[Bug 259152] [iscsi] ExpDataSN mismatch in SCSI Response (unable to connect or authenticate to OCI oracle iscsi block devices)

From: <bugzilla-noreply_at_freebsd.org>
Date: Fri, 15 Oct 2021 16:04:48 UTC

John Baldwin <jhb@FreeBSD.org> changed:

           What    |Removed                     |Added
                 CC|                            |jhb@FreeBSD.org,
                   |                            |mav@FreeBSD.org,
                   |                            |trasz@FreeBSD.org

--- Comment #5 from John Baldwin <jhb@FreeBSD.org> ---
Please redo the packet capture with '-s 0'.  The SCSI response PDUs are in the
same packet as the Data-In but can't be seen in WireShark since the captured
packets are truncated.  Given the small number of packets involved, using '-s
0' should be ok without risk of dropping packets.

Note that since there is a Data-In, the ExpDataSN field should be 1 in the SCSI
Response.  It is a count of the Data-In PDUs sent in the reply to the original
request prior to the SCSI Response PDU that closes it:


The error messages in comment #4 suggest a completely garbled connection, e.g.
if there are extra bytes in the stream that cause the client to read PDU
headers at the wrong offset.  'PDU 0x0' means that it thinks the opcode of the
PDU is 0 (NOP-Out).  However, NOP-out PDUs are only sent by the initiator
(client), not the target.

I wonder if OCI is using AHS (which isn't generally supported in FreeBSD's
iSCSI stack)?  A full pcap might help.

Oh, also, the LoginRespone shows 'iSCSIProtocolLevel=Not Understood" in the
reply.  That may be ok though as that just means we can't use features from RFC
7144 I think (reading in RFC 7144 which defines version 2 where as 7143 is
version 1).

You are receiving this mail because:
You are the assignee for the bug.