On 09/01/14(Thu) 13:25, Brett Mahar wrote:
> Hi bugs,
>
> On 28 Dec I rebuilt to -current and ever since then the computer has been
> halting at the ddb> prompt sometimes when I try to /shutdown. It also happens
> at the "clearing /tmp" stage when booting up, but that is a lot less
> frequent.
>
> I've rebuilt to -current a few more times since than and keep getting this
> freeze on random occasions.
>
> The 0x address seems to be different each time, all the rest is the same. Not
> sure if this a hardware fault or something else - the last -current I built
> prior to this one was on Dec 18 (was on holidays after that) and it never
> showed this problem.
>
> When shutting down, the screen shows (hand copied, could have typos):
>
> /etc/rc.shutdown complete
> syncing disks...done
> uvm_fault(0xfffffe82120c2718, 0x0, 0, 2) -> e
> kernel: page fault trap, code=0
> Stopped at task_del+0x57: movq %rdx, 0(%rax)
> ddb{0}>
Looks like the burner method of radeondrm(4) is not using the right
cookie, can you try the diff below and let me know if it fixes your
problem?
Index: radeon_fb.c
===================================================================
RCS file: /home/ncvs/src/sys/dev/pci/drm/radeon/radeon_fb.c,v
retrieving revision 1.3
diff -u -p -r1.3 radeon_fb.c
--- radeon_fb.c 5 Dec 2013 13:29:56 -0000 1.3
+++ radeon_fb.c 9 Jan 2014 11:52:20 -0000
@@ -494,7 +494,8 @@ bool radeon_fbdev_robj_is_fb(struct rade
void
radeondrm_burner(void *v, u_int on, u_int flags)
{
- struct radeon_device *rdev = v;
+ struct rasops_info *ri = v;
+ struct radeon_device *rdev = ri->ri_hw;
task_del(systq, &rdev->burner_task);
