On Fri, 20 May 2016, Junwang Zhao wrote: > On Thu, May 19, 2016 at 8:08 PM, Sage Weil <sage@xxxxxxxxxxxx> wrote: > > On Thu, 19 May 2016, Junwang Zhao wrote: > >> Hi Sage, Haomai, > >> Please check [1] and give some feedback, I am not sure about > >> 52436ded943b8f567adb584c50b789431b788a85 > >> > >> [1] https://github.com/zhjwpku/ceph/commits/wip-addr-work > > > > Great! I made several suggestions, and I have a question for everyone: > > what should we do if we have a "new" addr (say, for the v2 msgr protocol) > > and we need to encode it for a client that doesn't have the ADDR2 feature? > > Right now we are still stuffing the sockaddr_storage in their with > > whatever sockaddr data we have, but it is indistinguishable from a legacy > > address. Perhaps it sould encode as blank? (e.g., entity_addr_t())? Or > > blank, except with a magic nonce (0xffffffff maybe) so that you can tell > > that it is non-blank but not terribly useful/usable? > > I don't quite get this, can you explain this more specific? I now > consider the first encoded __u32 as the distinguish flag for the client > to identify legacy or not, a little bit confused :( Yeah, it is. Look forward a bit to where we have a new, incompatible wire protocol, and TYPE_MSGR2 = 2 (to go with TYPE_LEGACY = 1). How should we encode a msgr2 addr for a client that doesn't have the new ADDR2 feature? Currently, we'll set the legacy __u32 type = 0, and then encode same nonce and sockaddr, but that makes the msgr2 addr appear to be legacy addr to the legacy client. Instead, I'm suggesting that if type != TYPE_LEGACY and the target doesn't have the ADDR2 feature, we encode type = 0, nonce = -1 (or some other poison placeholder), and an empty sockaddr. That way the legacy code will see an addr that is not blank and also looks unusable (blank sockaddr, weird nonce). > BTW, i have add entity_addrvec_t[1], you may want to have a look. > > [1] https://github.com/zhjwpku/ceph/commits/wip-addr-work A few comments, but looks good! I will work on getting the discussion for the msgr2 protocol going next week, as we'll be needing that sooner rather than later. Thanks! sage -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
