Re: [PATCH 0/2] spi: Add support for non-blocking synchronous transfers

[Balaji Rao <balajirrao@openmoko.org>]

On Sun, Mar 01, 2009 at 01:49:19AM -0800, David Brownell wrote:
> On Saturday 28 February 2009, Balaji Rao wrote:
> > > That leaves un-answered the question of what to do when
> > > the SPI bus is busy performing some other transfer.  I
> > > looked at your [2/2] patch, and saw it ignoring that very
> > > basic issue ... this new call will just poke at the bus,
> > > trashing any transfer that was ongoing.
> > 
> > We use s3c24xx_gpio as the master, which is a very simple gpio based
> > bitbang. 
> > 
> > Yes, it is with this intention, interrupts are disabled around the
> > actual bitbang code, so that it completes without being interrupted.
> > Doesn't this guarantee atomicity ?
> 
> Atomicity isn't the issue so much as the fact that if the
> bus is in the middle of some transfer to one device,
> your patch lets another device trash that transmission.
> 
> I don't know how many more times I can say that your
> patches introduce DATA CORRUPTION to the system, but
> it's surely not many more times.

Yes, I get the point now. Sorry for not observing it earlier.

- Balaji