[PATCH v2] nvme-fabrics: Prevent KATO to be set as 0

Sagi Grimberg sagi at grimberg.me
Wed Sep 6 02:30:20 PDT 2017


>> My bad...
>>
>> So, please clarify it to me: do we want to allow a 0 value for KATO? If
>> yes, no piece of code is needed in the target to stop sending keep alives?
>> The way it "works" today (aka my 1st patch), if we set KATO as 0, we get
>> timeout right after connection. When I first set KATO to 0 in my debug,
>> I was expecting to disable keep alives, not getting timeouts heheh
>>
>>
>> So, the purpose of KATO == 0 is to:
>>
>> (a) Get timeouts right after the connection is complete and keep getting
>> them forever? Or...
>>
>> (b) Disable keep alives at all? (and in this case, we need to inform the
>> target to not send keep alives in case of a 0 valued kato).

Its just a simple way to test that the target does comply to the kato
by not sending it a keep-alive and wait for it to teardown the 
controller, seems useful to keep to have a sanity check that it
doesn't break.

So lets go with (a)



More information about the Linux-nvme mailing list