On Tue, 09 Jun 2009 14:23:45 -0400 Miles Nordin <car...@ivy.net> wrote:
> >>>>> "jcm" == James C McPherson <james.mcpher...@sun.com> writes: > > jcm> I get the impression that what you really want is for > jcm> Sun to supply you with a list that matches $arbitraryIHV's > jcm> marketing name to "will work with mpt". > > no, misimpression---I've found some cards that work with mpt already > but mpt is proprietary. > > And I don't think Sun has to do anything at all, only that it's much > harder than I expected to find a card with an open-source driver. > > And in spite of the promise, ``all new code in Solaris will be free,'' > new proprietary drivers are still being committed, and there isn't Sun > hardware for sale where all the drivers are liberated. I've gone over the changelog for the closed part of ON, and I have to go back a long time (_many_ builds) to find a new closed driver. Which ones are you claiming are new and closed? > jcm> As an example, for the mega_sas driver > > I think I was confused the last time around because mega_sas used to > be proprietary, right? I see the source checkin about a year ago. The mega_sas driver in OpenSolaris has _never_ been closed. Also, I'd appreciate it if you could not use proprietary when you actually mean closed source. > also it's x86-only. That's correct. > and works with the more expensive 1078 cards (with the built-in > RAID-on-a-card and the large proprietary firmware blob on the CPU > inside the card). > > I can't find the MegaCli tool source either. Linux is also > missing config tool source, but on Linux I can use a 1068 card that > doesn't need MegaCli because it's a simple controller rather than > RAID-on-a-card. We don't have the MegaCLI source either. Ask LSI for it. > Linux has open drivers for all these cards, and for the old Marvell > SATA card, too. Yes I understand that a lot of Solaris source is > free, but I'm trying to communicate the effect, which is that I'm > constantly looking over my back and feeling like I got screwed or > tricked or otherwise misinvested myself. If I use Ubuntu this does > not happen. I don't have to worry about it at all. I don't like > wasting my life worrying about what's free and what isn't, on such > tiny granularity with so many gotchyas. > > jcm> for the mega_sas driver we have __67__ pci ids, Some of which > jcm> specify the full subvendor and subdevice IDs as > > yeah this is part of what's confusing. mega_sas is stealing cards > away from other drivers by providing more specific matches. so even > though mpt lists fewer tuples, mpt matches more cards: if you expand > the match against that pciids.sourceforge.net that you mentioned, mpt > will cut a huger swath than mega_sas. > > but this whole line of reasoning is flawed because most of > pciids.sourceforge.net is historical: only a couple cards of each kind > are currently manufactured at any point in time, and only a few of > those aren't stupidly priced. so although you offer these lists to > try to create this impression there is some huge chaotic market, if > you actually shop them out, in practice there are just zero, one, or > two cards-of-the-month that everyone's buying. You appear to be ignoring the fact that most of those PCI ids are for versions of the chip bolted onto motherboards or sold as add-in card options by other system vendors. That's where the volume is, and yes, there definitely _is_ volume there. > /etc/driver_aliases is pretty useful! Other OS's don't have something > so easy. The only annoying thing is that it comes into existence only > as a result of executing the installer---you can't find it in > opengrok. I'm embarassed it took me this long to understand that the > pci ID matching information is hidden inside the post-install scripts. > There is an extremely short driver_aliases stub for each architecture > in OpenGrok which functions nicely to fool people like me. :) That's correct, we don't have a "master" copy of driver_aliases in the ON gate - it's a file which is assembled on installation and updated afterwards. James C. McPherson -- Senior Kernel Software Engineer, Solaris Sun Microsystems http://blogs.sun.com/jmcp http://www.jmcp.homeunix.com/blog Kernel Conference Australia - http://au.sun.com/sunnews/events/2009/kernel _______________________________________________ driver-discuss mailing list driver-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/driver-discuss
