[PATCH 1/3] fsi: occ: Force sequence numbering per OCC

Eddie James eajames at linux.ibm.com
Wed Jul 21 23:27:48 AEST 2021


On Wed, 2021-07-21 at 02:37 +0000, Joel Stanley wrote:
> On Fri, 16 Jul 2021 at 15:19, Eddie James <eajames at linux.ibm.com>
> wrote:
> > Set and increment the sequence number during the submit operation.
> > This prevents sequence number conflicts between different users of
> > the interface. A sequence number conflict may result in a user
> > getting an OCC response meant for a different command. Since the
> > sequence number is now modified, the checksum must be calculated
> > and
> > set before submitting the command.
> > 
> > Signed-off-by: Eddie James <eajames at linux.ibm.com>
> 
> Reviewed-by: Joel Stanley <joel at jms.id.au>
> 
> > @@ -479,11 +483,26 @@ int fsi_occ_submit(struct device *dev, const
> > void *request, size_t req_len,
> >                 return -EINVAL;
> >         }
> > 
> > +       /* Checksum the request, ignoring first byte (sequence
> > number). */
> > +       for (i = 1; i < req_len - 2; ++i)
> > +               checksum += byte_request[i];
> > +
> 
> This could go below, after you've got the sequence number, so the
> checksumming all happens in the same spot?

It definitely could, I had the idea to do the checksumming outside the
mutex in case it took a long time? Probably not worth it though.

> 
> The driver has become a bit of a maze, I can't tell how you're
> deciding what goes in fsi_occ_submit vs occ_write vs occ_putsram. If
> oyu have some ideas on how to simplify it then I would welcome those
> changes.

Well, it doesn't really matter in fsi_occ_submit vs occ_putsram, as the
latter is only called in the former. occ_write wouldn't be used by the
hwmon interface, which is why we're moving some of that to
fsi_occ_submit, to have more in common. Agree it could probably be
organized better but I don't immediately have a good idea how to do
that.

Thanks for the review!
Eddie

> 
> 
> 
> >         mutex_lock(&occ->occ_lock);
> > 
> > -       /* Extract the seq_no from the command (first byte) */
> > -       seq_no = *(const u8 *)request;
> > -       rc = occ_putsram(occ, request, req_len);
> > +       /*
> > +        * Get a sequence number and update the counter. Avoid a
> > sequence
> > +        * number of 0 which would pass the response check below
> > even if the
> > +        * OCC response is uninitialized. Any sequence number the
> > user is
> > +        * trying to send is overwritten since this function is the
> > only common
> > +        * interface to the OCC and therefore the only place we can
> > guarantee
> > +        * unique sequence numbers.
> > +        */
> > +       seq_no = occ->sequence_number++;
> > +       if (!occ->sequence_number)
> > +               occ->sequence_number = 1;
> > +       checksum += seq_no;
> > +
> > +       rc = occ_putsram(occ, request, req_len, seq_no, checksum);
> >         if (rc)
> >                 goto done;



More information about the openbmc mailing list