[PATCH v4 1/2] drm/xe: Userptr invalidation race with binds fixes

[Matthew Brost <matthew.brost@intel.com>]

Always wait on dma-resv bookkeep slots if userptr invalidation has raced
with a bind ensuring PTEs temporally setup to invalidated pages are
never accessed.

Fixup initial bind handling always add VMAs to invalidation list and
wait dma-resv bookkeep slots.

Always hold notifier across TLB invalidation in notifier to prevent a
UAF if an unbind races.

Including all of the above changes for Fixes patch in hopes of an easier
backport which fix a single patch.

v2:
 - Wait dma-resv bookkeep before issuing PTE zap (Thomas)
 - Support scratch page on invalidation (Thomas)
v3:
 - Drop clear of PTEs (Thomas)
v4:
 - Remove double dma-resv wait

Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: <stable@vger.kernel.org>
Fixes: e8babb280b5e ("drm/xe: Convert multiple bind ops into single job")
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
---
 drivers/gpu/drm/xe/xe_pt.c | 21 ++++++++++++---------
 drivers/gpu/drm/xe/xe_vm.c |  4 ++--
 2 files changed, 14 insertions(+), 11 deletions(-)

diff --git a/drivers/gpu/drm/xe/xe_pt.c b/drivers/gpu/drm/xe/xe_pt.c
index 1ddcc7e79a93..ffd23c3564c5 100644
--- a/drivers/gpu/drm/xe/xe_pt.c
+++ b/drivers/gpu/drm/xe/xe_pt.c
@@ -1215,9 +1215,6 @@ static int vma_check_userptr(struct xe_vm *vm, struct xe_vma *vma,
 	uvma = to_userptr_vma(vma);
 	notifier_seq = uvma->userptr.notifier_seq;
 
-	if (uvma->userptr.initial_bind && !xe_vm_in_fault_mode(vm))
-		return 0;
-
 	if (!mmu_interval_read_retry(&uvma->userptr.notifier,
 				     notifier_seq) &&
 	    !xe_pt_userptr_inject_eagain(uvma))
@@ -1226,6 +1223,8 @@ static int vma_check_userptr(struct xe_vm *vm, struct xe_vma *vma,
 	if (xe_vm_in_fault_mode(vm)) {
 		return -EAGAIN;
 	} else {
+		long err;
+
 		spin_lock(&vm->userptr.invalidated_lock);
 		list_move_tail(&uvma->userptr.invalidate_link,
 			       &vm->userptr.invalidated);
@@ -1234,19 +1233,23 @@ static int vma_check_userptr(struct xe_vm *vm, struct xe_vma *vma,
 		if (xe_vm_in_preempt_fence_mode(vm)) {
 			struct dma_resv_iter cursor;
 			struct dma_fence *fence;
-			long err;
 
 			dma_resv_iter_begin(&cursor, xe_vm_resv(vm),
 					    DMA_RESV_USAGE_BOOKKEEP);
 			dma_resv_for_each_fence_unlocked(&cursor, fence)
 				dma_fence_enable_sw_signaling(fence);
 			dma_resv_iter_end(&cursor);
-
-			err = dma_resv_wait_timeout(xe_vm_resv(vm),
-						    DMA_RESV_USAGE_BOOKKEEP,
-						    false, MAX_SCHEDULE_TIMEOUT);
-			XE_WARN_ON(err <= 0);
 		}
+
+		/*
+		 * We are temporally installing PTEs pointing to invalidated
+		 * pages, ensure VM is idle to avoid data corruption. PTEs fixed
+		 * up upon next exec or in rebind worker.
+		 */
+		err = dma_resv_wait_timeout(xe_vm_resv(vm),
+					    DMA_RESV_USAGE_BOOKKEEP,
+					    false, MAX_SCHEDULE_TIMEOUT);
+		XE_WARN_ON(err <= 0);
 	}
 
 	return 0;
diff --git a/drivers/gpu/drm/xe/xe_vm.c b/drivers/gpu/drm/xe/xe_vm.c
index 996000f2424e..9b2acb069a77 100644
--- a/drivers/gpu/drm/xe/xe_vm.c
+++ b/drivers/gpu/drm/xe/xe_vm.c
@@ -623,8 +623,6 @@ static bool vma_userptr_invalidate(struct mmu_interval_notifier *mni,
 		spin_unlock(&vm->userptr.invalidated_lock);
 	}
 
-	up_write(&vm->userptr.notifier_lock);
-
 	/*
 	 * Preempt fences turn into schedule disables, pipeline these.
 	 * Note that even in fault mode, we need to wait for binds and
@@ -647,6 +645,8 @@ static bool vma_userptr_invalidate(struct mmu_interval_notifier *mni,
 		XE_WARN_ON(err);
 	}
 
+	up_write(&vm->userptr.notifier_lock);
+
 	trace_xe_vma_userptr_invalidate_complete(vma);
 
 	return true;
-- 
2.34.1