Re: [PATCH] mcp23s08: support mcp23s17 variant

[Peter Korsgaard <jacmet@sunsite.dk>]

>>>>> "Grant" == Grant Likely <grant.likely@secretlab.ca> writes:

Hi,

 >> -       return (status < 0) ? status : rx[0];
 >> +
 >> +       if (mcp->type == MCP_TYPE_S17) {
 >> +               tx[1] <<= 1;
 >> +
 >> +               status = spi_write_then_read(mcp->spi, tx, sizeof tx, rx, 2);
 >> +               return (status < 0) ? status : (rx[0] | (rx[1] << 8));
 >> +       } else {
 >> +               status = spi_write_then_read(mcp->spi, tx, sizeof tx, rx, 1);
 >> +               return (status < 0) ? status : rx[0];
 >> +       }

 Grant> Rather than checking ->type for every transaction, would a set of
 Grant> callbacks for each type be better?  It would probably have lower
 Grant> overhead and be simpler to maintain.

We could. I didn't do it because the 8bit and 16bit cases are very
similar, and checking a simple integer is presumably a lot faster than
the spi access - But I can rework the patch if you insists.

 >> +++ b/include/linux/spi/mcp23s08.h
 >> @@ -2,21 +2,24 @@
 >>  /* FIXME driver should be able to handle IRQs...  */
 >> 
 >>  struct mcp23s08_chip_info {
 >> -       bool    is_present;             /* true iff populated */
 >> -       u8      pullups;                /* BIT(x) means enable pullup x */
 >> +       bool            is_present;     /* true if populated */
 >> +       unsigned        pullups;        /* BIT(x) means enable pullup x */
 >>  };

 Grant> Unrelated whitespace changes.

I wouldn't call it unrelated. It's to ensure it still lines up after the
s/u8/unsigned/.

-- 
Bye, Peter Korsgaard