Re: Framing Discussion

[Mohan Parthasarathy <Mohan.Parthasarathy@eng.sun.com>]

> 
> >> The design goal behind the framing discussion
> >> is avoidance of that copy.  In contrast to the
> >> typical case for NICs and TCP/IP, read data
> >> buffers for SCSI data are usually *not*
> >> interchangeable.
> >> 
> >Why are they not interchangeable ? This is an
> >assumption not stated anywhere.  Is there
> >a list of other assumptions that is documented ?
> >
> 
> Typically, the original data buffer is with the application that
> made the SCSI READ request - and this buffer may not have been
> allocated with the constraints that should be ready for a simple
> swapping. Assuming that the constraints are met, it would require
> VM page flipping, which is considered to be an implementation hack.
>

Buffer being allocated by the user application is just one case
where applications are doing raw read. I understand that it
is a problem when buffers are pre-allocated. But is this
the problem space we are addressing here ? Moreover,
page flipping is just one techinque for doing zero copy.

> A protocol level solution to locate the buffer and its offset for
> the incoming TCP segment is probably a better thing to have.
>
My point is it is not needed always. There are at least two cases
i can see here :

1) iscsi layer posts pre-allocated buffers for receiving data
   from network.

2) iscsi layer posts commands and willing to receive buffers
   handed by the transport layer. In this cases, i would assume
   the iscsi layer can easily separate the data belonging to
   the different commands it posted. I don't see why you need
   any other extra mechanism here.


So, what are the assumptions we are making on the iscsi/TCP
architecture ?

-mohan