Re: Recent 6.0_BETA crash

[Matthew Mondor <mm_lists%pulsar-zone.net@localhost>]

On Thu, 08 Mar 2012 14:04:49 +0300
Aleksey Cheusov <cheusov%tut.by@localhost> wrote:

>  >> I few minutes ago I updated the kernel and modules on my 6.0_BETA
>  >> to the latest netbsd-6 sources and enabled debugging kernel options.
>  >> 
>  >>    options        DEBUG           # expensive debugging checks/support
>  >>    makeoptions    DEBUG="-g"      # compile full symbol table
>  >> 
>  >> Userlevel was not updated.
>  >> After login in xdm the system crashed.
>  >>     at /srv/src_netbsd_current/sys/arch/x86/x86/pmap.c:3326
>  >> 
>  >>     3326            KASSERT(uvm_page_locked_p(pg));
>  >> 
>  >> Stacktrace is below. Any clue?
> 
> > It looks like the same problem as what makes the kernel occasionally panic
> > while running tests. It looks like a race condition causing memory 
> > corrution,
> > but this is hard to track down ...
> 
> On my system this crash is reproducible. At least it repeated three
> times with xdm login.  Does it make sense to send PR?

Other cases where I can reproduce a crash or freeze:

- Using tmpfs a lot, i.e. cvs co src and xsrc and build a full release
  on it (using -j4 on a 4 cores amd64).  The tmpfs shows 13GB free
  before I start, so I'm not sure the issue is related to a memory
  shortage.
- Playing a movie from a DVD using a puffs_cd9660, takes a while but
  eventually crashes

Unfortunately, I couldn't get a crash dump (although I do have a dump
device configured), so I'm considering connecting an RS232 cable to
this MB (it fortunately supports it) and finding again my null modem
cable such that I can instead try to get into ddb...

Sometimes the system just reboots, but most of the time it remains
frozen with the HDD light on until I reset.

I'm also unsure if I should file one PR for each case, at least until I
get more useful information...
-- 
Matt