Apart from Garrett I have had no response regarding my proposal to have mfi integrated into OpenSolaris.
Saying some vendor driver may be supplied sometime in the near future (where near obviously has a meaning different to the one I thought it had) is a poor excuse to ignore code that has been developed openly, is functional, and is able to be integrated now. The pain of replacing mfi with a possibly better driver later isn't comparable to not having any driver at all. Buying a brick isn't fun. If there is a final answer as to why mfi cannot be integrated then the RFE I have opened should be closed and this proposal should be rejected with a clear explanation as to why. If the integration of the LSI driver really is going to go ahead then a project to handle that integration should be proposed instead. dlg On 18/12/2007, at 12:34 PM, David Gwynne wrote: > On Mon, Dec 17, 2007 at 03:27:03PM -0800, Garrett D'Amore wrote: >> A bit of background here. >> >> 1) David's driver is really very, very clean. A nice bit of code. > > Thank you, that means a lot to me. > >> 2) LSI has a Solaris driver for this bit of hardware. It is closed >> source, and it includes functionality that no one else can easily >> reproduce because LSI does not give out programming specs. > > My boss has just installed the LSI driver on a machine here, so I > had a > quick look at what it presents to the operating system. Unfortunately > the things I had hoped would be there aren't. There's no management > tools, and I didn't see any FMA. > > I'm curious as to what advantage the LSI driver has over mine. I'd > hoped that LSI would be able to do something much better on Solaris. > >> 3) LSI and Sun are in negotiations of trying to bring the LSI code to >> OpenSolaris (most likely via usr/closed). I believe that this may >> be >> stalled/stymied due to lawyers. (When is involvement of lawyers >> *ever* >> a good thing?) >> >> So, personally I'm torn on this issue. On one hand, I hate the idea >> that LSI/Sun feet dragging can stall the project.... David has a >> solution that works today. On the other hand, the "better" >> (technically) solution is likely to appear at some point in the >> not-distant future, and the pain and suffering associated with >> resolving >> differences in the device path, etc. seems like it would be something >> I'd like to avoid if at all possible. >> >> Right now anyone can download David's code and install it. (David, >> do >> you have pkgadd'able packages set up for this? Maybe we can post >> those >> on a project page somewhere?) I like the idea of getting this >> stuff out >> there to enable folks to play with it now... sort of like frkit or >> Masa >> Murayama's drivers... useful but not yet ready for core ON? > > I've had packages up at http://www.itee.uq.edu.au/~dlg/mfi/ for > months. Last time we checked (which was months ago) there were > over 300 downloads of the package. I've had no complaints or bug > reports since I fixed partial DMA transfers in February. I know > people are using this. A lot of people. > >> If anyone from LSI is watching this, please pipe up. The lack of >> progress on this particular front is particularly frustrating. I >> can't >> imagine what kind of proprietary information LSI would want to >> protect >> by suppressing programming information, but then again, I've not seen >> the specs.... >> >> Btw, this particular problem (lack of LSI technical data) is >> frustrating >> to other OS' besides Solaris. For example, on NetBSD mpt for SAS >> adapters is consigned to rotten write performance due to lack of >> technical data about the write cache. > > I think you can fix that with the caching mode page against the scsi > disk itself. > > Anyway, I am frustrated with how long OpenSolaris has had open code > available to it to improve the system, and sat on their hands. > > dlg > >> >> >> -- Garrett >> >> David Gwynne wrote: >>> Hi, >>> >>> I have written a driver called mfi that supports the LSI MegaRAID >>> SAS >>> and Dell PERC 5 RAID controllers. This hardware currently lacks >>> support in OpenSolaris, but a lot of of the big vendors (it seems >>> like everyone except Sun) are now shipping these boards in their >>> machines. Integrating this driver could open doors for OpenSolaris >>> based distributions on these machines. >>> >>> I currently have a RFE 6556091 open against this code, however, I'd >>> like to see the process of integration done in the open and by the >>> community, hence this project proposal. >>> >>> The code for the driver is available at the following urls: >>> >>> http://www.itee.uq.edu.au/~dlg/mfi/ >>> https://svn.itee.uq.edu.au/repo/mfi/ (svn) >>> https://svn.itee.uq.edu.au/viewvc/mfi/ >>> >>> Feel free to contact me if you need further information. >>> >>> dlg >>> _______________________________________________ >>> driver-discuss mailing list >>> [email protected] >>> http://mail.opensolaris.org/mailman/listinfo/driver-discuss >>> >>> >> > _______________________________________________ > driver-discuss mailing list > [email protected] > http://mail.opensolaris.org/mailman/listinfo/driver-discuss _______________________________________________ driver-discuss mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/driver-discuss
