Latest release of BCM5700 driver for PE2650

Jeff Garzik jgarzik at
Tue Aug 5 13:08:00 CDT 2003

Steve_Boley at wrote:
>>BroadCom is actively supporting tg3 development.  If there are bugs, 
>>please let us know about them.
> Since they are actively supporting, why not merge the best of both and
> go to one.  I was informed by a little birdy flying around that bcom
> has actively engaged a kernel developer to clean up bcm5700 and that 
> is what redhat is supposed to officially support for Dell and broadcom
> adapters in upcoming adv server errata kernels.
> Officially I was told that we support bcm5700 on as 2.1 end of story.
> I'm not going to divulge any names or anything but this is where the
> info of broadcom utilizing kernel developer to clean the bcm and get it
> back into the mainstream kernels at redhat came from.  I know you work 
> for redhat so is that the truth or not.  I voiced concerns that different 
> info keeps getting presented and now this still continues.
> Jeff is broadcom going to integrate the codes and are you that developer
> or am I getting incorrect information?  
> I don't want to start any flames but this is a very confusing topic for 
> customers as well as what is working and what is not working.

Certainly, let me clarify.  Feel free to spread this info to other folks 
in Dell or Red Hat, it is public information (not NDA'd).

Red Hat supports what goes in the "upstream" Linux kernel, published on by Linus Torvalds and Marcelo Tosatti.  For a large number of 
technical reasons, the bcm5700 driver does not meet the criteria for 
being included in the Linux kernel.   For a long time, BroadCom was 
openly hostile to Linux and to cleaning up their driver (note: this has 
changed quite a bit recently).  So, the kernel community wrote its own 
Tigon3 driver, rather than wait years for BroadCom get its act together.

So, Red Hat supports the driver that the kernel community supports, 
which is tg3.  (yes, we support bcm5700, but only on AS2.1 -- and even 
there we recommend people switch to tg3 for new chip support)

As to what is going to happen in the future...  I don't see bcm5700 ever 
being merged into the upstream kernel.   bcm5700 is used in conference 
presentations as an example of how _not_ to write a driver!  If one were 
to go through many iterations of code changes to bcm5700, one would 
produce... tg3 :)  It's a lot of wasted work to clean up a really bad 
driver, just to duplicate support that's already in the kernel.  I would 
be really surprised, too, if BroadCom was going to ditch their 
abstraction layer.

Does that clear up the confusion?  Please, pelt me with more questions, 
if not...


More information about the Linux-PowerEdge mailing list