[Intel-xe] [PATCH 3/4] drm/xe: Add vm snapshot mutex for easily taking a vm snapshot during devcoredump

[Maarten at mblankhorst.nl]

From: Maarten Lankhorst <dev at lankhorst.se>

The devcoredump is done in fence signaling context. Because of this, we
cannot take any of the normal mutexes or we would invert.

Normal: Take vm->lock, dma_fence_wait()
Devcoredump: from dma_fence_wait() context, take vm->lock.

This doesn't work, and we only care about integrity, so take the locks
around additions and removals of vma's.

Signed-off-by: Maarten Lankhorst <dev at lankhorst.se>
---
 drivers/gpu/drm/xe/xe_vm.c       | 7 +++++++
 drivers/gpu/drm/xe/xe_vm_types.h | 5 +++++
 2 files changed, 12 insertions(+)

diff --git a/drivers/gpu/drm/xe/xe_vm.c b/drivers/gpu/drm/xe/xe_vm.c
index e17af7aa1ca1..b42ca1069c5b 100644
--- a/drivers/gpu/drm/xe/xe_vm.c
+++ b/drivers/gpu/drm/xe/xe_vm.c
@@ -1168,7 +1168,9 @@ static int xe_vm_insert_vma(struct xe_vm *vm, struct xe_vma *vma)
 	xe_assert(vm->xe, xe_vma_vm(vma) == vm);
 	lockdep_assert_held(&vm->lock);
 
+	mutex_lock(&vm->snap_mutex);
 	err = drm_gpuva_insert(&vm->gpuvm, &vma->gpuva);
+	mutex_unlock(&vm->snap_mutex);
 	XE_WARN_ON(err);	/* Shouldn't be possible */
 
 	return err;
@@ -1179,7 +1181,9 @@ static void xe_vm_remove_vma(struct xe_vm *vm, struct xe_vma *vma)
 	xe_assert(vm->xe, xe_vma_vm(vma) == vm);
 	lockdep_assert_held(&vm->lock);
 
+	mutex_lock(&vm->snap_mutex);
 	drm_gpuva_remove(&vma->gpuva);
+	mutex_unlock(&vm->snap_mutex);
 	if (vm->usm.last_fault_vma == vma)
 		vm->usm.last_fault_vma = NULL;
 }
@@ -1352,6 +1356,7 @@ struct xe_vm *xe_vm_create(struct xe_device *xe, u32 flags)
 	vm->flags = flags;
 
 	init_rwsem(&vm->lock);
+	mutex_init(&vm->snap_mutex);
 
 	INIT_LIST_HEAD(&vm->rebind_list);
 
@@ -1489,6 +1494,7 @@ struct xe_vm *xe_vm_create(struct xe_device *xe, u32 flags)
 	dma_resv_unlock(&vm->resv);
 	drm_gpuvm_destroy(&vm->gpuvm);
 err_put:
+	mutex_destroy(&vm->snap_mutex);
 	dma_resv_fini(&vm->resv);
 	for_each_tile(tile, xe, id)
 		xe_range_fence_tree_fini(&vm->rftree[id]);
@@ -1594,6 +1600,7 @@ void xe_vm_close_and_put(struct xe_vm *vm)
 	up_write(&vm->lock);
 
 	drm_gpuvm_destroy(&vm->gpuvm);
+	mutex_destroy(&vm->snap_mutex);
 
 	mutex_lock(&xe->usm.lock);
 	if (vm->flags & XE_VM_FLAG_FAULT_MODE)
diff --git a/drivers/gpu/drm/xe/xe_vm_types.h b/drivers/gpu/drm/xe/xe_vm_types.h
index 6852af366ea2..da42b1ee7aaa 100644
--- a/drivers/gpu/drm/xe/xe_vm_types.h
+++ b/drivers/gpu/drm/xe/xe_vm_types.h
@@ -182,6 +182,11 @@ struct xe_vm {
 	 * VM
 	 */
 	struct rw_semaphore lock;
+	/**
+	 * @snap_mutex: Mutex used to guard insertions and removals from gpuva,
+	 * so we can take a snapshot safely from devcoredump.
+	 */
+	struct mutex snap_mutex;
 
 	/**
 	 * @rebind_list: list of VMAs that need rebinding, and if they are
-- 
2.40.1