Re: LOR's, and a panic (ipf NAT related?)

Oleg Bulyzhin wrote:
> On Sun, Jul 22, 2007 at 04:31:37PM +0200, Attilio Rao wrote:
>>  Patrick Lamaiziere wrote:
>>> Le Sat, 21 Jul 2007 20:36:31 -0500,
>>> Eric Anderson <anderson_at_freebsd.org> a ?crit :
>>>> Today, on a -CURRENT from a few days ago (running ULE 3.0), I got a
>>>> panic:
>>>>
>>>> panic: Trying sleep, but thread marked as sleeping prohibited
>>>> cpuid = 0
>>>>
>>  This one should have been fixed in last ULE3.0 revision, could you please 
>>  update your src/sys and see if it goes away?
>>
>>  About the other LORs, you should see in the bz's page if they are alredy 
>>  listes, since it seems I remind at least one of them:
>>  http://sources.zabbadoz.net/freebsd/lor.html
>>
>>
>>  Thanks,
>>  Attilio
>>  _______________________________________________
>>  freebsd-current_at_freebsd.org mailing list
>>  http://lists.freebsd.org/mailman/listinfo/freebsd-current
>>  To unsubscribe, send any mail to "freebsd-current-unsubscribe_at_freebsd.org"
> 
> This panic is not ULE related. It came from wrong sx lock usage inside ipfilter.
> I tried to use recently imported (into -current) ipfilter 4.1.23 and found it
> is almost unusable (at least ipnat): it's very unstable (several hangs or
> deadlocks per day), it does leak memory.
> 
> I've spent some time on fixing ipfilter's bugs, then sent report to ipfilter's
> author. Unfortunately, i failed to get any feedback.
> 
> You can get patch here:
> http://people.freebsd.org/~oleg/patches/ipfilter.r70.diff
> 
> I would not swear it's 100% correct, but it should make ipfilter much more
> stable. It does following:
> - sx locks converted to rwlocks. (this should be done, cause pfil(4) uses
>   rwlocks, which are not sleepable).
> - fixed memory leak inside nat_getnext()
> - fixed several improper checks of array's boundary.
> - added some missing mutex_destroy() calls
> - ipfilter's attaching/detaching procedure changed a bit in order to fix some
>   LORs.
> 


Lost track of this - did it ever get ironed out?


Eric