Open gregjoyce opened 7 years ago
It is definitely not a violation of RFC 5041 to send 0 or any other value. The problem is really Chelsio hardware. I'm working on a fix that will ignore the stag value for a 0B READ REQUEST. That fix is in firmware. But the HW will fail an ingress tagged DDP message with an stag of 0 regardless of the payload length (which is a violation of 5041). But that cannot be fixed with existing hardware.
Bernard, will you set it to 1 in the name of interoperability?
Hang on. This might be fixed in chelsio's hw, but not enabled...
This may be a cxgb4 problem but it's preventing siw-cxgb4 interop. iw_cxgb4 has undergone interop testing with several Intel iwarp implementations so that may be an indication that Intel is setting it to a non-zero value. Does the RFC specify that a 0 length DDP message MUST have a stag of value zero? It may not be a violation to send a non-zero stag.