Bhanu Gollapudi wrote: > Hi Joe, > > Thanks a lot for reviewing the patches and providing valuable feedback. > I'll incorporate required changes and resend the patches soon.
Thanks. See more below. > On Thu, 2010-05-13 at 16:30 -0700, Joe Eykholt wrote: >> Joe Eykholt wrote: >>> Joe Eykholt wrote: >>>> Bhanu Gollapudi wrote: >>>>> Unsolicited PRLO is now handled by sending LS_ACC and then >>>>> re-login to the target port. >>>>> >>>>> Signed-off-by: Bhanu Prakash Gollapudi <[email protected]> >>>> From FC-LS-1.2 and 1.3, the PRLO LS_ACC response >>>> should have a parameter response page >>>> with a response code, and a payload length field of 20, >>>> so it's not as simple as using seq_els_rsp_send(). >>>> >>>> I think that's why it did reject before ... it was a cheat. >>>> >>>> Otherwise it looks good. >>> Actually, on further reading, table 3 in FC-LS-3 says that >>> a process login is required for PRLO, so the existing case >> I should've said "a N_Port Login is required for PRLO". >> >>> in fc_rport_recv_els_req() was correct. The requestor >>> should handle a reject with reason unable, and >>> the explanation "Port login required". > > I interpreted "a N_Port Login is required" as a session is required > beteen the initiator and the target. As long as the initiator logged in > (PLOGI, PRLI) the target can send a PRLO. > > As per FC-LS section 6.3 "N_Port Login between two Nx_Ports is complete > when each Nx_Port has received the Service Parameters of the other > Nx_Port. This may be accomplished by either implicit or explicit N_Port > Login" > > So, I interpret that the PRLO can be sent by the target even if it didnt > originate the PLOGI. Please let me know your thoughts. I agree with that interpretation, but if we don't find the rdata, then either the rport was logged out by sending a LOGO, or neither side sent a PLOGI, so the reject we generated in that case was appropriate. Regards, Joe _______________________________________________ devel mailing list [email protected] http://www.open-fcoe.org/mailman/listinfo/devel
