2003-03-05 15:59:52

by walt

[permalink] [raw]
Subject: 2.4.21-pre5-ac1: Broadcom gigabit ethernet quirk introduced

Hi Alan,

I have an ASUS A7VX8 motherboard with built-in Broadcom gigabit
ethernet chip which has been working perfectly right up through
-pre4-ac7 and now has developed a strange problem starting with
-pre5-ac1.

After rebooting the machine the Broadcom chip appears to be
configured normally, i.e. 'ifconfig eth0' looks perfectly
normal, and the routing table is the same as always.

The problem is that the chip simply doesn't transmit any packets
until I manually do 'ifconfig eth0 down' and 'ifconfig eth0 up'
at which point the chip starts working normally.

The output of 'ifconfig eth0' is identical before and after this
down/up cycle. The internal state of the chip is clearly changed
by this manoeuver, however.

Can you think of any change between -pre4-ac7 and -pre5-ac1 which
might have affected the way the Broadcom chip is initialized?



2003-03-05 17:57:59

by Jeff Garzik

[permalink] [raw]
Subject: Re: 2.4.21-pre5-ac1: Broadcom gigabit ethernet quirk introduced

On Wed, Mar 05, 2003 at 09:37:09AM -0800, walt wrote:
> Hi Alan,
>
> I have an ASUS A7VX8 motherboard with built-in Broadcom gigabit
> ethernet chip which has been working perfectly right up through
> -pre4-ac7 and now has developed a strange problem starting with
> -pre5-ac1.

Is this difference present in plain-jane 2.4.21-pre4 versus 2.4.21-pre5?

Jeff



2003-03-05 19:00:32

by walt

[permalink] [raw]
Subject: Re: 2.4.21-pre5-ac1: Broadcom gigabit ethernet quirk introduced

Jeff Garzik wrote:
> On Wed, Mar 05, 2003 at 09:37:09AM -0800, walt wrote:
>
>>Hi Alan,
>>
>>I have an ASUS A7VX8 motherboard with built-in Broadcom gigabit
>>ethernet chip which has been working perfectly right up through
>>-pre4-ac7 and now has developed a strange problem starting with
>>-pre5-ac1.
>
>
> Is this difference present in plain-jane 2.4.21-pre4 versus 2.4.21-pre5?

Hi Jeff,

I compiled a plain -pre5 just for you ;-)

Yes, the difference apparently has nothing to do with Alan. And (as
another data point) I tried 2.5.64 and I find exactly the same problem
with the Broadcom chip: if I do the ifconfig down/up cycle the chip
starts working normally again. I haven't played with 2.5.x very much
so I can't tell you when the Broadcom problem started there. If you
have a theory to test, I could compile an older 2.5 if you like.