Re: ntpd panic but only in a domu

[Sarton O'Brien <bsd-xen%roguewrt.org@localhost>]

Kimi wrote:
> On 28/02/2008, Sarton O'Brien <bsd-xen%roguewrt.org@localhost> wrote:
>   
> > Sarton O'Brien wrote:
> >  > I'm still seeing this but only in a domu and only on a domu with significant
> >  > memory usage (ie. a domu runng slapd/postgress and a domu running
> >  > apache/zope) and using the 'disable kernel' option  doesn't help, if there is
> >  > anything else I can provide, let me know. I can trigger this at will:
> >  >
> >  > panic: uvm_map_pageable: stale map
> >  > Stopped in pid 273.1 (ntpd) at  netbsd:breakpoint+0x4:  popl    %ebp
> >  > db> bt
> >  > breakpoint(c0432597,cc525b18,c0448080,c028cb4e,ca292320) at
> >  > netbsd:breakpoint+0x
> >  > 4
> >  > panic(c041dd73,bb96f000,bb989000,7,1) at netbsd:panic+0x1bf
> >  > uvm_map_pageable(ca292310,bb96f000,bb989000,0,1) at
> >  > netbsd:uvm_map_pageable+0x44
> >  > 7
> >  > uvm_mmap(ca292310,cc525cc8,1a000,5,7) at netbsd:uvm_mmap+0x50d
> >  > sys_mmap(cc4b7620,cc525d00,cc525d28,c0399d15,c038f086) at
> >  > netbsd:sys_mmap+0x342
> >  > syscall(cc525d48,1f,1f,bfbf001f,bba6001f) at netbsd:syscall+0xb8
> >  >
> >
> >
> > Am I the only one seeing this? I've been holding off on domu updates for
> >  about 3 months due to this issue, as it seems timed is definately not a
> >  suitable replacement for ntpd, not at least with dom0 rejecting
> >  broadcasts. I'd feel less lonely with at least a 'me too' :)
> >
> >     
>
> I can not say me too, as I run ntpdate & timed both for DOM0 and
> DOMUs. not sure about HVM yet.
>
>   

ntpdate and timed both work ok. The problem I had with timed was only 
when it was run within dom0, there was some broadcast issue. Running it 
from a domu worked fine but the time keeping on a domu is not as 
accurate IIUC.

> I run a ntpd on the firewall along with timed to service everything on
> my networks.
>
> are you running powerd? I noticed it doesn't work under an AMD64 DOMU,
> powerd complaining about not finding /dev/power
>
> I'll see if I can re-create the same thing.
>
>   

I'm on i386 and I do run powerd. No issues there. That stuff is pretty 
solid.

The problem is occurring when /etc/rc.d/ntpd is invoked, either manually 
or at boot. It doesn't happen in dom0, only a domu ... and only one with 
a significant amount of memory usage.

Every other domu is fine and they are all theoretically identical with 
regards to kernel and userland. I'm not quite sure how to recreate the 
problem from scratch, all I know is two out of four that I have at the 
moment, have this problem.

Sarton