--- Bernard Metzler, PhD Tech. Leader High Performance I/O, Principal Research Staff IBM Zurich Research Laboratory Saeumerstrasse 4 CH-8803 Rueschlikon, Switzerland +41 44 724 8605 -----"Doug Ledford" <dledford@xxxxxxxxxx> wrote: ----- >To: "Bernard Metzler" <bmt@xxxxxxxxxxxxxx>, >linux-rdma@xxxxxxxxxxxxxxx >From: "Doug Ledford" <dledford@xxxxxxxxxx> >Date: 08/22/2019 04:44PM >Cc: krishna2@xxxxxxxxxxx >Subject: [EXTERNAL] Re: [PATCH] RDMA/siw: Enable SGL's with mixed >memory types > >On Thu, 2019-08-22 at 14:16 +0200, Bernard Metzler wrote: >> This patch enables the transmission of work requests with SGL's >> of mixed types, e.g. kernel buffers and PBL regions referenced >> by same work request. This enables iSER as a kernel client. >> >> Reported-by: Krishnamraju Eraparaju <krishna2@xxxxxxxxxxx> >> Tested-by: Krishnamraju Eraparaju <krishna2@xxxxxxxxxxx> >> Signed-off-by: Bernard Metzler <bmt@xxxxxxxxxxxxxx> > >Hi Bernard, > >Commit subject and message are much better this time. However, it's >rc5 >already, and we *just* merged siw this merge cycle, so I'd rather >have >this in the next -rc pull and not in for-next so that siw "just >works" >across the board on initial release. Your language in a commit >message >makes all the difference in the world in terms of whether or not a >commit should go to for-rc. In this case, you used "Enable >SGL's...". >Enable is new feature language. For the rc cycles, you need Fix >language. Something like this: > >RDMA/siw: Fix SGE element mapping issues > >Most upper layer kernel modules submit WQEs where the SG list entries >are all of a single type. iSER in particular, however, will send us >WQEs with mixed SG types: sge[0] = kernel buffer, sge[1] = PBL >region. >Check and set is_kva on each SG entry individually instead of >assuming >the first SGE type carries through to the last. This fixes iSER over >siw. > >Same patch, but the difference in wording makes a world of difference >in >terms of whether or not Linus will give you the evil eye for sending >it >in an -rc cycle. And really, you didn't care about enabling SGLs >with >mixed memory types. It's not like that's some sort of sought after >feature. It was what was needed to fix siw. So just remember that >in >the future. Fix language for fixes, enable language for features. Makes sense, I'll try hard to adhere to that in the future. >The >difference does matter ;-) > >Please resubmit with a fixed commit message and a Fixes: tag. > >-- >Doug Ledford <dledford@xxxxxxxxxx> > GPG KeyID: B826A3330E572FDD > Fingerprint = AE6B 1BDA 122B 23B4 265B 1274 B826 A333 0E57 2FDD > [attachment "signature.asc" removed by Bernard Metzler/Zurich/IBM]
