From owner-freebsd-current  Fri Oct 31 06:22:10 1997
Return-Path: <owner-freebsd-current>
Received: (from root@localhost)
          by hub.freebsd.org (8.8.7/8.8.7) id GAA09405
          for current-outgoing; Fri, 31 Oct 1997 06:22:10 -0800 (PST)
          (envelope-from owner-freebsd-current)
Received: from sinbin.demos.su (sinbin.demos.su [194.87.5.31])
          by hub.freebsd.org (8.8.7/8.8.7) with SMTP id GAA09347
          for <current@FreeBSD.ORG>; Fri, 31 Oct 1997 06:21:48 -0800 (PST)
          (envelope-from bag@sinbin.demos.su)
Received: by sinbin.demos.su id RAA02226;
  (8.6.12/D) Fri, 31 Oct 1997 17:20:45 +0300
From: bag@sinbin.demos.su (Alex G. Bulushev)
Message-Id: <199710311420.RAA02226@sinbin.demos.su>
Subject: Re: current/nfs
In-Reply-To: <19971029133643.17087@demos.su> from "Mikhail A. Sokolov" at "Oct 29, 97 01:36:43 pm"
X-ELM-OSV: (Our standard violations) no-mime=1; no-hdr-encoding=1
To: smp@csn.net
Date: Fri, 31 Oct 1997 17:20:45 +0300 (MSK)
Cc: current@FreeBSD.ORG, mishania@demos.su
X-Mailer: ELM [version 2.4ME+ PL32 (25)]
Content-Type: text
Sender: owner-freebsd-current@FreeBSD.ORG
X-Loop: FreeBSD.org
Precedence: bulk

> On Tue, Oct 28, 1997 at 09:16:04AM -0700, Steve Passe wrote:
> # Hi,
> # > One machine runs 19971011-current and 'hangs' pretty regularily, then 
> # > tries to reboot, tries to sync and there hangs. The machine has pretty hard
> # > task to be an nfs server for 5 others with up to 6000 interrupts/second, 
> # > motherboard is intel PR440FX, 2xppro200, there are enough swap/memory space.
> # 
> # be more specific, what messages come out when it hangs/reboots?
> 
> 'Cannot switch to CPU 0' \n 'Syncing disks' (or what's that message..) and
> won't reboot. That's last one I caught yesterday, all other cases were just
> 'Syncing' and there we stale. Now I made it non-SMP :-(, awaiting ... 

now we use identical 4 fbsd boxes with intel pr440fx, each with 2 intel EE/100
100base-T full-duplex, one is nfs server and 3 nfs clients
nfs clients works fine with 3.0-971011 and  971012-SNAP kernel/SMP

nfs server hangs 1-3 times a day with SMP kernel 971011 and 971012
latest src's not tested ...

with nonSMP kernel 971012 it works without troubles:

 5:14ÐÐ  up 3 days,  1:01, 23 users, load averages: 0.47, 0.54, 0.56


> 
> Btw, what's that when it has 2 cpu's, the average load is _always_ >1,3, where
> when the second processor is not enabled, it's _always_ (exept when they 
> start some perl :-)) > 0,60 ?
> 
> It's peer (2.2.2+security fixes) would casually yelp that nfs server is not
> responding, the find it (in a second or so) and finally hang in a frozen 
> state with a 'nfsbiostat panic'.
> 
> 
> #> yell it can't start second CPU, - mp_lock 1000005, abort trap. It will do ok
> #> on 233Mhz with 19971011-current, as well as 970926-SNAP.
>  
> # this is probably the race fixed by:

with 971012-SNAP it works on 200 and 233 MHz, thanks ...

   Alex.

> 
> Thank you.
> 
> # Steve Passe
> 
> -- 
> -mishania, stressed.
>