Re: Rocket Port update patch

John Baldwin wrote:

> On Tuesday 01 November 2005 07:44 pm, Doug Ambrisko wrote:
> 
>>Scott Long writes:
>>| Doug Ambrisko wrote:
>>| > I've brought in the changes to the driver from the Comtrol web site.
>>| > This greatly improves the RocketPort HW support for those devices.
>>| > The changes are at:
>>| > 	http://www.ambrisko.com/doug/rp_current.patch
>>| >
>>| > I'd like to commit it so things just work out of the box.  I tried to
>>| > maintain all the various FreeBSD & Comtrol changes.
>>| >
>>| > It works here on my previously unsupported card.
>>|
>>| One thing stands out on the brief skim.  Why on earth does the code need
>>| to keep a global count of the number of adapters, and why does it bail
>>| if there are more than 4 adapters?
>>
>>I don't know.  Do we have a contact there to ask questions?  I don't have
>>more then one 16 port adapter to see if there is a problem.
>>
>>That brings up a point, we need to scope the functions and stuff so
>>they don't conflict with other stuff.  I kind-of hate to change
>>all the names to make it easier to merge in changes later.
>>Should we change all the sFlushRxFIFO type stuff to rp_s_flush_rx_fifo
>>and put the globals into the softc's?  Obviously they've put some
>>effort into supporting FreeBSD which is a good thing.  I don't know
>>how things go between them and us.
> 
> 
> It seems odd that they want to use explicit bzero() rather than M_ZERO as 
> well.  Is that to support really old versions of FreeBSD that didn't have 
> M_ZERO?  (I think that's <= 4.2 or some such).
> 

The usual answer to these kinds of questions is that most driver 
developers at hardware companies are more accustomed to Solaris and
especially Linux.  Since the global counts were added somewhat recently,
my guess is that it's more a mirror of the linux way of thinking.

Scott