[illumos-Developer] HEADS UP: open source iprb

[Garrett D'Amore]

I'm pretty familiar with iprb, and I *think* we have the same fixes.  I
don't reflash the the EEPROM, but in my experience the EEPROM flashing
code is broken in Solaris anyway -- it doesn't actually write
persistently to the flash -- at least not on the cards I've seen.

What I've done is a different workaround for this problem, which I think
will solve the problem, based on the *BSD code I've read.  (Basically,
we have code to monitor for receive hangs, and we insert a CU_NOP before
ever doing CU_RESUME, which should avoid the known erratum affecting the
PCI bus hang.

That said, if someone has one of these devices and can test my driver,
I'd love to hear the results -- good or bad.

	- Garrett

On Mon, 2010-12-06 at 10:33 -0800, Mike Riley wrote:
> On 12/04/10 17:51, Garrett D'Amore wrote:
> > Yesterday I integrated
> >
> > 435 Need an open iprb(7d)
> >
> > This means that the closed iprb is no longer used, and instead a fully
> > open replacement is used instead.  There should not be any noticeable
> > difference behavior, but if there are any bugs, please don't hesitate to
> > let me know or file a bug in redmine.
> 
> Hi Garrett,
> 
> I just remembered an old bug we fixed in iprb that you undoubtedly could not have made the fix 
> for.  Check out Solaris bug 4727020.
> 
> http://bugs.opensolaris.org/bugdatabase/view_bug.do?bug_id=4727020
> 
> Current iprb devices would not be a problem.  We only had the issue with Rev 12 versions of a 
> particular type.  It took Intel about 6 months to figure it out and provide a firmware fix that 
> the Solaris driver uploads the fix when it sees this revision of device.
> 
> Now this issue never showed up on any OS other than Solaris, and Intel wrote most of the code 
> (if not the entire driver) for every OS including Solaris, so it was surprising that it only 
> showed up on Solaris.  Something different about how we service events in the driver, I guess.
> 
> It might be worth noting that Rev 12 cards might experience a problem with your driver so people 
> know to replace them with another revision if they experience the issues reported in the bug. 
> It isn't clear that your version will have the problem, but if it does you won't have the code 
> fix from Intel that is in the Solaris iprb driver.
> 
> Mike
> 
> _______________________________________________
> Developer mailing list
> Developer at lists.illumos.org
> http://lists.illumos.org/m/listinfo/developer