Re: UVM Fault

[Manuel Bouyer <bouyer%antioche.eu.org@localhost>]

On Wed, Jun 18, 2008 at 11:16:13AM +1000, Sarton O'Brien wrote:
> Manuel Bouyer wrote:
> >On Mon, Jun 16, 2008 at 12:15:12PM +1000, Sarton O'Brien wrote:
> >>On Sat, 14 Jun 2008 06:37:25 pm Mike Pumford wrote:
> >>>John R. Shannon wrote:
> >>>>Started happening immediately upon upgrade to 4.99.64:
> >>>>
> >>>>DOMU runs squid proxy. Repeatable fault occurs on forst web access via
> >>>>squid. Works if previous 4.99.63 kernel is used in DOMU.
> >>>>
> >>>Seeing something similar here on a quad core 4.99.64 on amd64.
> >>Here's the trace with registers as requested:
> >>
> >>
> >>Jun 16 12:04:59 squid[12711]: Squid Parent: child process 21641 started
> >>uvm_fault(0xffffa000141a1720, 0x0, 1) -> e
> >>fatal page fault in supervisor mode
> >>trap type 6 code 0 rip ffffffff803a34a6 cs e030 rflags 10282 cr2  0 cpl 0 
> >>rsp ffffa00013be7810
> >>kernel: page fault trap, code=0
> >>Stopped in pid 16282.1 (pinger) at      netbsd:rb_tree_insert_node+0x26:  
> >>m
> >>ovq     0(%rax),%r14
> >>rb_tree_insert_node() at netbsd:rb_tree_insert_node+0x26
> >>uvm_pagealloc_strat() at netbsd:uvm_pagealloc_strat+0x2d2
> >
> >I've seen some commits from yamt@ about UVM and rb_tree. Maybe this is 
> >fixed
> >now ...
> >
> 
> No luck here. The domu fails to boot, now producing:
> 
> Starting cron.
> Wed Jun 18 11:12:20 EST 2008
> panic: kernel diagnostic assertion "!cv_has_waiters(&bp->b_done)" 
> failed: file "/usr/src/sys/kern/vfs_subr2.c", line 551

I think this one has been fixed today.

-- 
Manuel Bouyer <bouyer%antioche.eu.org@localhost>
     NetBSD: 26 ans d'experience feront toujours la difference
--