Don't attempt to clean device_pager backed objects at terminate time.
There is similar bogusness in the pageout daemon that will be fixed soon. This fixes a panic pointed out to me by Bruce Evans that occurs when /dev/mem is used to map managed memory.
This commit is contained in:
parent
673f8dc3be
commit
a465acdaf8
Notes:
svn2git
2020-12-20 02:59:44 +00:00
svn path=/head/; revision=5903
@ -61,7 +61,7 @@
|
||||
* any improvements or extensions that they make and grant Carnegie the
|
||||
* rights to redistribute these changes.
|
||||
*
|
||||
* $Id: vm_object.c,v 1.18 1995/01/13 13:30:24 davidg Exp $
|
||||
* $Id: vm_object.c,v 1.19 1995/01/24 10:13:14 davidg Exp $
|
||||
*/
|
||||
|
||||
/*
|
||||
@ -384,7 +384,8 @@ vm_object_terminate(object)
|
||||
* object are gone, so we don't need to lock it.
|
||||
*/
|
||||
|
||||
if ((object->flags & OBJ_INTERNAL) == 0) {
|
||||
if (((object->flags & OBJ_INTERNAL) == 0) &&
|
||||
object->pager && (object->pager->pg_type != PG_DEVICE)) {
|
||||
(void) vm_object_page_clean(object, 0, 0, TRUE, TRUE);
|
||||
}
|
||||
/*
|
||||
|
Loading…
Reference in New Issue
Block a user