[ogb-discuss] Project proposal - Integration (MFI driver for LSI network chipsets)

[James C. McPherson]

Hi Alan,

Alan DuBoff wrote:
> On Mon, 31 Dec 2007, James C. McPherson wrote:
> 
>> What we have with mfi is the situation where LSI's "oh we're really
>> truly going to give you this source real soon now" has taken, more
>> than 12 months (as far as I'm aware).
>>
>> How can *any* organisation think that this is useful to the people
>> who purchase their kit?
> 
> James,
> 
> I'll eply to your message, there's a bunch of them to wade through in this 
> thread, but you were the one that derived this as you are somehow related 
> to submitting some type of ARC case for the author.
> 
> I did find out about the megasas driver, and it's license.
> 
> I also asked my manager if he would reply to this thread, but I think it 
> is remotely possible that he will, I will pass along what I found out from 
> him.
> 
> Just before Xmas the agreement was received, signed by LSI to get the 
> driver putback, and this was about the time this thread started.

I heard that information as well, but from a back-channel and definitely
not out in the open, as it were. When it comes to taking delivery of the
code from LSI I'm taking a "Doubting Thomas" stance.

> The LSI driver is open source, I do not know exactly what license it is 
> under, but the driver portion is all distributable source.

That's another aspect - I've heard that it *could* be under CDDL, and
that it might be under some different Open license entirely. Again,
until we see the deliverable pieces, it's still a moot point because
from an OpenSolaris or even Solaris point of view, it doesn't exist.


> There are some 
> pieces that are not open, but those are the management utilities, and it 
> is up to Sun/LSI in how those are made available to the community. Since 
> LSI is the maintainer of those sources, they will most likely have the 
> biggest say, but I am not clear on how the roles are defined, or who did 
> the actual work. It is done under contract/nda between Sun/LSI, even 
> though the driver is being open sourced.
> You will see this putback within the next month I'm told, as of just 
> before Xmas, it is final.

Great. I'll keep my eyes peeled for it.

>> Whatever distros are out there will *all* have to deal with the
>> concept and problem of handling changes which move from one driver
>> to another, and how to make that happen in as pain-free a fashion
>> as possible.
> 
> I wonder how that will work for more than one of the same driver in open 
> source? It should be able to handle more than one driver to support a 
> given device also, and the user should be able to select which one they 
> use. Or it should be handled by specific distributions in which drivers 
> they package for a given device, and a distribution could elect to devise 
> a system that would allow for more than one driver for the same device I 
> 'spose. That would give the user an alternative driver, in this case the 
> mfi could be offered along with he megasas driver.

As I've mentioned to you before, I've had some ideas about this
issue and I'm currently putting them together into a discussion
paper which I will be sending out in the next week or two.

I don't see this as an insurmountable problem, just one that we
haven't had to deal with much (if at all) in Solaris-land.



James C. McPherson
--
Senior Kernel Software Engineer, Solaris
Sun Microsystems
http://blogs.sun.com/jmcp	http://www.jmcp.homeunix.com/blog