On Alan's advice, rather than do a wholesale conversion on a single

architecture from page queue lock to a hashed array of page locks
(based on a patch by Jeff Roberson), I've implemented page lock
support in the MI code and have only moved vm_page's hold_count
out from under page queue mutex to page lock. This changes
pmap_extract_and_hold on all pmaps.

Supported by: Bitgravity Inc.

Discussed with: alc, jeffr, and kib

1 files changed, 10 insertions, 4 deletions

diff --git a/sys/vm/vm_glue.c b/sys/vm/vm_glue.c
index 4eeaa4d..288c5d7 100644
--- a/sys/vm/vm_glue.c
+++ b/sys/vm/vm_glue.c
@@ -257,16 +257,18 @@ vm_imgact_hold_page(vm_object_t object, vm_ooffset_t offset)
 		if (m == NULL)
 			goto out;
 		if (rv != VM_PAGER_OK) {
+			vm_page_lock(m);
 			vm_page_lock_queues();
 			vm_page_free(m);
 			vm_page_unlock_queues();
+			vm_page_unlock(m);
 			m = NULL;
 			goto out;
 		}
 	}
-	vm_page_lock_queues();
+	vm_page_lock(m);
 	vm_page_hold(m);
-	vm_page_unlock_queues();
+	vm_page_unlock(m);
 	vm_page_wakeup(m);
 out:
 	VM_OBJECT_UNLOCK(object);
@@ -300,9 +302,9 @@ vm_imgact_unmap_page(struct sf_buf *sf)
 	m = sf_buf_page(sf);
 	sf_buf_free(sf);
 	sched_unpin();
-	vm_page_lock_queues();
+	vm_page_lock(m);
 	vm_page_unhold(m);
-	vm_page_unlock_queues();
+	vm_page_unlock(m);
 }
 
 void
@@ -434,10 +436,12 @@ vm_thread_stack_dispose(vm_object_t ksobj, vm_offset_t ks, int pages)
 		m = vm_page_lookup(ksobj, i);
 		if (m == NULL)
 			panic("vm_thread_dispose: kstack already missing?");
+		vm_page_lock(m);
 		vm_page_lock_queues();
 		vm_page_unwire(m, 0);
 		vm_page_free(m);
 		vm_page_unlock_queues();
+		vm_page_unlock(m);
 	}
 	VM_OBJECT_UNLOCK(ksobj);
 	vm_object_deallocate(ksobj);
@@ -524,9 +528,11 @@ vm_thread_swapout(struct thread *td)
 		if (m == NULL)
 			panic("vm_thread_swapout: kstack already missing?");
 		vm_page_dirty(m);
+		vm_page_lock(m);
 		vm_page_lock_queues();
 		vm_page_unwire(m, 0);
 		vm_page_unlock_queues();
+		vm_page_unlock(m);
 	}
 	VM_OBJECT_UNLOCK(ksobj);
 }