Re: [Scst-devel] qla2x00tgt does not detect cable disconnect?

[ido b. <id...@gm...>]

The problem I described before causes real security/data integrity issue:
I am connecting initiator A and setting a lun masking group for it. I
expose some luns and see them correctly on initiator A.
Then I disconnect the FC cable from that initiator to the FC switch,
and connect initiator B on the same port.
The result is that I see all luns that I exposed to initiator A on initiator B.

In the qla2xxx port_database I see the new port in the "port name
list" (loop ID 8082), however it seems scst does change the connected
sessions.

Does this scenario supposed to be covered by the qla2xxx/qla2x00tgt drivers?



On Sun, Feb 27, 2011 at 9:20 AM, ido benda <id...@gm...> wrote:
> Thank you for your reply Marc,
>
> The situation is as you described, but according to FCP spec, the
> switch, as a fabric, should send a "RSCN", once a logged in port has
> disappeared from the fabric.
> (similar to the indication when a port is logged in for the first time
> - which we do get).
>
> Thanks,
> Ido
>
> On Fri, Feb 25, 2011 at 1:21 AM, Marc Smith <mar...@mc...> wrote:
>> Oh, I think I see what you mean.
>>
>> SCST_Target---------Link_A-----------FC_Switch---------Link_B-------Initiator
>>
>>
>> And you are taking out 'Link_B' (removing) -- right?
>>
>> So, your SCST_Target HBA never sees a "disconnect" because it never is
>> disconnected. =)
>> Its still plugged into the switch so it still has light. Thats how its
>> supposed to work.
>> Think of it as like an Ethernet switch.
>>
>> Is that what you mean or am I still missing what you're trying to
>> explain. =)
>>
>> On Thu, Feb 24, 2011 at 11:33 AM, ido benda <id...@gm...> wrote:
>>> Hi Marc,
>>>
>>> I think I need to explain my self a bit better:
>>> I am using a rack mounted blade server, so I can not disconnect the
>>> cable from the FC switch to the server HBA.
>>> What I do disconnect, is the cable from the initiator to the FC switch.
>>>
>>> I assumed I will get an event in the qla2xxx/qla2x00tgt driver (which I don't).
>>>
>>> In any case, the output of the link_state and speed does not change
>>> when i disconnect the initiator-FC switch cable:
>>>
>>> [root@f16 ~]# cat /sys/class/fc_host/host12/device/scsi_host/host12/link_state
>>> Link Up - F_Port
>>> [root@f16 ~]# cat /sys/class/fc_host/host12/speed
>>> 8 Gbit
>>>
>>> Ido
>>>
>>> On Thu, Feb 24, 2011 at 5:39 PM, Marc Smith <mar...@mc...> wrote:
>>>> So, when you pull the cable out of the HBA, what do the following look like?
>>>> /sys/class/fc_host/hostX/device/scsi_host/hostX/link_state
>>>> /sys/class/fc_host/hostX/speed
>>>>
>>>> When I have a link on mine, I see stuff like this:
>>>> cat /sys/class/fc_host/host3/device/scsi_host/host3/link_state
>>>> Link Up - Ready
>>>>   Fabric
>>>> cat /sys/class/fc_host/host3/speed
>>>> 4 Gbit
>>>>
>>>> When the cable is disconnected (eg, no light), I see this:
>>>> cat /sys/class/fc_host/host3/device/scsi_host/host3/link_state
>>>> Link Down
>>>> cat /sys/class/fc_host/host3/speed
>>>> unknown
>>>>
>>>>
>>>> What do you see?
>>>>
>>>>
>>>> --Marc
>>>>
>>>>
>>>> On Sun, Feb 20, 2011 at 9:23 AM, ido benda <id...@gm...> wrote:
>>>>> Hi,
>>>>>
>>>>> I have started testing scst + qla2xxt on RHEL6 (before i was using
>>>>> centos 5.5 with qla_isp).
>>>>> it seems that the qla2xxx driver does not detect a disconnection of
>>>>> the FC cable. The same behavior was observed with qla_isp (I tested
>>>>> this with two different FC switches).
>>>>>
>>>>> Is this being tested by anyone? could this be a problem with the HBA
>>>>> firmware/FC switch?
>>>>>
>>>>> any insight will be appreciated.
>>>>>
>>>>> Ido
>>>>>
>>>>> ------------------------------------------------------------------------------
>>>>> The ultimate all-in-one performance toolkit: Intel(R) Parallel Studio XE:
>>>>> Pinpoint memory and threading errors before they happen.
>>>>> Find and fix more than 250 security defects in the development cycle.
>>>>> Locate bottlenecks in serial and parallel code that limit performance.
>>>>> http://p.sf.net/sfu/intel-dev2devfeb
>>>>> _______________________________________________
>>>>> Scst-devel mailing list
>>>>> Scs...@li...
>>>>> https://lists.sourceforge.net/lists/listinfo/scst-devel
>>>>>
>>>>
>>>
>>
>