[RFC5040 question] RDMA/siw: received Terminate is reported only as a CM close

Hello,

RFC 5040 Section 5.4 says that the RDMAP layer MUST deliver a
received Terminate Message to the ULP. Section 2.1 defines Delivery
as informing the ULP or consumer that a particular Message is
available for use.

https://www.rfc-editor.org/rfc/rfc5040.html#section-5.4

On Linux master commit
075b74841bd0065a3bda3440873c747938e69b68, siw_proc_terminate()
logs the Terminate layer, error type, and error code, optionally
parses the reported header for debugging, and returns -ECONNRESET:

https://github.com/torvalds/linux/blob/075b74841bd0065a3bda3440873c747938e69b68/drivers/infiniband/sw/siw/siw_qp_rx.c#L887-L957

The receive loop treats that return specially: it suppresses the
normal RDMAP completion and schedules the LLP close:

https://github.com/torvalds/linux/blob/075b74841bd0065a3bda3440873c747938e69b68/drivers/infiniband/sw/siw/siw_qp_rx.c#L1409-L1469

The close worker reports only IW_CM_EVENT_CLOSE with status zero:

https://github.com/torvalds/linux/blob/075b74841bd0065a3bda3440873c747938e69b68/drivers/infiniband/sw/siw/siw_cm.c#L1119-L1129

The CLOSE upcall does not include Terminate-specific information:

https://github.com/torvalds/linux/blob/075b74841bd0065a3bda3440873c747938e69b68/drivers/infiniband/sw/siw/siw_cm.c#L351-L409

Based on static source analysis, the consumer appears to receive only
a generic connection-close indication and flushed work requests. I
could not find a path that identifies the received Terminate to the
consumer or makes its information available for later query.

I have not reproduced this at runtime. Is IW_CM_EVENT_CLOSE considered
sufficient delivery of the Terminate Message under RFC 5040 Section
5.4, or is a Terminate-specific indication expected? Is there another
path exposing the received Terminate that I missed?

I realize that RFC 5040 does not prescribe a particular Linux API, so
I am asking about the intended mapping rather than proposing a
specific fix.

This finding came from an academic source-level RFC consistency review.

Best regards