Re: Broadcom 4400 driver bm44

Re: Broadcom 4400 driver bm44

[Pekka Pietikainen]

On Wed, Jul 07, 2004 at 02:21:49PM -0400, James Hubbard wrote:
> I have two apps (AppA and AppB) that communicate via multicast.   The
> initial discovery process allows the apps to find each other.  The
> apps send hearbeats periodically to advertise that they are on the
> network.  After a period of time,  AppB can't see AppA.   I would say
> that AppB stops seeing AppA after the initial discovery.  I say this
> If I use this command: 
> ifconfig eth0 allmulti
> AppB receives all of the multicast data and can receive data from
> AppA.  This in turn  causes a problem where standard communication via
> TCP does not function.  Pings, ssh, etc do not work. If I disable
> allmulti. All other communication works except for multicast.

Hiya

I'll try to reproduce the problem. It's entirely possible that there's 
something wrong in the multicast filter code. What would be helpful is
some sort of quick testcase that shows this problem. I tried
a minimal multicast test program (mcast.py from the python distribution)
and that worked ok. ifconfig eth1 allmulti did cause the connection
to my home box to hang, so that I can reproduce at least ;) ;) ;)

Re: Broadcom 4400 driver bm44

[Pekka Pietikainen]

On Thu, Jul 08, 2004 at 10:00:26AM -0400, James Hubbard wrote:
> If I can make some time I'll look into writing a test case here also. 
> I just don't know when it will be though.  I've got the source for the
> net comms layer for the apps.  Perhaps I can pull out some code that
> may demonstrate the error.
> 
> Thanks for the help.
Ok, after some digging around ALLMULTI breaking is not a bug, the
vendor-provided driver seems to break in similar ways, at least on 2.6,
so that's a feature unless they fix it in theirs so I can copy the fix :-) 

I put a lots'o'debugging + some bit-flipping reordering to look
more like bcm4400 version at http://www.ee.oulu.fi/~pp/b44-test.tgz.
Probably won't fix the problem, but it's worth a shot I suppose.

Should compile with 

make -C /lib/modules/<version>/build modules SUBDIRS=$PWD

As a workaround I suppose ifconfig eth1 promisc might do the trick too.

Re: Broadcom 4400 driver bm44

[James Hubbard]

On Thu, 8 Jul 2004 23:43:46 +0300, Pekka Pietikainen <pp@ee.oulu.fi> wrote:
> Ok, after some digging around ALLMULTI breaking is not a bug, the
> vendor-provided driver seems to break in similar ways, at least on 2.6,
> so that's a feature unless they fix it in theirs so I can copy the fix :-)
> 
> I put a lots'o'debugging + some bit-flipping reordering to look
> more like bcm4400 version at http://www.ee.oulu.fi/~pp/b44-test.tgz.
> Probably won't fix the problem, but it's worth a shot I suppose.
> 
> Should compile with
> 
> make -C /lib/modules/<version>/build modules SUBDIRS=$PWD
> 
> As a workaround I suppose ifconfig eth1 promisc might do the trick too.
> 

Hello,

Just to let you know, I've not forgotten about this problem.  I've got
another Dell laptop  that I can reproduce it on.  I'm going to try to
get the stock 2.6.7 kernel on it and try it out.  Hopefully I can get
you some more information by the end of the week.

James


-- 
James Hubbard
http://www.mcs.uvawise.edu/~jhubbard