From mboxrd@z Thu Jan 1 00:00:00 1970 From: aeasi@marvell.com (Arun Easi) Date: Mon, 6 May 2019 15:38:24 -0700 Subject: [PATCH nvme-cli rfc 0/6] Support discovery log change events In-Reply-To: <3b7117cf-db69-01a3-d286-a2f98f3e8f5b@broadcom.com> References: <20190223023257.21227-1-sagi@grimberg.me> <3b524b54-8f13-24f5-4367-34a4d02c59b3@broadcom.com> <5c64286b-3fbf-9dfe-ef8a-3f72e630f11d@broadcom.com> <44f637f2-0c0d-cef4-05b3-c5e67ec907bb@grimberg.me> <9d0e3080-baa7-6f55-2548-097523bb3a83@suse.de> <03163041-9b14-3662-edb8-1d69f20f681f@grimberg.me> <2a08f760-0aeb-d788-3210-72ec9d6eb9bc@grimberg.me> <3b7117cf-db69-01a3-d286-a2f98f3e8f5b@broadcom.com> Message-ID: Hi James, On Fri, 26 Apr 2019, 1:14pm, James Smart wrote: > On 4/26/2019 12:10 PM, Sagi Grimberg wrote: > > > > Given that FC does not need a persistent discovery controller to get the > > event, its fine that it will create a temporary discovery controller. > > Given that FC and IP do different things, we can separate them. But what > > I wanted to understand, is if I export a second subsystem on the same > > FC port, will it generate an FC event? > > Well - FC does have an RSCN notification to cause the rescan of the FC > port, if the FC port as the NVME Discovery attribute, but Christoph > rejected the 3 patches about a year ago. I'll be reposting it. So the > answer should eventually be - yes, it will get an FC event if the > subsystem list at the FC port changes. > So, just wondering, for the case of subsystem addition to an existing FC port, is it recommended or mandated in the spec that the target FC port send out a RSCN? The RSCN should be from the target FC port, not the switch, right? I could not find such a reference in the FC-NVME spec version I have (r1.15). On a related note, from a FC LLDD perspective, then, are they required to invoke nvme_fc_rescan_remoteport() upon RSCN reception? Regards, -Arun