FreeBSD/EC2 lives!

[Colin Percival]

On 12/20/10 13:21, Collin Forbes wrote:
> the last line I see before the panic is:
> 
>     checking whether mlock is broken... 

Yep, known issue -- this is PR kern/140313.

> I'm a debugging n00b however.  What do you need in terms of a debugging 
> report?  I have files in /var/crash as a result of the two panics at 
> exactly the same place.
> 
> -rw-r--r--  1 root  wheel          2 Dec 20 21:16 bounds
> -rw-------  1 root  wheel     149546 Dec 20 20:45 core.txt.0
> -rw-------  1 root  wheel     140384 Dec 20 21:16 core.txt.1
> -rw-------  1 root  wheel        519 Dec 20 20:45 info.0
> -rw-------  1 root  wheel        518 Dec 20 21:16 info.1
> -rw-r--r--  1 root  wheel          5 Dec 11 21:52 minfree
> -rw-------  1 root  wheel  118460416 Dec 20 20:45 vmcore.0
> -rw-------  1 root  wheel  102764544 Dec 20 21:16 vmcore.1

Can you look in the core.txt.[01] files and make sure that the backtraces
match the backtrace at
  http://www.freebsd.org/cgi/query-pr.cgi?pr=kern/140313

Assuming they're the same, I don't think I need any more data... of course,
if you want to dig through the FreeBSD kernel source code and figure out
why the vm page queue mutex is not owned, that would be great. :-)

(My guess is that the VM code was adjusted to push that lock down (i.e.,
hold it for less time) and the xen code wasn't modified at the same time,
but I haven't had time to investigate this fully yet.)

-- 
Colin Percival
Security Officer, FreeBSD | freebsd.org | The power to serve
Founder / author, Tarsnap | tarsnap.com | Online backups for the truly paranoid