Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 20 Mar 2006 16:11:59 +0900
From:      Pyun YongHyeon <pyunyh@gmail.com>
To:        Frank Behrens <frank@pinky.sax.de>
Cc:        freebsd-current@freebsd.org
Subject:   Re: call for sk(4) testers
Message-ID:  <20060320071159.GA76305@cdnetworks.co.kr>
In-Reply-To: <200603200626.k2K6Qq2M029733@pinky.frank-behrens.de>
References:  <200603171426.k2HEQJ9L085859@pinky.frank-behrens.de> <200603200626.k2K6Qq2M029733@pinky.frank-behrens.de>

next in thread | previous in thread | raw e-mail | index | archive | help
On Mon, Mar 20, 2006 at 07:26:48AM +0100, Frank Behrens wrote:
 > Pyun YongHyeon <pyunyh@gmail.com> wrote on 20 Mar 2006 10:36:
 > > Thanks for your report.
 > > If you find any unusual things related with sk(4) please let me know.
 > 
 > Unfortunalety I must report an issue. Yesterday I had after several 
 > weeks of fine running:
 > Mar 19 17:59:08 <kern.crit> moon kernel: sk0: watchdog timeout
 > Mar 19 17:59:08 <kern.notice> moon kernel: sk0: link state changed to DOWN
 > 

Stock sk(4) had a flaw on sending TX command to NIC due to hardware
related races. Rev. 1.90 of if_sk.c tried to fix it by keep resending
the start TX command if driver detects pending packets to be
transmitted. Since the check is done in interrupt handler it would
fail to detect the stuck condition if the first TX command was lost.
I modified the driver to enable a TX polling timer to reissue TX
command periodically as stated in SK NET GENESIS data sheet.
New driver is available at:
http://people.freebsd.org/~yongari/sk/if_sk.c
http://people.freebsd.org/~yongari/sk/if_skreg.h

 > The problem is not the watchdog timeout message itself, but that the 

No. The watchdog timeout message is serious one. It wouldn't recover
from its stuck state without manual interface down/up procedure.

 > links goes down. Fortunately I have in my crontab still a "safety 
 > belt", calling every 12 minutes
 > (ifconfig sk0 | fgrep active >/dev/null) || (ifconfig sk0 down; ifconfig sk0 up; ifconfig sk0)
 > 
 > So the system could recover without user interaction:
 > Mar 19 18:12:01 <kern.notice> moon kernel: sk0: link state changed to UP
 > Mar 19 18:12:01 <kern.crit> moon kernel: sk0: phy failed to come ready
 > 

Yup, this is one thing I'd like to fix.
>From time to time I saw the PHY message during sk(4) module load but
it seems to work correctly in spite of the dereadful message(I think
the same thing happens on stock driver too). So I guess you may
safely ignore that message.

If you still see "watchdog timeout message" please let me know.


-- 
Regards,
Pyun YongHyeon



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20060320071159.GA76305>