Re: [nfsv4] Barry Leiba's Discuss on draft-ietf-nfsv4-rpcrdma-cm-pvt-data-07: (with DISCUSS and COMMENT)

[Chuck Lever <[e-mail]>]

> On Feb 13, 2020, at 9:30 AM, Barry Leiba <[e-mail]> wrote:
> 
> Hi, Chuck, and thanks for the quick response.
> 
>>> — Section 5.2 —
>>> 
>>>  A sender computes the encoded
>>>  value by dividing the buffer size, in octets, by 1024 and subtracting
>>>  one from the result.
>>> 
>>> Is the buffer size necessarily a multiple of 1024?  If so, where is that
>>> specified?  If not, what is the encoded value when the buffer size is, say,
>>> 2000?  Is it zero?  Or one?
>> 
>> Good catch! Buffer sizes are not constrained to 1024-byte length alignment.
>> 
>> Further, if a sender posts, say, a 2032-byte message to a receiver that uses
>> a 2000-byte buffer, a Receive error occurs that typically results in
>> connection loss.
>> 
>> IMO Section 5.2 should instruct the sender to round the actual buffer length
>> down to the nearest 1024-byte multiple before encoding. Would that clarify
>> the issue for you?
> 
> Yes, that would be quite clear.

Looking at Section 5.2 now, it appears that the encoding computation performs
the rounding implicitly. IMO instructing that the sender rounds down first
would be redundant. Here's what I've come up with instead:

   Inline threshold sizes from 1KB to 256KB can be represented in the
   Send Size and Receive Size fields.  These inline threshold values
   provide a pair of 1024-octet-aligned maximum message lengths that
   guarantee Send and Receive operations do not fail due to length
   errors.

   A sender computes the encoded value by dividing its actual buffer
   size, in octets, by 1024 and subtracting one from the result.  A
   receiver decodes this value by performing the inverse set of
   operations: it adds one to the encoded value and then multiplies that
   result by 1024.

   The client uses the smaller of its own send size and the server's
   reported receive size as the client-to-server inline threshold.  The
   server uses the smaller of its own send size and the clients's
   reported receive size as the server-to-client inline threshold.


Perhaps not as clear as one could hope. Comments/suggestions are welcome.


--
Chuck Lever