Regarding the transfer mechanism for Large SPDM Message (CHUNK_SEND and CHUNK_GET)

[KwonHyunSoo]

Hi,

According to 10.26.3 Additional chunk transfer requirements of the SPDM v1.2 version 1.2.0 (or 1.2.1), due to retries, an SPDM endpoint might re-send or re-retrieve the current chunk number in the transfer.
However, it seems that libspdm does not consider such re-send mechanism in that the responder returns the error response when receiving the sequence number in SPDM request is not equal to "send_info->chunk_seq_no + 1" as follows:

libspdm/library/spdm_responder_lib/libspdm_rsp_chunk_send_ack.c

Lines 147 to 153 in 09be8c3

	if (spdm_request->chunk_seq_no != send_info->chunk_seq_no + 1
	|| spdm_request->header.param2 != send_info->chunk_handle
	|| spdm_request->chunk_size > calc_max_chunk_size
	|| spdm_request->chunk_size + send_info->chunk_bytes_transferred
	> send_info->large_message_size) {
	status = LIBSPDM_STATUS_INVALID_MSG_FIELD;
	}

How can the responder check whether the current CHUNK_SEND command is re-transmitted or not?

Thanks,
H. Kwon

[jyao1]

libspdm does not support retry so far.
ref: #491

[steven-bellock]

Retries with chunking are a bit more straightforward than other cases since chunking includes the sequence number. Also the SPDM specification does not require that a Responder support retries during chunking, just that

When transferring a large SPDM message, an SPDM endpoint shall be prohibited from transferring a chunk sequence number (that is, ChunkSeqNo) less than the current chunk sequence number.