cxgbe tom: Permit rcv_nxt mismatches on FIN for iSCSI connections on T6.

[jhb]

Summary

The remote peer might send a FIN in the middle of a burst of data
PDUs. In the case of T6 with data PDU completion moderation, the
driver would not have seen these PDUs since the final PDU in the burst
was never received resulting in a stale rcv_nxt when the FIN is
received.

While here, invert the logic in the condition to be more readable and
always set tp->rcv_nxt from the sequence number in the CPL. This sets
the proper value of rcv_nxt for FINs on connections with data received
but not reported via a CPL (e.g. a partial iSCSI PDU burst interrupted
by a FIN).

Reported by: Jithesh Arakkan @ Chelsio

[jhb]

sys/dev/cxgbe/tom/t4_cpl_io.c

1372

My only question here is if you want this conditional on T6 as the mismatches should only occur on T6 and later due to completion moderation. T5 should in theory still always have a valid rcv_nxt.

[np]

sys/dev/cxgbe/tom/t4_cpl_io.c

1372

Yes, let's keep it as tight as possible. These seq# checks have caught legitimate problems in the past.

[jhb]

Mmm, adding T6 to this started to make it unwieldy, so I flipped the condition to put the the KASSERT in the else clause with an empty if body, but I also noticed that technically we would report the wrong rcv_nxt value for the FIN when we actually had the proper value in the CPL, so I've further refined this to do the check before updating tp->rcv_nxt, and to just update rcv_nxt to the value in the CPL. I'll do some testing before updating the version here.

[jhb]

• Restrict check to T6.
• Invert condition in check.
• Always use seq from CPL as seq # of FIN.

[jhb]

I only did light testing of this with iSCSI, but I tested both "plain" TCP offload as well as iSCSI offload connections.