On 11/09/2014 09:26 PM, Jeremy Kerr wrote: > Hi Corey, > > Thanks for the review. > >>> IPMI folks: the IPMI driver could do with a little review, as it's >>> not a conventional BT/KCS/SMI SI, in that the low-level send/recv >>> interface will handle the entire message at once. >> Handling the entire message at once should be fine, as that's what >> this driver level is designed to do for the message handler. That >> part all looks correct. The code itself looks good, but I have a >> couple of high-level comments. >> >> The driver at this level can receive more than one message to handle >> at a time, so it needs some sort of queue. This is to allow multiple >> users and to allow the message handler to send its own commands while >> other commands are going on. You might argue that the queuing should >> be done in ipmi_msghandler, and you would probably be right. > Ah, that's what I'd been assuming was being done - I missed the > xmit_list in the si_intf code. It'd be great if this could be in the > generic msghandler code, otherwise I'd just be duplicating the si_intf > logic. > >> I'll look at doing that. If that is the case, then your NULL check >> for current message should probably be a BUG_ON(). > OK, I'll update this when the msghandler bit is implemented.
I've finally finished that, you can get it at: git://git.code.sf.net/p/openipmi/linux-ipmi using the for-next branch. This is what goes into linux-next. >> Do you need to handle any BMC flags? Particularly incoming events? > Not at this stage - we may in future though. Ok. That will complicate things a bit when they are added, but not very much. Thanks, -corey _______________________________________________ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
