Re: [PATCH v10] i2c-designware: make SDA hold time configurable

[Christian Ruppert <christian.ruppert@abilis.com>]

On Wed, Jul 03, 2013 at 04:44:09PM +0200, Arnd Bergmann wrote:
> On Wednesday 03 July 2013, Christian Ruppert wrote:
> > On Wed, Jul 03, 2013 at 04:20:03PM +0200, Arnd Bergmann wrote:
> > > On Wednesday 03 July 2013, Christian Ruppert wrote:
> > > > On Wed, Jul 03, 2013 at 01:43:11PM +0200, Arnd Bergmann wrote:
> > > > > On Wednesday 26 June 2013, Wolfram Sang wrote:
> > > > > >   On Wed, Jun 26, 2013 at 10:55:06AM +0200, Christian Ruppert wrote:
> > > > > > > This patch makes the SDA hold time configurable through device tree.
> > > > > > > 
> > > > > > > Signed-off-by: Christian Ruppert <christian.ruppert@abilis.com>
> > > > > > > Signed-off-by: Pierrick Hascoet <pierrick.hascoet@abilis.com>
> > > > > > 
> > > > > > Applied to for-next, thanks for keeping at it and providing lots of
> > > > > > useful information. Much appreciated!
> > > > > 
> > > > > Sorry, but I got a regression that I didn't find reported elsewhere
> > > > > so far, even though it breaks a lot of the ARM defconfig builds:
> > > > > 
> > > > > drivers/built-in.o: In function `dw_i2c_probe':
> > > > > /git/arm-soc/drivers/i2c/busses/i2c-designware-platdrv.c:125: undefined reference to `__udivdi3'
> > > > > 
> > > > > I suspect you want something like the change below.
> > > > 
> > > > This looks similar to a patch Vincent Stehle submitted yesterday, see
> > > > https://lkml.org/lkml/2013/7/2/145
> > > 
> > > Thanks for the link. Actually his patch looks wrong to me, because
> > > 
> > >  dev->sda_hold_time = div_u64((u64)ic_clk * ht + 500000, 1000000); 
> > > 
> > > assigns the division remainder to sda_hold_time, not the quotient.
> > 
> > Hrmmm... At least when I tested it this morning on an ARC architecture
> > it worked as intended and returned the quotient. Does that mean we have
> > an issue with this function on ARC? Can anyone who knows these functions
> > better than I comment?
> 
> ARC just uses the generic version of div_u64, which is defined in lib/div64.c.
> 
> I suspect that the division remainder just happens to work well enough for
> you to not cause any run-time error. You could try adding a printk
> in that function to show the values you get on ARC.

That's what I did and they were identical to the original values
calculated with /. I just looked at include/linux/math64.h and found the
following comment:

/**
 * div_u64 - unsigned 64bit divide with 32bit divisor
 *
 * This is the most common 64bit divide and should be used if possible,
 * as many 32bit archs can optimize this variant better than a full 64bit
 * divide.
 */

Although this doesn't explicitly state what the function returns to me
this sounds more like the quotient is returned rather than the remainder?

-- 
  Christian Ruppert              ,          <christian.ruppert@abilis.com>
                                /|
  Tel: +41/(0)22 816 19-42     //|                 3, Chemin du Pré-Fleuri
                             _// | bilis Systems   CH-1228 Plan-les-Ouates