Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 11 Apr 2007 19:23:19 +0900
From:      Pyun YongHyeon <pyunyh@gmail.com>
To:        Stefan Lambrev <stefan.lambrev@sun-fish.com>
Cc:        freebsd-stable@freebsd.org
Subject:   Re: watchdog network card
Message-ID:  <20070411102319.GC1065@cdnetworks.co.kr>
In-Reply-To: <461CB2D8.7040303@sun-fish.com>
References:  <20060327093011.GA21070@math.jussieu.fr> <200603280747.55047.joao@matik.com.br> <20070410042415.GA50510@duncan.reilly.home> <200704100954.20900.joao@matik.com.br> <461CA204.9030700@sun-fish.com> <20070411090617.GB1065@cdnetworks.co.kr> <461CB2D8.7040303@sun-fish.com>

next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, Apr 11, 2007 at 01:05:12PM +0300, Stefan Lambrev wrote:
 > Hi,
 > 
 > -cut-
 > > > I thought getting latest nfe source will help but it doesn't.
 > > > 
 > >  
 > -cut-
 > > > 
 > > > nfe0: watchdog timeout (missed Tx interrupts) -- recovering
 > > > 
 > >
 > >According to the above message, it seems that you use new nfe(4). :-)
 > >  
 > Yes this is is with latest nfe, but I've got watchdog timeouts with 
 > previous version of this driver too.

Without polling(4) on shared interrupt system it's expected one.

 > > > interrupt                          total       rate
 > > > irq1: atkbd0                        6660          0
 > > > irq12: psm0                       153054          0
 > > > irq15: ata1                       778797          1
 > > > irq16: pcm0                      2735727          7
 > > > irq17: skc0                    115669786        296
 > > > irq18: nvidia0                  24802500         63
 > > > irq21: ohci0+                    2418887          6
 > > > irq22: nfe0 ehci0               92319117        236
 > >          ^^^^^^^^^^
 > >You are using shared interrut so it's possible to get occasional
 > >watchdog timeouts. polling(4) should fix your issue here.
 > >  
 > I'll try this, and test :), It wasn't hard to generate generate watchdog 
 > timeout.
 > But isn't polling bad under heavy load and busy CPU(s) ?

Could be. YMMV.
You couldn't also get best performance if interrupt is shared with
other devices.

 > 
 > > > cpu0: timer                    780827522       2000
 > > > Total                         1019712050       2612
 > > > 
 > > > FreeBSD 6.2-STABLE i386
 > > > 
 > > > nfe0: <NVIDIA nForce4 CK804 MCP9 Networking Adapter> port 0xb000-0xb007 
 > > > mem 0xd5000000-0xd5000fff irq 22 at device 10.0 on pci0
 > > > 
 > >
 > >nfe(4) should be teached to use MSI/MSI-X for PCI-Express/PCI-X based
 > >adapters but it's not done yet.
 > >
 > >  
 > 
 > 

-- 
Regards,
Pyun YongHyeon



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20070411102319.GC1065>