There is a real deadlock as well as sleeping in atomic() bug in here, if
the bo put happens to be the last ref, since bo destruction wants to
grab the same spinlock and sleeping locks. Fix that by dropping the ref
using xe_bo_put_deferred(), and moving the final commit outside of the
lock. Dropping the lock around the put is tricky since the bo can go
out of scope and delete itself from the list, making it difficult to
navigate to the next list entry.
Fixes: 0845233388f8 ("drm/xe: Implement fdinfo memory stats printing")
Closes: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/2727
Signed-off-by: Matthew Auld <matthew.auld@xxxxxxxxx>
Cc: Himal Prasad Ghimiray <himal.prasad.ghimiray@xxxxxxxxx>
Cc: Tejas Upadhyay <tejas.upadhyay@xxxxxxxxx>
Cc: "Thomas Hellström" <thomas.hellstrom@xxxxxxxxxxxxxxx>
Cc: <stable@xxxxxxxxxxxxxxx> # v6.8+
Reviewed-by: Matthew Brost <matthew.brost@xxxxxxxxx>
Reviewed-by: Tejas Upadhyay <tejas.upadhyay@xxxxxxxxx>
---
drivers/gpu/drm/xe/xe_drm_client.c | 6 +++++-
1 file changed, 5 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/xe/xe_drm_client.c b/drivers/gpu/drm/xe/xe_drm_client.c
index e64f4b645e2e..badfa045ead8 100644
--- a/drivers/gpu/drm/xe/xe_drm_client.c
+++ b/drivers/gpu/drm/xe/xe_drm_client.c
@@ -196,6 +196,7 @@ static void show_meminfo(struct drm_printer *p, struct drm_file *file)
struct xe_drm_client *client;
struct drm_gem_object *obj;
struct xe_bo *bo;
+ LLIST_HEAD(deferred);
unsigned int id;
u32 mem_type;
@@ -215,11 +216,14 @@ static void show_meminfo(struct drm_printer *p, struct drm_file *file)
list_for_each_entry(bo, &client->bos_list, client_link) {
if (!kref_get_unless_zero(&bo->ttm.base.refcount))
continue;
+
bo_meminfo(bo, stats);
- xe_bo_put(bo);
+ xe_bo_put_deferred(bo, &deferred);
}
spin_unlock(&client->bos_lock);
+ xe_bo_put_commit(&deferred);
+
for (mem_type = XE_PL_SYSTEM; mem_type < TTM_NUM_MEM_TYPES; ++mem_type) {
if (!xe_mem_type_to_name[mem_type])
continue;
--
2.46.0
