Subject: Re: trouble with 10baseT autoselect on wm0
To: None <firstname.lastname@example.org>
From: Steven M. Bellovin <email@example.com>
Date: 04/28/2005 00:12:54
In message <20050428033240.GA16279@panix.com>, Thor Lancelot Simon writes:
>On Wed, Apr 27, 2005 at 11:25:24PM -0400, Thor Lancelot Simon wrote:
>> On Wed, Apr 27, 2005 at 11:08:00PM -0400, Christos Zoulas wrote:
>> > Please file a PR. Drivers that issue a reset to enter and exit promiscuous
>> > mode should be fixed where possible.
>> This actually causes dhclient to *crash the firmware* of some tigon2 cards
>> by resetting them too frequently. Ugly, ugly, ugly. I no longer have a ti
>> card to reproduce the bug for a PR, but I distinctly remember the Happy
>> Fun Symptom.
>I just looked at the ti driver assuming this would be an easy fix. But it
>isn't. The ti driver *already* does the right thing, and it may be that
>the wm driver does, too -- I haven't looked. If only the promiscuous mode
>is changing, ti_ioctl doens't reset the chip.
>But *dhclient* explicitly downs and ups the interface. So it will in fact
>cause this problem with any interface that renegotiates the link under
>these circumstances, whether the driver works right or not. Suck.
I've already posted my long rant on how dhclient should behave; one
more piece of brokenness doesn't surprise me.
I've filed a PR on my wm problem: kern/30078
--Prof. Steven M. Bellovin, http://www.cs.columbia.edu/~smb