Re: [issue530] crash from latest -HEAD

[Matthew Dillon <dillon@xxxxxxxxxxxxxxxxxxxx>]

:On Thu, Mar 08, 2007 at 01:13:36PM -0500, Peter Avalos wrote:
:>=20
:> I've compiled up a new kernel with that patch, so we'll see.  The first t=
:ime it
:> happened only after 13 hours of uptime.  The second time took 35 days.  I=
:f you
:> have any suggestions on how to trigger it, I'll see what I can do.
:>=20
:
:Now that I'm running with that patch, I'm getting plenty of those warnings.
:Is that expected?
:
:--Peter

    Lets try a slightly modified patch... this one only generates warnings
    if the page is on a paging queue.

    There is clearly a race happening somewhere, and I think the solution
    is to remove the page from the paging queues (since we are trying to
    destroy it anyhow), but I'd like to accumulate a bit more evidence.

					-Matt
					Matthew Dillon 
					<dillon@backplane.com>

Index: vm_object.c
===================================================================
RCS file: /cvs/src/sys/vm/vm_object.c,v
retrieving revision 1.29
diff -u -r1.29 vm_object.c
--- vm_object.c	28 Dec 2006 21:24:02 -0000	1.29
+++ vm_object.c	9 Mar 2007 05:15:41 -0000
@@ -460,6 +460,8 @@
 		vm_page_free(p);
 		mycpu->gd_cnt.v_pfree++;
 	} else {
+		if (p->queue != PQ_NONE)
+			kprintf("vm_object_terminate: Warning: Encountered wired page %p on queue %d\n", p, p->queue);
 		vm_page_busy(p);
 		vm_page_remove(p);
 		vm_page_wakeup(p);