Ah yes, Rob, but which tables are the correct ones, those in section 9 or those in section 10? Or are you suggesting that both sets of tables be reworked?

 

Thanks.

 

 

We don't really use "fill bytes" in SMP, since the request and response frames are always defined as containing structures that are multiples of 4 bytes.  The only one that does not do so is the ZONED BROADCAST request, but it already defines a "PAD" field on its own to round up to a multiple of 4 bytes.

 

Removing the "fill bytes" concept would be a good letter ballot comment on sas2r14.

--
Rob Elliott, elliott@hp.com
Hewlett-Packard Industry Standard Server Storage Advanced Technology

 

 

---

From: owner-t10@t10.org [mailto:owner-t10@t10.org] On Behalf Of Rich Deglin
Sent: Thursday, January 24, 2008 7:32 PM
To: t10@t10.org
Subject: SMP frame - issue with maximum number of valid data bytes

SAS 2 (r13) section 9.4.3, I think, clearly shows that a maximum length SMP RESPONSE frame consists of the following:

 

1 byte FRAME TYPE

1024 bytes RESPONSE BYTES

3 fill bytes

4 bytes CRC

 

In my reading of this section, this means that an SMP RESPONSE frame with a maximum length RESPONSE BYTES field has NO MORE THAN 1025 valid data bytes plus 3 fill bytes plus 4 CRC bytes = 1032 total bytes. The “header” mentioned in section 10.4.3.2 is not shown here.

 

Section 10.4.3.2 conflicts with this. It states that a maximum length SMP RESPONSE frame consists of the following:

 

1 byte FRAME TYPE

1 byte FUNCTION

1 byte FUNCTION RESULT

1 byte RESPONSE LENGTH

1024 bytes RESPONSE BYTES

0 fill bytes

4 bytes CRC

 

In my reading of this section, this means that an SMP RESPONSE frame with a maximum length RESPONSE BYTES field has EXACTLY 1028 valid data bytes plus 0 fill bytes plus 4 CRC bytes = 1032 total bytes. A 4 byte “header” is mentioned in the text of this section but is not explicitly defined, which I would contend is an oversight in the document.

 

The SMP REQUEST frame descriptions demonstrate the same issue.

 

Please explain the apparent conflict. If there is a real problem here in the spec, is a proposal necessary to correct it?

 

Also note that the same issues exist in the SAS 1.1 spec, but I am guessing that no problems occurred in practice, due to the fact that the lengths of the SAS 1.1 defined SMP REQUESTs and RESPONSEs do not get anywhere near the limits.