On 7 Apr 2016, at 13:41, Eric Blake <eblake@...696...> wrote: > On 04/07/2016 03:43 AM, Alex Bligh wrote: >> >> I think "[PATCHv5] docs/proto.md: Clarify SHOULD / MUST / MAY etc" >> is probably ready to merge now. Wouter's points were all addressed >> in v5, and Eric OK'd v5 apart from some issues re NBD_OPT_GO that >> weren't actually introduced by the patch. > > Agreed. I can supply the followup for NBD_OPT_GO, if you'd like. Yes please. >> EB: [Nbd] [PATCH v2] doc: Add NBD_CMD_BLOCK_STATUS extension > > Discussion includes the ability to limit the commands' output, whether > from the client or from the server; whether to pick a single format that > lets dirtiness be exposed without a flag. I suspect that a > proof-of-concept implementation for qemu's needs may be necessary before > this one can be finalized. > >> EB: [Nbd] [PATCH v2] doc: Allow NBD_CMD_FLAG_NO_HOLE during NBD_CMD_WRITE > > Discussion includes that it may be better to add a new NBD_OPT_ for > communicating client preference for default server behavior, rather than > the client having to set the flag on every WRITE. This one should be > easier to come up with v3 than BLOCK_STATUS. I haven't been following these closely, but with both of these it seems to me they might be easier if we had the concept of 'block size' for a given export, in which a pile of other useful things (e.g. maximum 'lookahead' for a new hole, bitmap granularity etc). Making the bitmaps support unaligned reads seems a lot of work for little reason. So that's ... > EB: extension to allow server to advertise min/preferred/max sizing ... this one! -- Alex Bligh
Attachment:
signature.asc
Description: Message signed with OpenPGP using GPGMail