bge(4) failure, Dell 12G hardware, BCM5720C

YongHyeon PYUN pyunyh at gmail.com
Mon Mar 19 23:58:46 UTC 2012


On Mon, Mar 19, 2012 at 02:27:11PM -0700, Sean Bruno wrote:
> On Fri, 2012-02-24 at 10:06 -0800, YongHyeon PYUN wrote:
> > On Thu, Feb 23, 2012 at 10:23:23AM -0800, Sean Bruno wrote:
> > > 
> > > > As you see ukphy(4) was attached to bge2 so it may cause various
> > > > issues.
> > > > Is bge2 ASF/IPMI enabled interface?  It seems ASF handling in
> > > > bge(4) causes more trouble on recent controllers.  Unfortunately
> > > > disabling ASF may also trigger other problems like NMI.
> > > > I believe bge(4) should always honor ASF/IMPI firmware instead of
> > > > relying on hw.bge.allow_asf tunable and have to strictly follow
> > > > firmware handshake sequence.  Just ignoring ASF/IMPI firmware seems
> > > > to confuse firmware.   
> > > > Unfortunately all these information is undocumented and fixing it
> > > > requires real hardware access.
> > > 
> > > ASF/IPMI -- I've tried disabling things, but it just fails miserably.  I
> > > can at least get the host to a login via serial console and poke at
> > > things.
> > > 
> > > Do you want me to rig up a test for you on this box?  I suspect I can do
> > > something temporarily in the freebsd cluster with this box.
> > > 
> > 
> > Hmm, I still have to understand what is correct handshake sequence
> > for ASF/IPMI firmware. This handshake may be related with suspend/
> > resume as well as WOL. I'll let you know when I have experimental
> > patch.
> > 
> > > Sean
> > > 
> 
> Hrm, looking at the BCM57XX programming guide, I see that there's a
> section on ASF.  Is that what you're looking for or is there a different
> bit of information you are trying to find?
> 

Yes the public data sheet mentions ASF specific registers. However
what we want to know is communication interface between ASF/IPM
firmware and driver.  This would be done via several undocumented
mailbox registers for each events(driver up/down, WOL
configuration, suspend and resume etc).  bge(4) does not care about
registers that make ASF work.

> Sean
> 
> ref.  http://www.broadcom.com/collateral/pg/57XX-PG105-R.pdf
> 


More information about the freebsd-net mailing list