drm/xe: Allow per queue programming of COMMON_SLICE_CHICKEN3 bit13

Similar to i915's commit cebc13de7e704b1355bea208a9f9cdb042c74588
("drm/i915: Whitelist COMMON_SLICE_CHICKEN3 for UMD access"), except
that instead of putting the register on the allowlist for UMD to
program, the KMD is doing the programming at context initialization
based on a queue creation flag.

This is a recommended tuning setting for both gen12 and Xe_HP
platforms.

If a render queue is created with
DRM_XE_EXEC_QUEUE_SET_STATE_CACHE_PERF_FIX, COMMON_SLICE_CHICKEN3 will
be programmed at initialization to enable the render color cache to
key with BTP+BTI (binding table pool + binding table entry) instead of
just BTI (binding table entry). This enables the UMD to avoid emitting
render-target-cache-flush + stall-at-pixel-scoreboard every time a
binding table entry pointing to a render target is changed.

v2: Use xe_lrc_write_ring()

v3: Update xe_query.c to report availability

v4: Rename defines to add DISABLE_

v5: update commit message

v6: rebase

Mesa MR: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39982

Bspec: 73993, 73994, 72161, 31870, 68331
Acked-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: José Roberto de Souza <jose.souza@intel.com>
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Link: https://patch.msgid.link/20260306075504.1288676-1-lionel.g.landwerlin@intel.com

6 files changed, 33 insertions, 1 deletions

diff --git a/drivers/gpu/drm/xe/regs/xe_gt_regs.h b/drivers/gpu/drm/xe/regs/xe_gt_regs.h
index 8e6df9dcd137..4ee88f629c02 100644
--- a/drivers/gpu/drm/xe/regs/xe_gt_regs.h
+++ b/drivers/gpu/drm/xe/regs/xe_gt_regs.h
@@ -183,6 +183,7 @@
 
 #define COMMON_SLICE_CHICKEN3				XE_REG(0x7304, XE_REG_OPTION_MASKED)
 #define XEHP_COMMON_SLICE_CHICKEN3			XE_REG_MCR(0x7304, XE_REG_OPTION_MASKED)
+#define   DISABLE_STATE_CACHE_PERF_FIX			REG_BIT(13)
 #define   DG1_FLOAT_POINT_BLEND_OPT_STRICT_MODE_EN	REG_BIT(12)
 #define   XEHP_DUAL_SIMD8_SEQ_MERGE_DISABLE		REG_BIT(12)
 #define   BLEND_EMB_FIX_DISABLE_IN_RCC			REG_BIT(11)
diff --git a/drivers/gpu/drm/xe/xe_exec_queue.c b/drivers/gpu/drm/xe/xe_exec_queue.c
index 6166b1a81433..a11021d36f87 100644
--- a/drivers/gpu/drm/xe/xe_exec_queue.c
+++ b/drivers/gpu/drm/xe/xe_exec_queue.c
@@ -353,6 +353,9 @@ static int __xe_exec_queue_init(struct xe_exec_queue *q, u32 exec_queue_flags)
 	if (!(exec_queue_flags & EXEC_QUEUE_FLAG_KERNEL))
 		flags |= XE_LRC_CREATE_USER_CTX;
 
+	if (q->flags & EXEC_QUEUE_FLAG_DISABLE_STATE_CACHE_PERF_FIX)
+		flags |= XE_LRC_DISABLE_STATE_CACHE_PERF_FIX;
+
 	err = q->ops->init(q);
 	if (err)
 		return err;
@@ -978,6 +981,17 @@ static int exec_queue_set_multi_queue_priority(struct xe_device *xe, struct xe_e
 	return q->ops->set_multi_queue_priority(q, value);
 }
 
+static int exec_queue_set_state_cache_perf_fix(struct xe_device *xe, struct xe_exec_queue *q,
+					       u64 value)
+{
+	if (XE_IOCTL_DBG(xe, q->class != XE_ENGINE_CLASS_RENDER))
+		return -EOPNOTSUPP;
+
+	q->flags |= value != 0 ? EXEC_QUEUE_FLAG_DISABLE_STATE_CACHE_PERF_FIX : 0;
+
+	return 0;
+}
+
 typedef int (*xe_exec_queue_set_property_fn)(struct xe_device *xe,
 					     struct xe_exec_queue *q,
 					     u64 value);
@@ -990,6 +1004,8 @@ static const xe_exec_queue_set_property_fn exec_queue_set_property_funcs[] = {
 	[DRM_XE_EXEC_QUEUE_SET_PROPERTY_MULTI_GROUP] = exec_queue_set_multi_group,
 	[DRM_XE_EXEC_QUEUE_SET_PROPERTY_MULTI_QUEUE_PRIORITY] =
 							exec_queue_set_multi_queue_priority,
+	[DRM_XE_EXEC_QUEUE_SET_DISABLE_STATE_CACHE_PERF_FIX] =
+							exec_queue_set_state_cache_perf_fix,
 };
 
 /**
@@ -1085,7 +1101,8 @@ static int exec_queue_user_ext_set_property(struct xe_device *xe,
 			 ext.property != DRM_XE_EXEC_QUEUE_SET_PROPERTY_PXP_TYPE &&
 			 ext.property != DRM_XE_EXEC_QUEUE_SET_HANG_REPLAY_STATE &&
 			 ext.property != DRM_XE_EXEC_QUEUE_SET_PROPERTY_MULTI_GROUP &&
-			 ext.property != DRM_XE_EXEC_QUEUE_SET_PROPERTY_MULTI_QUEUE_PRIORITY))
+			 ext.property != DRM_XE_EXEC_QUEUE_SET_PROPERTY_MULTI_QUEUE_PRIORITY &&
+			 ext.property != DRM_XE_EXEC_QUEUE_SET_DISABLE_STATE_CACHE_PERF_FIX))
 		return -EINVAL;
 
 	idx = array_index_nospec(ext.property, ARRAY_SIZE(exec_queue_set_property_funcs));
diff --git a/drivers/gpu/drm/xe/xe_exec_queue_types.h b/drivers/gpu/drm/xe/xe_exec_queue_types.h
index a1f3938f4173..8ce78e0b1d50 100644
--- a/drivers/gpu/drm/xe/xe_exec_queue_types.h
+++ b/drivers/gpu/drm/xe/xe_exec_queue_types.h
@@ -134,6 +134,8 @@ struct xe_exec_queue {
 #define EXEC_QUEUE_FLAG_LOW_LATENCY		BIT(5)
 /* for migration (kernel copy, clear, bind) jobs */
 #define EXEC_QUEUE_FLAG_MIGRATE			BIT(6)
+/* for programming COMMON_SLICE_CHICKEN3 on first submission */
+#define EXEC_QUEUE_FLAG_DISABLE_STATE_CACHE_PERF_FIX	BIT(7)
 
 	/**
 	 * @flags: flags for this exec queue, should statically setup aside from ban
diff --git a/drivers/gpu/drm/xe/xe_lrc.c b/drivers/gpu/drm/xe/xe_lrc.c
index ebab5d78f7cc..a1f856eff4ee 100644
--- a/drivers/gpu/drm/xe/xe_lrc.c
+++ b/drivers/gpu/drm/xe/xe_lrc.c
@@ -14,6 +14,7 @@
 #include "instructions/xe_gfxpipe_commands.h"
 #include "instructions/xe_gfx_state_commands.h"
 #include "regs/xe_engine_regs.h"
+#include "regs/xe_gt_regs.h"
 #include "regs/xe_lrc_layout.h"
 #include "xe_bb.h"
 #include "xe_bo.h"
@@ -1446,6 +1447,7 @@ static int xe_lrc_ctx_init(struct xe_lrc *lrc, struct xe_hw_engine *hwe, struct
 	struct xe_device *xe = gt_to_xe(gt);
 	struct iosys_map map;
 	u32 arb_enable;
+	u32 state_cache_perf_fix[3];
 	int err;
 
 	/*
@@ -1546,6 +1548,13 @@ static int xe_lrc_ctx_init(struct xe_lrc *lrc, struct xe_hw_engine *hwe, struct
 	arb_enable = MI_ARB_ON_OFF | MI_ARB_ENABLE;
 	xe_lrc_write_ring(lrc, &arb_enable, sizeof(arb_enable));
 
+	if (init_flags & XE_LRC_DISABLE_STATE_CACHE_PERF_FIX) {
+		state_cache_perf_fix[0] = MI_LOAD_REGISTER_IMM | MI_LRI_NUM_REGS(1);
+		state_cache_perf_fix[1] = COMMON_SLICE_CHICKEN3.addr;
+		state_cache_perf_fix[2] = _MASKED_BIT_ENABLE(DISABLE_STATE_CACHE_PERF_FIX);
+		xe_lrc_write_ring(lrc, state_cache_perf_fix, sizeof(state_cache_perf_fix));
+	}
+
 	map = __xe_lrc_seqno_map(lrc);
 	xe_map_write32(lrc_to_xe(lrc), &map, lrc->fence_ctx.next_seqno - 1);
 
diff --git a/drivers/gpu/drm/xe/xe_lrc.h b/drivers/gpu/drm/xe/xe_lrc.h
index 48f7c26cf129..e7c975f9e2d9 100644
--- a/drivers/gpu/drm/xe/xe_lrc.h
+++ b/drivers/gpu/drm/xe/xe_lrc.h
@@ -49,6 +49,7 @@ struct xe_lrc_snapshot {
 #define XE_LRC_CREATE_RUNALONE		BIT(0)
 #define XE_LRC_CREATE_PXP		BIT(1)
 #define XE_LRC_CREATE_USER_CTX		BIT(2)
+#define XE_LRC_DISABLE_STATE_CACHE_PERF_FIX	BIT(3)
 
 struct xe_lrc *xe_lrc_create(struct xe_hw_engine *hwe, struct xe_vm *vm,
 			     void *replay_state, u32 ring_size, u16 msix_vec, u32 flags);
diff --git a/drivers/gpu/drm/xe/xe_query.c b/drivers/gpu/drm/xe/xe_query.c
index 34db266b723f..4852fdcb4b95 100644
--- a/drivers/gpu/drm/xe/xe_query.c
+++ b/drivers/gpu/drm/xe/xe_query.c
@@ -340,6 +340,8 @@ static int query_config(struct xe_device *xe, struct drm_xe_device_query *query)
 			DRM_XE_QUERY_CONFIG_FLAG_HAS_NO_COMPRESSION_HINT;
 	config->info[DRM_XE_QUERY_CONFIG_FLAGS] |=
 			DRM_XE_QUERY_CONFIG_FLAG_HAS_LOW_LATENCY;
+	config->info[DRM_XE_QUERY_CONFIG_FLAGS] |=
+		DRM_XE_QUERY_CONFIG_FLAG_HAS_DISABLE_STATE_CACHE_PERF_FIX;
 	config->info[DRM_XE_QUERY_CONFIG_MIN_ALIGNMENT] =
 		xe->info.vram_flags & XE_VRAM_FLAGS_NEED64K ? SZ_64K : SZ_4K;
 	config->info[DRM_XE_QUERY_CONFIG_VA_BITS] = xe->info.va_bits;