> On Mar 22, 2017, at 9:09 AM, Sagi Grimberg <sagi@xxxxxxxxxxx> wrote: > > >>> I agree this needs to happen, but turns out you don't have any >>> guarantees of the maximum size of the sq depending on your max_sge >>> parameter. >> >> That's true. However, this is meant to be temporary while I'm >> working out details of the rdma_rw API conversion. More work >> in this area comes in the next series: >> >> http://git.linux-nfs.org/?p=cel/cel-2.6.git;a=log;h=refs/heads/nfsd-rdma-rw-api > > Thanks for the pointer... > >>> I'd recommend having a fall-back shrinked size sq allocation >>> impllemented like srpt does. >> >> Agree it should be done. Would it be OK to wait until the dust >> settles here, or do you think it's a hard requirement for >> accepting this series? > > It isn't and can definitely be added incrementally... Roughly speaking, I think there needs to be an rdma_rw API that assists the ULP with setting its CQ and SQ sizes, since rdma_rw hides the registration mode (one of which, at least, consumes more SQEs than the other). I'd like to introduce one new function call that surfaces the factor used to compute how many additional SQEs that rdma_rw will need. The ULP will invoke it before allocating new Send CQs. I'll try to provide an RFC in the nfsd-rdma-rw-api topic branch. -- Chuck Lever -- To unsubscribe from this list: send the line "unsubscribe linux-nfs" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
