9(engine->i915)) {
- ret = gen9_init_indirectctx_bb(engine,
- &wa_ctx->indirect_ctx,
- batch,
- &offset);
- if (ret)
- goto out;
-
- ret = gen9_init_perctx_bb(engine,
- &wa_ctx->per_ctx,
- batch,
- &offset);
- if (ret)
+ /*
+ * Emit the two workaround batch buffers, recording the offset from the
+ * start of the workaround batch buffer object for each and their
+ * respective sizes.
+ */
+ for (i = 0; i < ARRAY_SIZE(wa_bb_fn); i++) {
+ wa_bb[i]->offset = batch_ptr - batch;
+ if (WARN_ON(!IS_ALIGNED(wa_bb[i]->offset, CACHELINE_BYTES))) {
+ ret = -EINVAL;
goto out;
+ }
+ batch_ptr = wa_bb_fn[i](engine, batch_ptr);
+ wa_bb[i]->size = batch_ptr - (batch + wa_bb[i]->offset);
}
+ BUG_ON(batch_ptr - batch > CTX_WA_BB_OBJ_SIZE);
out:
kunmap_atomic(batch);
- if (ret)
- lrc_destroy_wa_ctx_obj(engine);
return ret;
}
@@ -1685,7 +1600,7 @@ void intel_logical_ring_cleanup(struct intel_engine_cs *engine)
intel_engine_cleanup_common(engine);
- lrc_destroy_wa_ctx_obj(engine);
+ lrc_destroy_wa_ctx(engine);
engine->i915 = NULL;
dev_priv->engine[engine->id] = NULL;
kfree(engine);
@@ -1971,15 +1886,14 @@ static void execlists_init_reg_state(u32 *reg_state,
u32 ggtt_offset = i915_ggtt_offset(wa_ctx->vma);
reg_state[CTX_RCS_INDIRECT_CTX+1] =
- (ggtt_offset + wa_ctx->indirect_ctx.offset * sizeof(uint32_t)) |
- (wa_ctx->indirect_ctx.size / CACHELINE_DWORDS);
+ (ggtt_offset + wa_ctx->indirect_ctx.offset) |
+ (wa_ctx->indirect_ctx.size / CACHELINE_BYTES);
reg_state[CTX_RCS_INDIRECT_CTX_OFFSET+1] =
intel_lr_indirect_ctx_offset(engine) << 6;
reg_state[CTX_BB_PER_CTX_PTR+1] =
- (ggtt_offset + wa_ctx->per_ctx.offset * sizeof(uint32_t)) |
- 0x01;
+ (ggtt_offset + wa_ctx->per_ctx.offset) | 0x01;
}
}
reg_state[CTX_LRI_HEADER_1] = MI_LOAD_REGISTER_IMM(9) | MI_LRI_FORCE_POSTED;
--
2.9.3
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 21+ messages in thread
* Re: [PATCH v5 1/2] drm/i915: Tidy workaround batch buffer emission
2017-02-16 13:54 ` [PATCH v5 " Tvrtko Ursulin
@ 2017-02-16 14:03 ` Chris Wilson
2017-02-17 7:58 ` [PATCH v6] " Tvrtko Ursulin
1 sibling, 0 replies; 21+ messages in thread
From: Chris Wilson @ 2017-02-16 14:03 UTC (permalink / raw)
To: Tvrtko Ursulin; +Cc: Intel-gfx
On Thu, Feb 16, 2017 at 01:54:02PM +0000, Tvrtko Ursulin wrote:
> From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
>
> Use the "*batch++ = " style as in the ring emission for better
> readability and also simplify the logic a bit by consolidating
> the offset and size calculations and overflow checking. The
> latter is a programming error so it is not required to check
> for it after each write to the object, but rather do it once the
> whole state has been written and fail the driver if something
> went wrong.
>
> v2: Rebase.
>
> v3: Keep track of offsets and sizes in bytes for simplicity
> and rename function pointer variable to _fn suffix.
> (Chris Wilson)
>
> v4: Fix size calc broken in v3 and add alignment warning. (Chris Wilson)
>
> v5: Fix return code.
>
> Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> (v2)
> Cc: Chris Wilson <chris@chris-wilson.co.uk>
> ---
> +static u32 *gen9_init_indirectctx_bb(struct intel_engine_cs *engine, u32 *batch)
> {
> /* Pad to end of cacheline */
> - while (index % CACHELINE_DWORDS)
> - wa_ctx_emit(batch, index, MI_NOOP);
> + while ((unsigned long)batch % CACHELINE_BYTES)
> + *batch++ = MI_NOOP;
>
> - return wa_ctx_end(wa_ctx, *offset = index, CACHELINE_DWORDS);
> + return batch;
> }
>
> -static int gen9_init_perctx_bb(struct intel_engine_cs *engine,
> - struct i915_wa_ctx_bb *wa_ctx,
> - uint32_t *batch,
> - uint32_t *offset)
> +static u32 *gen9_init_perctx_bb(struct intel_engine_cs *engine, u32 *batch)
> {
> - uint32_t index = wa_ctx_start(wa_ctx, *offset, CACHELINE_DWORDS);
> + *batch++ = MI_BATCH_BUFFER_END;
>
> - wa_ctx_emit(batch, index, MI_BATCH_BUFFER_END);
> -
> - return wa_ctx_end(wa_ctx, *offset = index, 1);
> + return batch;
> }
> + /*
> + * Emit the two workaround batch buffers, recording the offset from the
> + * start of the workaround batch buffer object for each and their
> + * respective sizes.
> + */
> + for (i = 0; i < ARRAY_SIZE(wa_bb_fn); i++) {
> + wa_bb[i]->offset = batch_ptr - batch;
> + if (WARN_ON(!IS_ALIGNED(wa_bb[i]->offset, CACHELINE_BYTES))) {
> + ret = -EINVAL;
> goto out;
> + }
Looks like a trap! If we add a third wa bb and forget to align the end
of the second.
Not a problem today, so
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
-Chris
--
Chris Wilson, Intel Open Source Technology Centre
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 21+ messages in thread
* ✓ Fi.CI.BAT: success for Assorted engine tidy (rev3)
2017-02-16 12:23 [PATCH 0/5] Assorted engine tidy Tvrtko Ursulin
` (4 preceding siblings ...)
2017-02-16 12:23 ` [PATCH 5/5] drm/i915: Consolidate gen8_emit_pipe_control Tvrtko Ursulin
@ 2017-02-16 18:22 ` Patchwork
2017-02-17 8:22 ` ✗ Fi.CI.BAT: failure for Assorted engine tidy (rev4) Patchwork
6 siblings, 0 replies; 21+ messages in thread
From: Patchwork @ 2017-02-16 18:22 UTC (permalink / raw)
To: Tvrtko Ursulin; +Cc: intel-gfx
== Series Details ==
Series: Assorted engine tidy (rev3)
URL : https://patchwork.freedesktop.org/series/19774/
State : success
== Summary ==
Series 19774v3 Assorted engine tidy
https://patchwork.freedesktop.org/api/1.0/series/19774/revisions/3/mbox/
fi-bdw-5557u total:252 pass:241 dwarn:0 dfail:0 fail:0 skip:11
fi-bsw-n3050 total:252 pass:213 dwarn:0 dfail:0 fail:0 skip:39
fi-bxt-j4205 total:252 pass:233 dwarn:0 dfail:0 fail:0 skip:19
fi-bxt-t5700 total:83 pass:70 dwarn:0 dfail:0 fail:0 skip:12
fi-byt-j1900 total:252 pass:225 dwarn:0 dfail:0 fail:0 skip:27
fi-byt-n2820 total:252 pass:221 dwarn:0 dfail:0 fail:0 skip:31
fi-hsw-4770 total:252 pass:236 dwarn:0 dfail:0 fail:0 skip:16
fi-hsw-4770r total:252 pass:236 dwarn:0 dfail:0 fail:0 skip:16
fi-ilk-650 total:252 pass:202 dwarn:0 dfail:0 fail:0 skip:50
fi-ivb-3520m total:252 pass:234 dwarn:0 dfail:0 fail:0 skip:18
fi-ivb-3770 total:252 pass:234 dwarn:0 dfail:0 fail:0 skip:18
fi-kbl-7500u total:252 pass:234 dwarn:0 dfail:0 fail:0 skip:18
fi-skl-6260u total:252 pass:242 dwarn:0 dfail:0 fail:0 skip:10
fi-skl-6700hq total:252 pass:235 dwarn:0 dfail:0 fail:0 skip:17
fi-skl-6700k total:252 pass:230 dwarn:4 dfail:0 fail:0 skip:18
fi-skl-6770hq total:252 pass:242 dwarn:0 dfail:0 fail:0 skip:10
fi-snb-2520m total:252 pass:224 dwarn:0 dfail:0 fail:0 skip:28
fi-snb-2600 total:252 pass:223 dwarn:0 dfail:0 fail:0 skip:29
b29710e319e4b58d994065f5e776f6f66674d7c9 drm-tip: 2017y-02m-16d-15h-35m-40s UTC integration manifest
2e52ebf drm/i915: Consolidate gen8_emit_pipe_control
8ab28ee drm/i915: Tidy workaround batch buffer emission
dd052d5 drm/i915: Move common workaround code to intel_engine_cs
328eff8 drm/i915: Simplify cleanup path in intel_engines_init
a3d7d79 drm/i915: Make int __intel_ring_space static
== Logs ==
For more details see: https://intel-gfx-ci.01.org/CI/Patchwork_3858/
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 21+ messages in thread
* [PATCH v6] drm/i915: Tidy workaround batch buffer emission
2017-02-16 13:54 ` [PATCH v5 " Tvrtko Ursulin
2017-02-16 14:03 ` Chris Wilson
@ 2017-02-17 7:58 ` Tvrtko Ursulin
2017-02-17 8:16 ` Chris Wilson
1 sibling, 1 reply; 21+ messages in thread
From: Tvrtko Ursulin @ 2017-02-17 7:58 UTC (permalink / raw)
To: Intel-gfx
From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Use the "*batch++ = " style as in the ring emission for better
readability and also simplify the logic a bit by consolidating
the offset and size calculations and overflow checking. The
latter is a programming error so it is not required to check
for it after each write to the object, but rather do it once the
whole state has been written and fail the driver if something
went wrong.
v2: Rebase.
v3: Keep track of offsets and sizes in bytes for simplicity
and rename function pointer variable to _fn suffix.
(Chris Wilson)
v4: Fix size calc broken in v3 and add alignment warning. (Chris Wilson)
v5: Fix return code.
v6: I added an exit from loop in v5 but forgot to put back
the object teardown.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> (v5)
Cc: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/intel_lrc.c | 318 ++++++++++++++-------------------------
1 file changed, 117 insertions(+), 201 deletions(-)
diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c
index ee431d39ce06..5154661a38cf 100644
--- a/drivers/gpu/drm/i915/intel_lrc.c
+++ b/drivers/gpu/drm/i915/intel_lrc.c
@@ -890,18 +890,6 @@ static int execlists_request_alloc(struct drm_i915_gem_request *request)
return ret;
}
-#define wa_ctx_emit(batch, index, cmd) \
- do { \
- int __index = (index)++; \
- if (WARN_ON(__index >= (PAGE_SIZE / sizeof(uint32_t)))) { \
- return -ENOSPC; \
- } \
- batch[__index] = (cmd); \
- } while (0)
-
-#define wa_ctx_emit_reg(batch, index, reg) \
- wa_ctx_emit((batch), (index), i915_mmio_reg_offset(reg))
-
/*
* In this WA we need to set GEN8_L3SQCREG4[21:21] and reset it after
* PIPE_CONTROL instruction. This is required for the flush to happen correctly
@@ -918,56 +906,31 @@ static int execlists_request_alloc(struct drm_i915_gem_request *request)
* This WA is also required for Gen9 so extracting as a function avoids
* code duplication.
*/
-static inline int gen8_emit_flush_coherentl3_wa(struct intel_engine_cs *engine,
- uint32_t *batch,
- uint32_t index)
-{
- uint32_t l3sqc4_flush = (0x40400000 | GEN8_LQSC_FLUSH_COHERENT_LINES);
-
- wa_ctx_emit(batch, index, (MI_STORE_REGISTER_MEM_GEN8 |
- MI_SRM_LRM_GLOBAL_GTT));
- wa_ctx_emit_reg(batch, index, GEN8_L3SQCREG4);
- wa_ctx_emit(batch, index, i915_ggtt_offset(engine->scratch) + 256);
- wa_ctx_emit(batch, index, 0);
-
- wa_ctx_emit(batch, index, MI_LOAD_REGISTER_IMM(1));
- wa_ctx_emit_reg(batch, index, GEN8_L3SQCREG4);
- wa_ctx_emit(batch, index, l3sqc4_flush);
-
- wa_ctx_emit(batch, index, GFX_OP_PIPE_CONTROL(6));
- wa_ctx_emit(batch, index, (PIPE_CONTROL_CS_STALL |
- PIPE_CONTROL_DC_FLUSH_ENABLE));
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
-
- wa_ctx_emit(batch, index, (MI_LOAD_REGISTER_MEM_GEN8 |
- MI_SRM_LRM_GLOBAL_GTT));
- wa_ctx_emit_reg(batch, index, GEN8_L3SQCREG4);
- wa_ctx_emit(batch, index, i915_ggtt_offset(engine->scratch) + 256);
- wa_ctx_emit(batch, index, 0);
-
- return index;
-}
-
-static inline uint32_t wa_ctx_start(struct i915_wa_ctx_bb *wa_ctx,
- uint32_t offset,
- uint32_t start_alignment)
+static u32 *
+gen8_emit_flush_coherentl3_wa(struct intel_engine_cs *engine, u32 *batch)
{
- return wa_ctx->offset = ALIGN(offset, start_alignment);
-}
-
-static inline int wa_ctx_end(struct i915_wa_ctx_bb *wa_ctx,
- uint32_t offset,
- uint32_t size_alignment)
-{
- wa_ctx->size = offset - wa_ctx->offset;
-
- WARN(wa_ctx->size % size_alignment,
- "wa_ctx_bb failed sanity checks: size %d is not aligned to %d\n",
- wa_ctx->size, size_alignment);
- return 0;
+ *batch++ = MI_STORE_REGISTER_MEM_GEN8 | MI_SRM_LRM_GLOBAL_GTT;
+ *batch++ = i915_mmio_reg_offset(GEN8_L3SQCREG4);
+ *batch++ = i915_ggtt_offset(engine->scratch) + 256;
+ *batch++ = 0;
+
+ *batch++ = MI_LOAD_REGISTER_IMM(1);
+ *batch++ = i915_mmio_reg_offset(GEN8_L3SQCREG4);
+ *batch++ = 0x40400000 | GEN8_LQSC_FLUSH_COHERENT_LINES;
+
+ *batch++ = GFX_OP_PIPE_CONTROL(6);
+ *batch++ = PIPE_CONTROL_CS_STALL | PIPE_CONTROL_DC_FLUSH_ENABLE;
+ *batch++ = 0;
+ *batch++ = 0;
+ *batch++ = 0;
+ *batch++ = 0;
+
+ *batch++ = MI_LOAD_REGISTER_MEM_GEN8 | MI_SRM_LRM_GLOBAL_GTT;
+ *batch++ = i915_mmio_reg_offset(GEN8_L3SQCREG4);
+ *batch++ = i915_ggtt_offset(engine->scratch) + 256;
+ *batch++ = 0;
+
+ return batch;
}
/*
@@ -985,42 +948,28 @@ static inline int wa_ctx_end(struct i915_wa_ctx_bb *wa_ctx,
* MI_BATCH_BUFFER_END will be added to perctx batch and both of them together
* makes a complete batch buffer.
*/
-static int gen8_init_indirectctx_bb(struct intel_engine_cs *engine,
- struct i915_wa_ctx_bb *wa_ctx,
- uint32_t *batch,
- uint32_t *offset)
+static u32 *gen8_init_indirectctx_bb(struct intel_engine_cs *engine, u32 *batch)
{
- uint32_t scratch_addr;
- uint32_t index = wa_ctx_start(wa_ctx, *offset, CACHELINE_DWORDS);
-
/* WaDisableCtxRestoreArbitration:bdw,chv */
- wa_ctx_emit(batch, index, MI_ARB_ON_OFF | MI_ARB_DISABLE);
+ *batch++ = MI_ARB_ON_OFF | MI_ARB_DISABLE;
/* WaFlushCoherentL3CacheLinesAtContextSwitch:bdw */
- if (IS_BROADWELL(engine->i915)) {
- int rc = gen8_emit_flush_coherentl3_wa(engine, batch, index);
- if (rc < 0)
- return rc;
- index = rc;
- }
+ if (IS_BROADWELL(engine->i915))
+ batch = gen8_emit_flush_coherentl3_wa(engine, batch);
+ *batch++ = GFX_OP_PIPE_CONTROL(6);
+ *batch++ = PIPE_CONTROL_FLUSH_L3 | PIPE_CONTROL_GLOBAL_GTT_IVB |
+ PIPE_CONTROL_CS_STALL | PIPE_CONTROL_QW_WRITE;
/* WaClearSlmSpaceAtContextSwitch:bdw,chv */
/* Actual scratch location is at 128 bytes offset */
- scratch_addr = i915_ggtt_offset(engine->scratch) + 2 * CACHELINE_BYTES;
-
- wa_ctx_emit(batch, index, GFX_OP_PIPE_CONTROL(6));
- wa_ctx_emit(batch, index, (PIPE_CONTROL_FLUSH_L3 |
- PIPE_CONTROL_GLOBAL_GTT_IVB |
- PIPE_CONTROL_CS_STALL |
- PIPE_CONTROL_QW_WRITE));
- wa_ctx_emit(batch, index, scratch_addr);
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
+ *batch++ = i915_ggtt_offset(engine->scratch) + 2 * CACHELINE_BYTES;
+ *batch++ = 0;
+ *batch++ = 0;
+ *batch++ = 0;
/* Pad to end of cacheline */
- while (index % CACHELINE_DWORDS)
- wa_ctx_emit(batch, index, MI_NOOP);
+ while ((unsigned long)batch % CACHELINE_BYTES)
+ *batch++ = MI_NOOP;
/*
* MI_BATCH_BUFFER_END is not required in Indirect ctx BB because
@@ -1028,7 +977,7 @@ static int gen8_init_indirectctx_bb(struct intel_engine_cs *engine,
* in the register CTX_RCS_INDIRECT_CTX
*/
- return wa_ctx_end(wa_ctx, *offset = index, CACHELINE_DWORDS);
+ return batch;
}
/*
@@ -1040,58 +989,38 @@ static int gen8_init_indirectctx_bb(struct intel_engine_cs *engine,
* This batch is terminated with MI_BATCH_BUFFER_END and so we need not add padding
* to align it with cacheline as padding after MI_BATCH_BUFFER_END is redundant.
*/
-static int gen8_init_perctx_bb(struct intel_engine_cs *engine,
- struct i915_wa_ctx_bb *wa_ctx,
- uint32_t *batch,
- uint32_t *offset)
+static u32 *gen8_init_perctx_bb(struct intel_engine_cs *engine, u32 *batch)
{
- uint32_t index = wa_ctx_start(wa_ctx, *offset, CACHELINE_DWORDS);
-
/* WaDisableCtxRestoreArbitration:bdw,chv */
- wa_ctx_emit(batch, index, MI_ARB_ON_OFF | MI_ARB_ENABLE);
-
- wa_ctx_emit(batch, index, MI_BATCH_BUFFER_END);
+ *batch++ = MI_ARB_ON_OFF | MI_ARB_ENABLE;
+ *batch++ = MI_BATCH_BUFFER_END;
- return wa_ctx_end(wa_ctx, *offset = index, 1);
+ return batch;
}
-static int gen9_init_indirectctx_bb(struct intel_engine_cs *engine,
- struct i915_wa_ctx_bb *wa_ctx,
- uint32_t *batch,
- uint32_t *offset)
+static u32 *gen9_init_indirectctx_bb(struct intel_engine_cs *engine, u32 *batch)
{
- int ret;
- struct drm_i915_private *dev_priv = engine->i915;
- uint32_t index = wa_ctx_start(wa_ctx, *offset, CACHELINE_DWORDS);
-
/* WaFlushCoherentL3CacheLinesAtContextSwitch:skl,bxt,glk */
- ret = gen8_emit_flush_coherentl3_wa(engine, batch, index);
- if (ret < 0)
- return ret;
- index = ret;
+ batch = gen8_emit_flush_coherentl3_wa(engine, batch);
/* WaDisableGatherAtSetShaderCommonSlice:skl,bxt,kbl,glk */
- wa_ctx_emit(batch, index, MI_LOAD_REGISTER_IMM(1));
- wa_ctx_emit_reg(batch, index, COMMON_SLICE_CHICKEN2);
- wa_ctx_emit(batch, index, _MASKED_BIT_DISABLE(
- GEN9_DISABLE_GATHER_AT_SET_SHADER_COMMON_SLICE));
- wa_ctx_emit(batch, index, MI_NOOP);
+ *batch++ = MI_LOAD_REGISTER_IMM(1);
+ *batch++ = i915_mmio_reg_offset(COMMON_SLICE_CHICKEN2);
+ *batch++ = _MASKED_BIT_DISABLE(
+ GEN9_DISABLE_GATHER_AT_SET_SHADER_COMMON_SLICE);
+ *batch++ = MI_NOOP;
/* WaClearSlmSpaceAtContextSwitch:kbl */
/* Actual scratch location is at 128 bytes offset */
- if (IS_KBL_REVID(dev_priv, 0, KBL_REVID_A0)) {
- u32 scratch_addr =
- i915_ggtt_offset(engine->scratch) + 2 * CACHELINE_BYTES;
-
- wa_ctx_emit(batch, index, GFX_OP_PIPE_CONTROL(6));
- wa_ctx_emit(batch, index, (PIPE_CONTROL_FLUSH_L3 |
- PIPE_CONTROL_GLOBAL_GTT_IVB |
- PIPE_CONTROL_CS_STALL |
- PIPE_CONTROL_QW_WRITE));
- wa_ctx_emit(batch, index, scratch_addr);
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
+ if (IS_KBL_REVID(engine->i915, 0, KBL_REVID_A0)) {
+ *batch++ = GFX_OP_PIPE_CONTROL(6);
+ *batch++ = PIPE_CONTROL_FLUSH_L3 | PIPE_CONTROL_GLOBAL_GTT_IVB |
+ PIPE_CONTROL_CS_STALL | PIPE_CONTROL_QW_WRITE;
+ *batch++ = i915_ggtt_offset(engine->scratch) +
+ 2 * CACHELINE_BYTES;
+ *batch++ = 0;
+ *batch++ = 0;
+ *batch++ = 0;
}
/* WaMediaPoolStateCmdInWABB:bxt,glk */
@@ -1109,41 +1038,37 @@ static int gen9_init_indirectctx_bb(struct intel_engine_cs *engine,
* possible configurations, to avoid duplication they are
* not shown here again.
*/
- u32 eu_pool_config = 0x00777000;
- wa_ctx_emit(batch, index, GEN9_MEDIA_POOL_STATE);
- wa_ctx_emit(batch, index, GEN9_MEDIA_POOL_ENABLE);
- wa_ctx_emit(batch, index, eu_pool_config);
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
- wa_ctx_emit(batch, index, 0);
+ *batch++ = GEN9_MEDIA_POOL_STATE;
+ *batch++ = GEN9_MEDIA_POOL_ENABLE;
+ *batch++ = 0x00777000;
+ *batch++ = 0;
+ *batch++ = 0;
+ *batch++ = 0;
}
/* Pad to end of cacheline */
- while (index % CACHELINE_DWORDS)
- wa_ctx_emit(batch, index, MI_NOOP);
+ while ((unsigned long)batch % CACHELINE_BYTES)
+ *batch++ = MI_NOOP;
- return wa_ctx_end(wa_ctx, *offset = index, CACHELINE_DWORDS);
+ return batch;
}
-static int gen9_init_perctx_bb(struct intel_engine_cs *engine,
- struct i915_wa_ctx_bb *wa_ctx,
- uint32_t *batch,
- uint32_t *offset)
+static u32 *gen9_init_perctx_bb(struct intel_engine_cs *engine, u32 *batch)
{
- uint32_t index = wa_ctx_start(wa_ctx, *offset, CACHELINE_DWORDS);
-
- wa_ctx_emit(batch, index, MI_BATCH_BUFFER_END);
+ *batch++ = MI_BATCH_BUFFER_END;
- return wa_ctx_end(wa_ctx, *offset = index, 1);
+ return batch;
}
-static int lrc_setup_wa_ctx_obj(struct intel_engine_cs *engine, u32 size)
+#define CTX_WA_BB_OBJ_SIZE (PAGE_SIZE)
+
+static int lrc_setup_wa_ctx(struct intel_engine_cs *engine)
{
struct drm_i915_gem_object *obj;
struct i915_vma *vma;
int err;
- obj = i915_gem_object_create(engine->i915, PAGE_ALIGN(size));
+ obj = i915_gem_object_create(engine->i915, CTX_WA_BB_OBJ_SIZE);
if (IS_ERR(obj))
return PTR_ERR(obj);
@@ -1165,78 +1090,70 @@ static int lrc_setup_wa_ctx_obj(struct intel_engine_cs *engine, u32 size)
return err;
}
-static void lrc_destroy_wa_ctx_obj(struct intel_engine_cs *engine)
+static void lrc_destroy_wa_ctx(struct intel_engine_cs *engine)
{
i915_vma_unpin_and_release(&engine->wa_ctx.vma);
}
+typedef u32 *(*wa_bb_func_t)(struct intel_engine_cs *engine, u32 *batch);
+
static int intel_init_workaround_bb(struct intel_engine_cs *engine)
{
struct i915_ctx_workarounds *wa_ctx = &engine->wa_ctx;
- uint32_t *batch;
- uint32_t offset;
+ struct i915_wa_ctx_bb *wa_bb[2] = { &wa_ctx->indirect_ctx,
+ &wa_ctx->per_ctx };
+ wa_bb_func_t wa_bb_fn[2];
struct page *page;
+ void *batch, *batch_ptr;
+ unsigned int i;
int ret;
- WARN_ON(engine->id != RCS);
+ if (WARN_ON(engine->id != RCS || !engine->scratch))
+ return -EINVAL;
- /* update this when WA for higher Gen are added */
- if (INTEL_GEN(engine->i915) > 9) {
- DRM_ERROR("WA batch buffer is not initialized for Gen%d\n",
- INTEL_GEN(engine->i915));
+ switch (INTEL_GEN(engine->i915)) {
+ case 9:
+ wa_bb_fn[0] = gen9_init_indirectctx_bb;
+ wa_bb_fn[1] = gen9_init_perctx_bb;
+ break;
+ case 8:
+ wa_bb_fn[0] = gen8_init_indirectctx_bb;
+ wa_bb_fn[1] = gen8_init_perctx_bb;
+ break;
+ default:
+ MISSING_CASE(INTEL_GEN(engine->i915));
return 0;
}
- /* some WA perform writes to scratch page, ensure it is valid */
- if (!engine->scratch) {
- DRM_ERROR("scratch page not allocated for %s\n", engine->name);
- return -EINVAL;
- }
-
- ret = lrc_setup_wa_ctx_obj(engine, PAGE_SIZE);
+ ret = lrc_setup_wa_ctx(engine);
if (ret) {
DRM_DEBUG_DRIVER("Failed to setup context WA page: %d\n", ret);
return ret;
}
page = i915_gem_object_get_dirty_page(wa_ctx->vma->obj, 0);
- batch = kmap_atomic(page);
- offset = 0;
-
- if (IS_GEN8(engine->i915)) {
- ret = gen8_init_indirectctx_bb(engine,
- &wa_ctx->indirect_ctx,
- batch,
- &offset);
- if (ret)
- goto out;
+ batch = batch_ptr = kmap_atomic(page);
- ret = gen8_init_perctx_bb(engine,
- &wa_ctx->per_ctx,
- batch,
- &offset);
- if (ret)
- goto out;
- } else if (IS_GEN9(engine->i915)) {
- ret = gen9_init_indirectctx_bb(engine,
- &wa_ctx->indirect_ctx,
- batch,
- &offset);
- if (ret)
- goto out;
-
- ret = gen9_init_perctx_bb(engine,
- &wa_ctx->per_ctx,
- batch,
- &offset);
- if (ret)
- goto out;
+ /*
+ * Emit the two workaround batch buffers, recording the offset from the
+ * start of the workaround batch buffer object for each and their
+ * respective sizes.
+ */
+ for (i = 0; i < ARRAY_SIZE(wa_bb_fn); i++) {
+ wa_bb[i]->offset = batch_ptr - batch;
+ if (WARN_ON(!IS_ALIGNED(wa_bb[i]->offset, CACHELINE_BYTES))) {
+ ret = -EINVAL;
+ break;
+ }
+ batch_ptr = wa_bb_fn[i](engine, batch_ptr);
+ wa_bb[i]->size = batch_ptr - (batch + wa_bb[i]->offset);
}
-out:
+ BUG_ON(batch_ptr - batch > CTX_WA_BB_OBJ_SIZE);
+
kunmap_atomic(batch);
if (ret)
- lrc_destroy_wa_ctx_obj(engine);
+ lrc_destroy_wa_ctx(engine);
return ret;
}
@@ -1685,7 +1602,7 @@ void intel_logical_ring_cleanup(struct intel_engine_cs *engine)
intel_engine_cleanup_common(engine);
- lrc_destroy_wa_ctx_obj(engine);
+ lrc_destroy_wa_ctx(engine);
engine->i915 = NULL;
dev_priv->engine[engine->id] = NULL;
kfree(engine);
@@ -1971,15 +1888,14 @@ static void execlists_init_reg_state(u32 *reg_state,
u32 ggtt_offset = i915_ggtt_offset(wa_ctx->vma);
reg_state[CTX_RCS_INDIRECT_CTX+1] =
- (ggtt_offset + wa_ctx->indirect_ctx.offset * sizeof(uint32_t)) |
- (wa_ctx->indirect_ctx.size / CACHELINE_DWORDS);
+ (ggtt_offset + wa_ctx->indirect_ctx.offset) |
+ (wa_ctx->indirect_ctx.size / CACHELINE_BYTES);
reg_state[CTX_RCS_INDIRECT_CTX_OFFSET+1] =
intel_lr_indirect_ctx_offset(engine) << 6;
reg_state[CTX_BB_PER_CTX_PTR+1] =
- (ggtt_offset + wa_ctx->per_ctx.offset * sizeof(uint32_t)) |
- 0x01;
+ (ggtt_offset + wa_ctx->per_ctx.offset) | 0x01;
}
}
reg_state[CTX_LRI_HEADER_1] = MI_LOAD_REGISTER_IMM(9) | MI_LRI_FORCE_POSTED;
--
2.9.3
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 21+ messages in thread
* Re: [PATCH v6] drm/i915: Tidy workaround batch buffer emission
2017-02-17 7:58 ` [PATCH v6] " Tvrtko Ursulin
@ 2017-02-17 8:16 ` Chris Wilson
0 siblings, 0 replies; 21+ messages in thread
From: Chris Wilson @ 2017-02-17 8:16 UTC (permalink / raw)
To: Tvrtko Ursulin; +Cc: Intel-gfx
On Fri, Feb 17, 2017 at 07:58:59AM +0000, Tvrtko Ursulin wrote:
> From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
>
> Use the "*batch++ = " style as in the ring emission for better
> readability and also simplify the logic a bit by consolidating
> the offset and size calculations and overflow checking. The
> latter is a programming error so it is not required to check
> for it after each write to the object, but rather do it once the
> whole state has been written and fail the driver if something
> went wrong.
>
> v2: Rebase.
>
> v3: Keep track of offsets and sizes in bytes for simplicity
> and rename function pointer variable to _fn suffix.
> (Chris Wilson)
>
> v4: Fix size calc broken in v3 and add alignment warning. (Chris Wilson)
>
> v5: Fix return code.
>
> v6: I added an exit from loop in v5 but forgot to put back
> the object teardown.
>
> Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> (v5)
> Cc: Chris Wilson <chris@chris-wilson.co.uk>
> + /*
> + * Emit the two workaround batch buffers, recording the offset from the
> + * start of the workaround batch buffer object for each and their
> + * respective sizes.
> + */
> + for (i = 0; i < ARRAY_SIZE(wa_bb_fn); i++) {
> + wa_bb[i]->offset = batch_ptr - batch;
> + if (WARN_ON(!IS_ALIGNED(wa_bb[i]->offset, CACHELINE_BYTES))) {
> + ret = -EINVAL;
> + break;
> + }
> + batch_ptr = wa_bb_fn[i](engine, batch_ptr);
> + wa_bb[i]->size = batch_ptr - (batch + wa_bb[i]->offset);
> }
>
> -out:
> + BUG_ON(batch_ptr - batch > CTX_WA_BB_OBJ_SIZE);
Ok, I didn't spot it last time, so hopefully not missing anything this
time!
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
-Chris
--
Chris Wilson, Intel Open Source Technology Centre
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 21+ messages in thread
* ✗ Fi.CI.BAT: failure for Assorted engine tidy (rev4)
2017-02-16 12:23 [PATCH 0/5] Assorted engine tidy Tvrtko Ursulin
` (5 preceding siblings ...)
2017-02-16 18:22 ` ✓ Fi.CI.BAT: success for Assorted engine tidy (rev3) Patchwork
@ 2017-02-17 8:22 ` Patchwork
2017-02-17 11:50 ` Tvrtko Ursulin
6 siblings, 1 reply; 21+ messages in thread
From: Patchwork @ 2017-02-17 8:22 UTC (permalink / raw)
To: Tvrtko Ursulin; +Cc: intel-gfx
== Series Details ==
Series: Assorted engine tidy (rev4)
URL : https://patchwork.freedesktop.org/series/19774/
State : failure
== Summary ==
Series 19774v4 Assorted engine tidy
https://patchwork.freedesktop.org/api/1.0/series/19774/revisions/4/mbox/
Test kms_pipe_crc_basic:
Subgroup suspend-read-crc-pipe-a:
pass -> FAIL (fi-ivb-3520m)
Subgroup suspend-read-crc-pipe-b:
pass -> DMESG-FAIL (fi-ivb-3520m)
Subgroup suspend-read-crc-pipe-c:
pass -> DMESG-FAIL (fi-ivb-3520m)
fi-bdw-5557u total:252 pass:241 dwarn:0 dfail:0 fail:0 skip:11
fi-bsw-n3050 total:252 pass:213 dwarn:0 dfail:0 fail:0 skip:39
fi-bxt-j4205 total:252 pass:233 dwarn:0 dfail:0 fail:0 skip:19
fi-bxt-t5700 total:83 pass:70 dwarn:0 dfail:0 fail:0 skip:12
fi-byt-j1900 total:252 pass:225 dwarn:0 dfail:0 fail:0 skip:27
fi-byt-n2820 total:252 pass:221 dwarn:0 dfail:0 fail:0 skip:31
fi-hsw-4770 total:252 pass:236 dwarn:0 dfail:0 fail:0 skip:16
fi-hsw-4770r total:252 pass:236 dwarn:0 dfail:0 fail:0 skip:16
fi-ilk-650 total:252 pass:202 dwarn:0 dfail:0 fail:0 skip:50
fi-ivb-3520m total:252 pass:231 dwarn:0 dfail:2 fail:1 skip:18
fi-ivb-3770 total:252 pass:234 dwarn:0 dfail:0 fail:0 skip:18
fi-kbl-7500u total:252 pass:234 dwarn:0 dfail:0 fail:0 skip:18
fi-skl-6260u total:252 pass:242 dwarn:0 dfail:0 fail:0 skip:10
fi-skl-6700hq total:252 pass:235 dwarn:0 dfail:0 fail:0 skip:17
fi-skl-6700k total:252 pass:230 dwarn:4 dfail:0 fail:0 skip:18
fi-skl-6770hq total:252 pass:242 dwarn:0 dfail:0 fail:0 skip:10
fi-snb-2520m total:252 pass:224 dwarn:0 dfail:0 fail:0 skip:28
fi-snb-2600 total:252 pass:223 dwarn:0 dfail:0 fail:0 skip:29
c2033e7aa383d062000e024c5fac5f46560327cd drm-tip: 2017y-02m-16d-20h-42m-04s UTC integration manifest
4f91a4f drm/i915: Consolidate gen8_emit_pipe_control
2fc291c drm/i915: Tidy workaround batch buffer emission
401470e drm/i915: Move common workaround code to intel_engine_cs
fc07f32 drm/i915: Simplify cleanup path in intel_engines_init
c7fb795 drm/i915: Make int __intel_ring_space static
== Logs ==
For more details see: https://intel-gfx-ci.01.org/CI/Patchwork_3870/
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 21+ messages in thread
* Re: ✗ Fi.CI.BAT: failure for Assorted engine tidy (rev4)
2017-02-17 8:22 ` ✗ Fi.CI.BAT: failure for Assorted engine tidy (rev4) Patchwork
@ 2017-02-17 11:50 ` Tvrtko Ursulin
0 siblings, 0 replies; 21+ messages in thread
From: Tvrtko Ursulin @ 2017-02-17 11:50 UTC (permalink / raw)
To: intel-gfx, Tvrtko Ursulin
On 17/02/2017 08:22, Patchwork wrote:
> == Series Details ==
>
> Series: Assorted engine tidy (rev4)
> URL : https://patchwork.freedesktop.org/series/19774/
> State : failure
>
> == Summary ==
>
> Series 19774v4 Assorted engine tidy
> https://patchwork.freedesktop.org/api/1.0/series/19774/revisions/4/mbox/
>
> Test kms_pipe_crc_basic:
> Subgroup suspend-read-crc-pipe-a:
> pass -> FAIL (fi-ivb-3520m)
> Subgroup suspend-read-crc-pipe-b:
> pass -> DMESG-FAIL (fi-ivb-3520m)
> Subgroup suspend-read-crc-pipe-c:
> pass -> DMESG-FAIL (fi-ivb-3520m)
https://bugs.freedesktop.org/show_bug.cgi?id=9984
e1000 driver is failing to suspend, with or without a WARN like this one:
[ 432.585013] WARNING: CPU: 3 PID: 8372 at kernel/irq/manage.c:1478
__free_irq+0x9f/0x280
[ 432.585015] Trying to free already-free IRQ 207
>
> fi-bdw-5557u total:252 pass:241 dwarn:0 dfail:0 fail:0 skip:11
> fi-bsw-n3050 total:252 pass:213 dwarn:0 dfail:0 fail:0 skip:39
> fi-bxt-j4205 total:252 pass:233 dwarn:0 dfail:0 fail:0 skip:19
> fi-bxt-t5700 total:83 pass:70 dwarn:0 dfail:0 fail:0 skip:12
> fi-byt-j1900 total:252 pass:225 dwarn:0 dfail:0 fail:0 skip:27
> fi-byt-n2820 total:252 pass:221 dwarn:0 dfail:0 fail:0 skip:31
> fi-hsw-4770 total:252 pass:236 dwarn:0 dfail:0 fail:0 skip:16
> fi-hsw-4770r total:252 pass:236 dwarn:0 dfail:0 fail:0 skip:16
> fi-ilk-650 total:252 pass:202 dwarn:0 dfail:0 fail:0 skip:50
> fi-ivb-3520m total:252 pass:231 dwarn:0 dfail:2 fail:1 skip:18
> fi-ivb-3770 total:252 pass:234 dwarn:0 dfail:0 fail:0 skip:18
> fi-kbl-7500u total:252 pass:234 dwarn:0 dfail:0 fail:0 skip:18
> fi-skl-6260u total:252 pass:242 dwarn:0 dfail:0 fail:0 skip:10
> fi-skl-6700hq total:252 pass:235 dwarn:0 dfail:0 fail:0 skip:17
> fi-skl-6700k total:252 pass:230 dwarn:4 dfail:0 fail:0 skip:18
> fi-skl-6770hq total:252 pass:242 dwarn:0 dfail:0 fail:0 skip:10
> fi-snb-2520m total:252 pass:224 dwarn:0 dfail:0 fail:0 skip:28
> fi-snb-2600 total:252 pass:223 dwarn:0 dfail:0 fail:0 skip:29
>
> c2033e7aa383d062000e024c5fac5f46560327cd drm-tip: 2017y-02m-16d-20h-42m-04s UTC integration manifest
> 4f91a4f drm/i915: Consolidate gen8_emit_pipe_control
> 2fc291c drm/i915: Tidy workaround batch buffer emission
> 401470e drm/i915: Move common workaround code to intel_engine_cs
> fc07f32 drm/i915: Simplify cleanup path in intel_engines_init
> c7fb795 drm/i915: Make int __intel_ring_space static
Pushed, thanks for the review.
Regards,
Tvrtko
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 21+ messages in thread
end of thread, other threads:[~2017-02-17 11:50 UTC | newest]
Thread overview: 21+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-02-16 12:23 [PATCH 0/5] Assorted engine tidy Tvrtko Ursulin
2017-02-16 12:23 ` [PATCH 1/5] drm/i915: Make int __intel_ring_space static Tvrtko Ursulin
2017-02-16 13:12 ` Chris Wilson
2017-02-16 12:23 ` [PATCH 2/5] drm/i915: Simplify cleanup path in intel_engines_init Tvrtko Ursulin
2017-02-16 13:15 ` Chris Wilson
2017-02-16 12:23 ` [PATCH 3/5] drm/i915: Move common workaround code to intel_engine_cs Tvrtko Ursulin
2017-02-16 13:18 ` Chris Wilson
2017-02-16 12:23 ` [PATCH 4/5] drm/i915: Tidy workaround batch buffer emission Tvrtko Ursulin
2017-02-16 13:20 ` Chris Wilson
2017-02-16 13:23 ` Tvrtko Ursulin
2017-02-16 13:28 ` Chris Wilson
2017-02-16 13:47 ` Tvrtko Ursulin
2017-02-16 13:51 ` [PATCH v4 1/2] " Tvrtko Ursulin
2017-02-16 13:54 ` [PATCH v5 " Tvrtko Ursulin
2017-02-16 14:03 ` Chris Wilson
2017-02-17 7:58 ` [PATCH v6] " Tvrtko Ursulin
2017-02-17 8:16 ` Chris Wilson
2017-02-16 12:23 ` [PATCH 5/5] drm/i915: Consolidate gen8_emit_pipe_control Tvrtko Ursulin
2017-02-16 18:22 ` ✓ Fi.CI.BAT: success for Assorted engine tidy (rev3) Patchwork
2017-02-17 8:22 ` ✗ Fi.CI.BAT: failure for Assorted engine tidy (rev4) Patchwork
2017-02-17 11:50 ` Tvrtko Ursulin
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.