1 files changed, 88 insertions, 56 deletions

diff --git a/drivers/gpu/drm/xe/xe_exec.c b/drivers/gpu/drm/xe/xe_exec.c
index 97eeb973e897..4d81210e41f5 100644
--- a/drivers/gpu/drm/xe/xe_exec.c
+++ b/drivers/gpu/drm/xe/xe_exec.c
@@ -8,16 +8,20 @@
 #include <drm/drm_device.h>
 #include <drm/drm_exec.h>
 #include <drm/drm_file.h>
-#include <drm/xe_drm.h>
+#include <uapi/drm/xe_drm.h>
 #include <linux/delay.h>
 
 #include "xe_bo.h"
 #include "xe_device.h"
 #include "xe_exec_queue.h"
+#include "xe_hw_engine_group.h"
 #include "xe_macros.h"
+#include "xe_pm.h"
 #include "xe_ring_ops_types.h"
 #include "xe_sched_job.h"
 #include "xe_sync.h"
+#include "xe_svm.h"
+#include "xe_trace.h"
 #include "xe_vm.h"
 
 /**
@@ -30,9 +34,9 @@
  * - Binding at exec time
  * - Flow controlling the ring at exec time
  *
- * In XE we avoid all of this complication by not allowing a BO list to be
+ * In Xe we avoid all of this complication by not allowing a BO list to be
  * passed into an exec, using the dma-buf implicit sync uAPI, have binds as
- * seperate operations, and using the DRM scheduler to flow control the ring.
+ * separate operations, and using the DRM scheduler to flow control the ring.
  * Let's deep dive on each of these.
  *
  * We can get away from a BO list by forcing the user to use in / out fences on
@@ -40,11 +44,6 @@
  * user knows an exec writes to a BO and reads from the BO in the next exec, it
  * is the user's responsibility to pass in / out fence between the two execs).
  *
- * Implicit dependencies for external BOs are handled by using the dma-buf
- * implicit dependency uAPI (TODO: add link). To make this works each exec must
- * install the job's fence into the DMA_RESV_USAGE_WRITE slot of every external
- * BO mapped in the VM.
- *
  * We do not allow a user to trigger a bind at exec time rather we have a VM
  * bind IOCTL which uses the same in / out fence interface as exec. In that
  * sense, a VM bind is basically the same operation as an exec from the user
@@ -58,8 +57,8 @@
  * behind any pending kernel operations on any external BOs in VM or any BOs
  * private to the VM. This is accomplished by the rebinds waiting on BOs
  * DMA_RESV_USAGE_KERNEL slot (kernel ops) and kernel ops waiting on all BOs
- * slots (inflight execs are in the DMA_RESV_USAGE_BOOKING for private BOs and
- * in DMA_RESV_USAGE_WRITE for external BOs).
+ * slots (inflight execs are in the DMA_RESV_USAGE_BOOKKEEP for private BOs and
+ * for external BOs).
  *
  * Rebinds / dma-resv usage applies to non-compute mode VMs only as for compute
  * mode VMs we use preempt fences and a rebind worker (TODO: add link).
@@ -101,9 +100,13 @@
 static int xe_exec_fn(struct drm_gpuvm_exec *vm_exec)
 {
 	struct xe_vm *vm = container_of(vm_exec->vm, struct xe_vm, gpuvm);
+	int ret;
 
 	/* The fence slot added here is intended for the exec sched job. */
-	return xe_vm_validate_rebind(vm, &vm_exec->exec, 1);
+	xe_vm_set_validation_exec(vm, &vm_exec->exec);
+	ret = xe_vm_validate_rebind(vm, &vm_exec->exec, 1);
+	xe_vm_set_validation_exec(vm, NULL);
+	return ret;
 }
 
 int xe_exec_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
@@ -118,12 +121,14 @@ int xe_exec_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
 	u64 addresses[XE_HW_ENGINE_MAX_INSTANCE];
 	struct drm_gpuvm_exec vm_exec = {.extra.fn = xe_exec_fn};
 	struct drm_exec *exec = &vm_exec.exec;
-	u32 i, num_syncs = 0, num_ufence = 0;
+	u32 i, num_syncs, num_ufence = 0;
+	struct xe_validation_ctx ctx;
 	struct xe_sched_job *job;
 	struct xe_vm *vm;
-	bool write_locked, skip_retry = false;
-	ktime_t end = 0;
+	bool write_locked;
 	int err = 0;
+	struct xe_hw_engine_group *group;
+	enum xe_hw_engine_group_execution_mode mode, previous_mode;
 
 	if (XE_IOCTL_DBG(xe, args->extensions) ||
 	    XE_IOCTL_DBG(xe, args->pad[0] || args->pad[1] || args->pad[2]) ||
@@ -134,18 +139,28 @@ int xe_exec_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
 	if (XE_IOCTL_DBG(xe, !q))
 		return -ENOENT;
 
-	if (XE_IOCTL_DBG(xe, q->flags & EXEC_QUEUE_FLAG_VM))
-		return -EINVAL;
+	if (XE_IOCTL_DBG(xe, q->flags & EXEC_QUEUE_FLAG_VM)) {
+		err = -EINVAL;
+		goto err_exec_queue;
+	}
 
 	if (XE_IOCTL_DBG(xe, args->num_batch_buffer &&
-			 q->width != args->num_batch_buffer))
-		return -EINVAL;
+			 q->width != args->num_batch_buffer)) {
+		err = -EINVAL;
+		goto err_exec_queue;
+	}
 
-	if (XE_IOCTL_DBG(xe, q->flags & EXEC_QUEUE_FLAG_BANNED)) {
+	if (XE_IOCTL_DBG(xe, q->ops->reset_status(q))) {
 		err = -ECANCELED;
 		goto err_exec_queue;
 	}
 
+	if (atomic_read(&q->job_cnt) >= XE_MAX_JOB_COUNT_PER_EXEC_QUEUE) {
+		trace_xe_exec_queue_reach_max_job_count(q, XE_MAX_JOB_COUNT_PER_EXEC_QUEUE);
+		err = -EAGAIN;
+		goto err_exec_queue;
+	}
+
 	if (args->num_syncs) {
 		syncs = kcalloc(args->num_syncs, sizeof(*syncs), GFP_KERNEL);
 		if (!syncs) {
@@ -156,15 +171,16 @@ int xe_exec_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
 
 	vm = q->vm;
 
-	for (i = 0; i < args->num_syncs; i++) {
-		err = xe_sync_entry_parse(xe, xef, &syncs[num_syncs++],
-					  &syncs_user[i], SYNC_PARSE_FLAG_EXEC |
+	for (num_syncs = 0; num_syncs < args->num_syncs; num_syncs++) {
+		err = xe_sync_entry_parse(xe, xef, &syncs[num_syncs],
+					  &syncs_user[num_syncs], NULL, 0,
+					  SYNC_PARSE_FLAG_EXEC |
 					  (xe_vm_in_lr_mode(vm) ?
 					   SYNC_PARSE_FLAG_LR_MODE : 0));
 		if (err)
 			goto err_syncs;
 
-		if (xe_sync_is_ufence(&syncs[i]))
+		if (xe_sync_is_ufence(&syncs[num_syncs]))
 			num_ufence++;
 	}
 
@@ -174,14 +190,23 @@ int xe_exec_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
 	}
 
 	if (xe_exec_queue_is_parallel(q)) {
-		err = __copy_from_user(addresses, addresses_user, sizeof(u64) *
-				       q->width);
+		err = copy_from_user(addresses, addresses_user, sizeof(u64) *
+				     q->width);
 		if (err) {
 			err = -EFAULT;
 			goto err_syncs;
 		}
 	}
 
+	group = q->hwe->hw_engine_group;
+	mode = xe_hw_engine_group_find_exec_mode(q);
+
+	if (mode == EXEC_MODE_DMA_FENCE) {
+		err = xe_hw_engine_group_get_mode(group, mode, &previous_mode);
+		if (err)
+			goto err_syncs;
+	}
+
 retry:
 	if (!xe_vm_in_lr_mode(vm) && xe_vm_userptr_check_repin(vm)) {
 		err = down_write_killable(&vm->lock);
@@ -192,7 +217,7 @@ retry:
 		write_locked = false;
 	}
 	if (err)
-		goto err_syncs;
+		goto err_hw_exec_mode;
 
 	if (write_locked) {
 		err = xe_vm_userptr_pin(vm);
@@ -213,6 +238,7 @@ retry:
 			fence = xe_sync_in_fence_get(syncs, num_syncs, q, vm);
 			if (IS_ERR(fence)) {
 				err = PTR_ERR(fence);
+				xe_vm_unlock(vm);
 				goto err_unlock_list;
 			}
 			for (i = 0; i < num_syncs; i++)
@@ -225,17 +251,21 @@ retry:
 		goto err_unlock_list;
 	}
 
-	vm_exec.vm = &vm->gpuvm;
-	vm_exec.flags = DRM_EXEC_INTERRUPTIBLE_WAIT;
-	if (xe_vm_in_lr_mode(vm)) {
-		drm_exec_init(exec, vm_exec.flags, 0);
-	} else {
-		err = drm_gpuvm_exec_lock(&vm_exec);
-		if (err) {
-			if (xe_vm_validate_should_retry(exec, err, &end))
-				err = -EAGAIN;
+	/*
+	 * It's OK to block interruptible here with the vm lock held, since
+	 * on task freezing during suspend / hibernate, the call will
+	 * return -ERESTARTSYS and the IOCTL will be rerun.
+	 */
+	err = xe_pm_block_on_suspend(xe);
+	if (err)
+		goto err_unlock_list;
+
+	if (!xe_vm_in_lr_mode(vm)) {
+		vm_exec.vm = &vm->gpuvm;
+		vm_exec.flags = DRM_EXEC_INTERRUPTIBLE_WAIT;
+		err = xe_validation_exec_lock(&ctx, &vm_exec, &xe->val);
+		if (err)
 			goto err_unlock_list;
-		}
 	}
 
 	if (xe_vm_is_closed_or_banned(q->vm)) {
@@ -244,10 +274,10 @@ retry:
 		goto err_exec;
 	}
 
-	if (xe_exec_queue_is_lr(q) && xe_exec_queue_ring_full(q)) {
-		err = -EWOULDBLOCK;	/* Aliased to -EAGAIN */
-		skip_retry = true;
-		goto err_exec;
+	if (xe_exec_queue_uses_pxp(q)) {
+		err = xe_vm_validate_protected(q->vm);
+		if (err)
+			goto err_exec;
 	}
 
 	job = xe_sched_job_create(q, xe_exec_queue_is_parallel(q) ?
@@ -259,9 +289,9 @@ retry:
 
 	/* Wait behind rebinds */
 	if (!xe_vm_in_lr_mode(vm)) {
-		err = drm_sched_job_add_resv_dependencies(&job->drm,
-							  xe_vm_resv(vm),
-							  DMA_RESV_USAGE_KERNEL);
+		err = xe_sched_job_add_deps(job,
+					    xe_vm_resv(vm),
+					    DMA_RESV_USAGE_KERNEL);
 		if (err)
 			goto err_put_job;
 	}
@@ -272,11 +302,7 @@ retry:
 		goto err_put_job;
 
 	if (!xe_vm_in_lr_mode(vm)) {
-		err = xe_sched_job_last_fence_add_dep(job, vm);
-		if (err)
-			goto err_put_job;
-
-		err = down_read_interruptible(&vm->userptr.notifier_lock);
+		err = xe_svm_notifier_lock_interruptible(vm);
 		if (err)
 			goto err_put_job;
 
@@ -292,15 +318,14 @@ retry:
 	xe_sched_job_arm(job);
 	if (!xe_vm_in_lr_mode(vm))
 		drm_gpuvm_resv_add_fence(&vm->gpuvm, exec, &job->drm.s_fence->finished,
-					 DMA_RESV_USAGE_BOOKKEEP, DMA_RESV_USAGE_WRITE);
+					 DMA_RESV_USAGE_BOOKKEEP,
+					 DMA_RESV_USAGE_BOOKKEEP);
 
 	for (i = 0; i < num_syncs; i++) {
 		xe_sync_entry_signal(&syncs[i], &job->drm.s_fence->finished);
 		xe_sched_job_init_user_fence(job, &syncs[i]);
 	}
 
-	if (xe_exec_queue_is_lr(q))
-		q->ring_ops->emit_job(job);
 	if (!xe_vm_in_lr_mode(vm))
 		xe_exec_queue_last_fence_set(q, vm, &job->drm.s_fence->finished);
 	xe_sched_job_push(job);
@@ -312,21 +337,28 @@ retry:
 		spin_unlock(&xe->ttm.lru_lock);
 	}
 
+	if (mode == EXEC_MODE_LR)
+		xe_hw_engine_group_resume_faulting_lr_jobs(group);
+
 err_repin:
 	if (!xe_vm_in_lr_mode(vm))
-		up_read(&vm->userptr.notifier_lock);
+		xe_svm_notifier_unlock(vm);
 err_put_job:
 	if (err)
 		xe_sched_job_put(job);
 err_exec:
-	drm_exec_fini(exec);
+	if (!xe_vm_in_lr_mode(vm))
+		xe_validation_ctx_fini(&ctx);
 err_unlock_list:
 	up_read(&vm->lock);
-	if (err == -EAGAIN && !skip_retry)
+	if (err == -EAGAIN)
 		goto retry;
+err_hw_exec_mode:
+	if (mode == EXEC_MODE_DMA_FENCE)
+		xe_hw_engine_group_put(group);
 err_syncs:
-	for (i = 0; i < num_syncs; i++)
-		xe_sync_entry_cleanup(&syncs[i]);
+	while (num_syncs--)
+		xe_sync_entry_cleanup(&syncs[num_syncs]);
 	kfree(syncs);
 err_exec_queue:
 	xe_exec_queue_put(q);