Aren
/
NetBSD
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Don't reorder page on the age queue if the lookup is done by the 

pagedaemon.  This mimics normal kernel behaviour where pmap_kentered
mappings are not tracked for references.  Without this change the
vnode pager's clustering could cause one page to be released by
the pagedaemon, and the rest of the pages in the pageout cluster
made unlikely candidates to be released soon.
This commit is contained in:
pooka 2010-10-13 11:01:04 +00:00
parent d6b38309ca
commit aeada54307
1 changed files with 13 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

17
sys/rump/librump/rumpkern/vm.c
View File

@ -1,4 +1,4 @@
/* $NetBSD: vm.c,v 1.96 2010/09/24 22:51:51 rmind Exp $ */
/* $NetBSD: vm.c,v 1.97 2010/10/13 11:01:04 pooka Exp $ */
/*
* Copyright (c) 2007-2010 Antti Kantee. All Rights Reserved.
@ -41,7 +41,7 @@
*/
#include <sys/cdefs.h>
__KERNEL_RCSID(0, "$NetBSD: vm.c,v 1.96 2010/09/24 22:51:51 rmind Exp $");
__KERNEL_RCSID(0, "$NetBSD: vm.c,v 1.97 2010/10/13 11:01:04 pooka Exp $");
#include <sys/param.h>
#include <sys/atomic.h>
@ -464,14 +464,23 @@ uvm_pageratop(vaddr_t va)
return pg;
}
/* Called with the vm object locked */
/*
* Called with the vm object locked.
*
* Put vnode object pages at the end of the access queue to indicate
* they have been recently accessed and should not be immediate
* candidates for pageout. Do not do this for lookups done by
* the pagedaemon to mimic pmap_kentered mappings which don't track
* access information.
*/
struct vm_page *
uvm_pagelookup(struct uvm_object *uobj, voff_t off)
{
struct vm_page *pg;
bool ispagedaemon = curlwp == uvm.pagedaemon_lwp;
pg = rb_tree_find_node(&uobj->rb_tree, &off);
if (pg && !UVM_OBJ_IS_AOBJ(pg->uobject)) {
if (pg && !UVM_OBJ_IS_AOBJ(pg->uobject) && !ispagedaemon) {
mutex_enter(&uvm_pageqlock);
TAILQ_REMOVE(&vmpage_lruqueue, pg, pageq.queue);
TAILQ_INSERT_TAIL(&vmpage_lruqueue, pg, pageq.queue);