Re: re(4) watchdog timeouts after a short time

[Joe Talbott <josepht@xxxxxxxxxx>]

On Thu, Dec 04, 2008 at 12:09:49PM +0800, Sepherosa Ziehau wrote:
> On Thu, Dec 4, 2008 at 8:04 AM, Joe Talbott <josepht@cstone.net> wrote:
> > On Wed, Dec 03, 2008 at 11:34:31AM +0800, Sepherosa Ziehau wrote:
> >> On Wed, Dec 3, 2008 at 11:24 AM, Joe Talbott <josepht@cstone.net> wrote:
> >> > On Wed, Dec 03, 2008 at 09:16:28AM +0800, Sepherosa Ziehau wrote:
> >> >> On Wed, Dec 3, 2008 at 7:00 AM, Joe Talbott <josepht@cstone.net> wrote:
> >> >> > I upgraded my 2.1-DEVELOPMENT box from 2.1-DEVELOPMENT from several
> >> >> > months ago and my re(4) a RTL8101E card only works for a few seconds
> >> >> > after boot.  After which I get watchdog timeouts and am unable to ping
> >> >> > a remote host.  I was unable to get any information from tcpdump.
> >> >> > Here is the relevant verbose dmesg output and sysctl output.  I have
> >> >> > tried disabling rx/tx csum and vlan{mtu,hwtagging} all to no avail.  Let
> >> >> > me know what further info is needed.
> >> >> >
> >> >> > re0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> mtu 1500
> >> >> >        options=1b<RXCSUM,TXCSUM,VLAN_MTU,VLAN_HWTAGGING>
> >> >> >        inet6 fe80::2a0:d1ff:fe60:25a3%re0 prefixlen 64 scopeid 0x1
> >> >> >        inet x.x.x.x netmask 0xffffffc0 broadcast x.x.x.x
> >> >> >        ether 00:a0:d1:60:25:a3
> >> >> >        media: Ethernet autoselect (100baseTX <full-duplex>)
> >> >> >        status: active
> >> >>
> >> >> Does sysctl hw.re0.imtype=0 help?
> >> >
> >> > I failed to mention I tried that as well.  It did not work.
> >>
> >> Try the attached patch.
> >
> > That works.  I'm not sure if that needs be removed for all cards to I
> > made this patch.  Also I'm not sure if the 8102 cards need this or
> > not.  Thanks for your help on this.
> 
> Thank you for testing.  I think we should not set max read request
> size on all 10/100 only NICs.  Please test following patch:
> http://leaf.dragonflybsd.org/~sephe/re.diff

That works.

Thanks,
Joe