page fault panic in vm system
dillon at apollo.backplane.com
Sat Jul 9 16:02:16 PDT 2011
:On Tue, Jul 05, 2011 at 08:47:46AM -1000, Peter Avalos wrote:
:> I got a panic 2 nights in a row (both at around 2:55am which is weird).
:I got another panic last night with mysqld as the current process. The
:kernel for this one is:
:DragonFly ylem.theshell.com 2.11-DEVELOPMENT DragonFly
:v220.127.116.111.g96956-DEVELOPMENT #54: Tue Jul 5 23:12:44 HST 2011
:root at ylem.theshell.com:/usr/obj/usr/src/sys/YLEM i386
:I wasn't able to get a core, but this was on the console. Not sure if
:much can be gained from this:
:Fatal trap 12: page fault while in kernel mode
:cpuid =3D 1; lapic.id =3D 02000000
:fault virtual address =3D 0x14
:fault code =3D supervisor read, page not present
:instruction pointer =3D 0x8:0xc02f2e20
This is vm_page_rb_tree_RB_LOOKUP() .. the red-black lookup code
for a vm_page in an object. Called from vm_page_lookup().
Currently the global vm_token protects that field. I didn't see any
particular race there but the problem could be related to a more
general reference count problem with VM objects.
More information about the Bugs