drm/xe: Invalidate TLB also on bind if in scratch page mode

For scratch table mode we need to cover the case where a scratch PTE might
have been pre-fetched and cached and used instead of that of the newly
bound vma.
For compute vms, invalidate TLB globally using GuC before signalling
bind complete. For !long-running vms, invalidate TLB at batch start.

Also document how TLB invalidation works.

v2:
- Fix a pointer to the comment about TLB invalidation (Jose Souza).
- Add a bool to the vm whether we want to invalidate TLB at batch start.
- Invalidate TLB also on BCS- and video engines at batch start where
  needed.
- Use BIT() macro instead of explicit shift.

Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Tested-by: José Roberto de Souza <jose.souza@intel.com> #v1
Reported-by: José Roberto de Souza <jose.souza@intel.com> #v1
Link: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/291
Closes: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/291
Acked-by: José Roberto de Souza <jose.souza@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

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

@@ -73,6 +73,7 @@
 #define   PIPE_CONTROL_STORE_DATA_INDEX			(1<<21)
 #define   PIPE_CONTROL_CS_STALL				(1<<20)
 #define   PIPE_CONTROL_GLOBAL_SNAPSHOT_RESET		(1<<19)
+#define	  PIPE_CONTROL_TLB_INVALIDATE			BIT(18)
 #define   PIPE_CONTROL_PSD_SYNC				(1<<17)
 #define   PIPE_CONTROL_QW_WRITE				(1<<14)
 #define   PIPE_CONTROL_DEPTH_STALL			(1<<13)

drivers/gpu/drm/xe/xe_pt.c

@@ -1297,7 +1297,20 @@ __xe_pt_bind_vma(struct xe_tile *tile, struct xe_vma *vma, struct xe_engine *e,
 
 	xe_vm_dbg_print_entries(tile_to_xe(tile), entries, num_entries);
 
-	if (rebind && !xe_vm_no_dma_fences(vma->vm)) {
+	/*
+	 * If rebind, we have to invalidate TLB on !LR vms to invalidate
+	 * cached PTEs point to freed memory. on LR vms this is done
+	 * automatically when the context is re-enabled by the rebind worker,
+	 * or in fault mode it was invalidated on PTE zapping.
+	 *
+	 * If !rebind, and scratch enabled VMs, there is a chance the scratch
+	 * PTE is already cached in the TLB so it needs to be invalidated.
+	 * on !LR VMs this is done in the ring ops preceding a batch, but on
+	 * non-faulting LR, in particular on user-space batch buffer chaining,
+	 * it needs to be done here.
+	 */
+	if ((rebind && !xe_vm_no_dma_fences(vm) && !vm->batch_invalidate_tlb) ||
+	    (!rebind && vm->scratch_bo[tile->id] && xe_vm_in_compute_mode(vm))) {
 		ifence = kzalloc(sizeof(*ifence), GFP_KERNEL);
 		if (!ifence)
 			return ERR_PTR(-ENOMEM);
@@ -1313,7 +1326,7 @@ __xe_pt_bind_vma(struct xe_tile *tile, struct xe_vma *vma, struct xe_engine *e,
 		LLIST_HEAD(deferred);
 
 		/* TLB invalidation must be done before signaling rebind */
-		if (rebind && !xe_vm_no_dma_fences(vma->vm)) {
+		if (ifence) {
 			int err = invalidation_fence_init(tile->primary_gt, ifence, fence,
 							  vma);
 			if (err) {

drivers/gpu/drm/xe/xe_ring_ops.c

@@ -15,6 +15,7 @@
 #include "xe_macros.h"
 #include "xe_sched_job.h"
 #include "xe_vm_types.h"
+#include "xe_vm.h"
 
 /*
  * 3D-related flags that can't be set on _engines_ that lack access to the 3D
@@ -74,9 +75,11 @@ static int emit_store_imm_ggtt(u32 addr, u32 value, u32 *dw, int i)
 	return i;
 }
 
-static int emit_flush_imm_ggtt(u32 addr, u32 value, u32 *dw, int i)
+static int emit_flush_imm_ggtt(u32 addr, u32 value, bool invalidate_tlb,
+			       u32 *dw, int i)
 {
-	dw[i++] = (MI_FLUSH_DW + 1) | MI_FLUSH_DW_OP_STOREDW;
+	dw[i++] = (MI_FLUSH_DW + 1) | MI_FLUSH_DW_OP_STOREDW |
+		(invalidate_tlb ? MI_INVALIDATE_TLB : 0);
 	dw[i++] = addr | MI_FLUSH_DW_USE_GTT;
 	dw[i++] = 0;
 	dw[i++] = value;
@@ -107,7 +110,8 @@ static int emit_flush_invalidate(u32 flag, u32 *dw, int i)
 	return i;
 }
 
-static int emit_pipe_invalidate(u32 mask_flags, u32 *dw, int i)
+static int emit_pipe_invalidate(u32 mask_flags, bool invalidate_tlb, u32 *dw,
+				int i)
 {
 	u32 flags = PIPE_CONTROL_CS_STALL |
 		PIPE_CONTROL_COMMAND_CACHE_INVALIDATE |
@@ -119,6 +123,9 @@ static int emit_pipe_invalidate(u32 mask_flags, u32 *dw, int i)
 		PIPE_CONTROL_QW_WRITE |
 		PIPE_CONTROL_STORE_DATA_INDEX;
 
+	if (invalidate_tlb)
+		flags |= PIPE_CONTROL_TLB_INVALIDATE;
+
 	flags &= ~mask_flags;
 
 	dw[i++] = GFX_OP_PIPE_CONTROL(6);
@@ -170,9 +177,17 @@ static void __emit_job_gen12_copy(struct xe_sched_job *job, struct xe_lrc *lrc,
 {
 	u32 dw[MAX_JOB_SIZE_DW], i = 0;
 	u32 ppgtt_flag = get_ppgtt_flag(job);
+	struct xe_vm *vm = job->engine->vm;
 
-	i = emit_store_imm_ggtt(xe_lrc_start_seqno_ggtt_addr(lrc),
-				seqno, dw, i);
+	if (vm->batch_invalidate_tlb) {
+		dw[i++] = preparser_disable(true);
+		i = emit_flush_imm_ggtt(xe_lrc_start_seqno_ggtt_addr(lrc),
+					seqno, true, dw, i);
+		dw[i++] = preparser_disable(false);
+	} else {
+		i = emit_store_imm_ggtt(xe_lrc_start_seqno_ggtt_addr(lrc),
+					seqno, dw, i);
+	}
 
 	i = emit_bb_start(batch_addr, ppgtt_flag, dw, i);
 
@@ -181,7 +196,7 @@ static void __emit_job_gen12_copy(struct xe_sched_job *job, struct xe_lrc *lrc,
 						job->user_fence.value,
 						dw, i);
 
-	i = emit_flush_imm_ggtt(xe_lrc_seqno_ggtt_addr(lrc), seqno, dw, i);
+	i = emit_flush_imm_ggtt(xe_lrc_seqno_ggtt_addr(lrc), seqno, false, dw, i);
 
 	i = emit_user_interrupt(dw, i);
 
@@ -210,6 +225,7 @@ static void __emit_job_gen12_video(struct xe_sched_job *job, struct xe_lrc *lrc,
 	struct xe_gt *gt = job->engine->gt;
 	struct xe_device *xe = gt_to_xe(gt);
 	bool decode = job->engine->class == XE_ENGINE_CLASS_VIDEO_DECODE;
+	struct xe_vm *vm = job->engine->vm;
 
 	dw[i++] = preparser_disable(true);
 
@@ -220,10 +236,16 @@ static void __emit_job_gen12_video(struct xe_sched_job *job, struct xe_lrc *lrc,
 		else
 			i = emit_aux_table_inv(gt, VE0_AUX_INV, dw, i);
 	}
+
+	if (vm->batch_invalidate_tlb)
+		i = emit_flush_imm_ggtt(xe_lrc_start_seqno_ggtt_addr(lrc),
+					seqno, true, dw, i);
+
 	dw[i++] = preparser_disable(false);
 
-	i = emit_store_imm_ggtt(xe_lrc_start_seqno_ggtt_addr(lrc),
-				seqno, dw, i);
+	if (!vm->batch_invalidate_tlb)
+		i = emit_store_imm_ggtt(xe_lrc_start_seqno_ggtt_addr(lrc),
+					seqno, dw, i);
 
 	i = emit_bb_start(batch_addr, ppgtt_flag, dw, i);
 
@@ -232,7 +254,7 @@ static void __emit_job_gen12_video(struct xe_sched_job *job, struct xe_lrc *lrc,
 						job->user_fence.value,
 						dw, i);
 
-	i = emit_flush_imm_ggtt(xe_lrc_seqno_ggtt_addr(lrc), seqno, dw, i);
+	i = emit_flush_imm_ggtt(xe_lrc_seqno_ggtt_addr(lrc), seqno, false, dw, i);
 
 	i = emit_user_interrupt(dw, i);
 
@@ -250,6 +272,7 @@ static void __emit_job_gen12_render_compute(struct xe_sched_job *job,
 	struct xe_gt *gt = job->engine->gt;
 	struct xe_device *xe = gt_to_xe(gt);
 	bool lacks_render = !(gt->info.engine_mask & XE_HW_ENGINE_RCS_MASK);
+	struct xe_vm *vm = job->engine->vm;
 	u32 mask_flags = 0;
 
 	dw[i++] = preparser_disable(true);
@@ -257,7 +280,9 @@ static void __emit_job_gen12_render_compute(struct xe_sched_job *job,
 		mask_flags = PIPE_CONTROL_3D_ARCH_FLAGS;
 	else if (job->engine->class == XE_ENGINE_CLASS_COMPUTE)
 		mask_flags = PIPE_CONTROL_3D_ENGINE_FLAGS;
-	i = emit_pipe_invalidate(mask_flags, dw, i);
+
+	/* See __xe_pt_bind_vma() for a discussion on TLB invalidations. */
+	i = emit_pipe_invalidate(mask_flags, vm->batch_invalidate_tlb, dw, i);
 
 	/* hsdes: 1809175790 */
 	if (has_aux_ccs(xe))

drivers/gpu/drm/xe/xe_vm.c

@@ -1237,11 +1237,13 @@ struct xe_vm *xe_vm_create(struct xe_device *xe, u32 flags)
 			if (err)
 				goto err_scratch_pt;
 		}
+		vm->batch_invalidate_tlb = true;
 	}
 
 	if (flags & DRM_XE_VM_CREATE_COMPUTE_MODE) {
 		INIT_WORK(&vm->preempt.rebind_work, preempt_rebind_work_func);
 		vm->flags |= XE_VM_FLAG_COMPUTE_MODE;
+		vm->batch_invalidate_tlb = false;
 	}
 
 	if (flags & DRM_XE_VM_CREATE_ASYNC_BIND_OPS) {

drivers/gpu/drm/xe/xe_vm_types.h

@@ -337,6 +337,9 @@ struct xe_vm {
 		/** @capture_once: capture only one error per VM */
 		bool capture_once;
 	} error_capture;
+
+	/** @batch_invalidate_tlb: Always invalidate TLB before batch start */
+	bool batch_invalidate_tlb;
 };
 
 #endif