drm/xe: Add Indirect Ring State support

When Indirect Ring State is enabled, the Ring Buffer state and
Batch Buffer state are context save/restored to/from Indirect
Ring State instead of the LRC. The Indirect Ring State is a 4K
page mapped in global GTT at a 4K aligned address. This address
is programmed in the INDIRECT_RING_STATE register of the
corresponding context's LRC.

v2: Fix kernel-doc, add bspec reference
v3: Fix typo in commit text

Bspec: 67296, 67139
Signed-off-by: Niranjana Vishwanathapura <niranjana.vishwanathapura@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240507224255.5059-3-niranjana.vishwanathapura@intel.com

drivers/gpu/drm/xe/regs/xe_engine_regs.h

@@ -125,6 +125,7 @@
 #define RING_EXECLIST_STATUS_HI(base)		XE_REG((base) + 0x234 + 4)
 
 #define RING_CONTEXT_CONTROL(base)		XE_REG((base) + 0x244, XE_REG_OPTION_MASKED)
+#define	  CTX_CTRL_INDIRECT_RING_STATE_ENABLE	REG_BIT(4)
 #define	  CTX_CTRL_INHIBIT_SYN_CTX_SWITCH	REG_BIT(3)
 #define	  CTX_CTRL_ENGINE_CTX_RESTORE_INHIBIT	REG_BIT(0)
 

drivers/gpu/drm/xe/regs/xe_lrc_layout.h

@@ -11,6 +11,7 @@
 #define CTX_RING_TAIL			(0x06 + 1)
 #define CTX_RING_START			(0x08 + 1)
 #define CTX_RING_CTL			(0x0a + 1)
+#define CTX_INDIRECT_RING_STATE		(0x26 + 1)
 #define CTX_PDP0_UDW			(0x30 + 1)
 #define CTX_PDP0_LDW			(0x32 + 1)
 
@@ -23,4 +24,10 @@
 #define CTX_INT_SRC_REPORT_REG		(CTX_LRI_INT_REPORT_PTR + 3)
 #define CTX_INT_SRC_REPORT_PTR		(CTX_LRI_INT_REPORT_PTR + 4)
 
+#define INDIRECT_CTX_RING_HEAD		(0x02 + 1)
+#define INDIRECT_CTX_RING_TAIL		(0x04 + 1)
+#define INDIRECT_CTX_RING_START		(0x06 + 1)
+#define INDIRECT_CTX_RING_START_UDW	(0x08 + 1)
+#define INDIRECT_CTX_RING_CTL		(0x0a + 1)
+
 #endif

drivers/gpu/drm/xe/xe_gt.c

@@ -160,7 +160,7 @@ static int emit_wa_job(struct xe_gt *gt, struct xe_exec_queue *q)
 
 	if (q->hwe->class == XE_ENGINE_CLASS_RENDER)
 		/* Big enough to emit all of the context's 3DSTATE */
-		bb = xe_bb_new(gt, xe_lrc_size(gt_to_xe(gt), q->hwe->class), false);
+		bb = xe_bb_new(gt, xe_gt_lrc_size(gt, q->hwe->class), false);
 	else
 		/* Just pick a large BB size */
 		bb = xe_bb_new(gt, SZ_4K, false);
@@ -244,7 +244,7 @@ int xe_gt_record_default_lrcs(struct xe_gt *gt)
 		xe_tuning_process_lrc(hwe);
 
 		default_lrc = drmm_kzalloc(&xe->drm,
-					   xe_lrc_size(xe, hwe->class),
+					   xe_gt_lrc_size(gt, hwe->class),
 					   GFP_KERNEL);
 		if (!default_lrc)
 			return -ENOMEM;
@@ -294,7 +294,7 @@ int xe_gt_record_default_lrcs(struct xe_gt *gt)
 		xe_map_memcpy_from(xe, default_lrc,
 				   &q->lrc[0].bo->vmap,
 				   xe_lrc_pphwsp_offset(&q->lrc[0]),
-				   xe_lrc_size(xe, hwe->class));
+				   xe_gt_lrc_size(gt, hwe->class));
 
 		gt->default_lrc[hwe->class] = default_lrc;
 put_nop_q:

drivers/gpu/drm/xe/xe_gt.h

@@ -8,6 +8,7 @@
 
 #include <drm/drm_util.h>
 
+#include "xe_device.h"
 #include "xe_device_types.h"
 #include "xe_hw_engine.h"
 
@@ -58,6 +59,12 @@ struct xe_hw_engine *xe_gt_hw_engine(struct xe_gt *gt,
 				     u16 instance,
 				     bool logical);
 
+static inline bool xe_gt_has_indirect_ring_state(struct xe_gt *gt)
+{
+	return gt->info.has_indirect_ring_state &&
+	       xe_device_uc_enabled(gt_to_xe(gt));
+}
+
 static inline bool xe_gt_is_media_type(struct xe_gt *gt)
 {
 	return gt->info.type == XE_GT_TYPE_MEDIA;

drivers/gpu/drm/xe/xe_gt_types.h

@@ -110,8 +110,6 @@ struct xe_gt {
 	struct {
 		/** @info.type: type of GT */
 		enum xe_gt_type type;
-		/** @info.id: Unique ID of this GT within the PCI Device */
-		u8 id;
 		/** @info.reference_clock: clock frequency */
 		u32 reference_clock;
 		/** @info.engine_mask: mask of engines present on GT */
@@ -124,6 +122,10 @@ struct xe_gt {
 		u64 __engine_mask;
 		/** @info.gmdid: raw GMD_ID value from hardware */
 		u32 gmdid;
+		/** @info.id: Unique ID of this GT within the PCI Device */
+		u8 id;
+		/** @info.has_indirect_ring_state: GT has indirect ring state support */
+		u8 has_indirect_ring_state:1;
 	} info;
 
 	/**

drivers/gpu/drm/xe/xe_guc_ads.c

@@ -267,7 +267,6 @@ static u32 engine_enable_mask(struct xe_gt *gt, enum xe_engine_class class)
 
 static size_t calculate_golden_lrc_size(struct xe_guc_ads *ads)
 {
-	struct xe_device *xe = ads_to_xe(ads);
 	struct xe_gt *gt = ads_to_gt(ads);
 	size_t total_size = 0, alloc_size, real_size;
 	int class;
@@ -276,7 +275,7 @@ static size_t calculate_golden_lrc_size(struct xe_guc_ads *ads)
 		if (!engine_enable_mask(gt, class))
 			continue;
 
-		real_size = xe_lrc_size(xe, class);
+		real_size = xe_gt_lrc_size(gt, class);
 		alloc_size = PAGE_ALIGN(real_size);
 		total_size += alloc_size;
 	}
@@ -774,7 +773,7 @@ static void guc_populate_golden_lrc(struct xe_guc_ads *ads)
 
 		xe_gt_assert(gt, gt->default_lrc[class]);
 
-		real_size = xe_lrc_size(xe, class);
+		real_size = xe_gt_lrc_size(gt, class);
 		alloc_size = PAGE_ALIGN(real_size);
 		total_size += alloc_size;
 

drivers/gpu/drm/xe/xe_guc_submit.c

@@ -677,7 +677,7 @@ static void submit_exec_queue(struct xe_exec_queue *q)
 	if (xe_exec_queue_is_parallel(q))
 		wq_item_append(q);
 	else
-		xe_lrc_write_ctx_reg(lrc, CTX_RING_TAIL, lrc->ring.tail);
+		xe_lrc_set_ring_tail(lrc, lrc->ring.tail);
 
 	if (exec_queue_suspended(q) && !xe_exec_queue_is_parallel(q))
 		return;

drivers/gpu/drm/xe/xe_lrc.h

@@ -21,14 +21,17 @@ int xe_lrc_init(struct xe_lrc *lrc, struct xe_hw_engine *hwe,
 		struct xe_exec_queue *q, struct xe_vm *vm, u32 ring_size);
 void xe_lrc_finish(struct xe_lrc *lrc);
 
-size_t xe_lrc_size(struct xe_device *xe, enum xe_engine_class class);
+size_t xe_gt_lrc_size(struct xe_gt *gt, enum xe_engine_class class);
 u32 xe_lrc_pphwsp_offset(struct xe_lrc *lrc);
 
+void xe_lrc_set_ring_tail(struct xe_lrc *lrc, u32 tail);
+u32 xe_lrc_ring_tail(struct xe_lrc *lrc);
 void xe_lrc_set_ring_head(struct xe_lrc *lrc, u32 head);
 u32 xe_lrc_ring_head(struct xe_lrc *lrc);
 u32 xe_lrc_ring_space(struct xe_lrc *lrc);
 void xe_lrc_write_ring(struct xe_lrc *lrc, const void *data, size_t size);
 
+u32 xe_lrc_indirect_ring_ggtt_addr(struct xe_lrc *lrc);
 u32 xe_lrc_ggtt_addr(struct xe_lrc *lrc);
 u32 *xe_lrc_regs(struct xe_lrc *lrc);
 

drivers/gpu/drm/xe/xe_lrc_types.h

@@ -20,10 +20,14 @@ struct xe_lrc {
 	 */
 	struct xe_bo *bo;
 
+	/** @size: size of lrc including any indirect ring state page */
+	u32 size;
+
 	/** @tile: tile which this LRC belongs to */
 	struct xe_tile *tile;
 
 	/** @flags: LRC flags */
+#define XE_LRC_FLAG_INDIRECT_RING_STATE		0x1
 	u32 flags;
 
 	/** @ring: submission ring state */

drivers/gpu/drm/xe/xe_pci.c

@@ -661,6 +661,7 @@ static int xe_info_init(struct xe_device *xe,
 		gt = tile->primary_gt;
 		gt->info.id = xe->info.gt_count++;
 		gt->info.type = XE_GT_TYPE_MAIN;
+		gt->info.has_indirect_ring_state = graphics_desc->has_indirect_ring_state;
 		gt->info.__engine_mask = graphics_desc->hw_engine_mask;
 		if (MEDIA_VER(xe) < 13 && media_desc)
 			gt->info.__engine_mask |= media_desc->hw_engine_mask;
@@ -678,6 +679,7 @@ static int xe_info_init(struct xe_device *xe,
 
 		gt = tile->media_gt;
 		gt->info.type = XE_GT_TYPE_MEDIA;
+		gt->info.has_indirect_ring_state = media_desc->has_indirect_ring_state;
 		gt->info.__engine_mask = media_desc->hw_engine_mask;
 		gt->mmio.adj_offset = MEDIA_GT_GSI_OFFSET;
 		gt->mmio.adj_limit = MEDIA_GT_GSI_LENGTH;

drivers/gpu/drm/xe/xe_pci_types.h

@@ -27,6 +27,7 @@ struct xe_graphics_desc {
 	u8 has_asid:1;
 	u8 has_atomic_enable_pte_bit:1;
 	u8 has_flat_ccs:1;
+	u8 has_indirect_ring_state:1;
 	u8 has_range_tlb_invalidation:1;
 	u8 has_usm:1;
 };
@@ -37,6 +38,8 @@ struct xe_media_desc {
 	u8 rel;
 
 	u64 hw_engine_mask;	/* hardware engines provided by media IP */
+
+	u8 has_indirect_ring_state:1;
 };
 
 struct gmdid_map {