8xx: i2c-algo-8xx - fixed timeout detection and transmission errors

[Joakim Tjernlund]

> 
> Hi Jocke,
> 
> I changed the force_reinit() to
> 
> // Disable interrupts.
> i2c->i2c_i2cmr = 0;
> i2c->i2c_i2cer = 0xff;
> // Clear enable
> i2c->i2c_i2mod &= ~1;
> // Reset internal state
> iip->iic_rstate = 0;
> iip->iic_tstate = 0;
> 
> This seems to work and is less code than my old force_reinit().
> On the other hand: This kind of CPM timeout will only occur on heavy I2C
> bus conflicts. They should not appear at all. The code will not be executed
> if a slave device does not answer! The original driver is from the year
> 2001 and nobody has mentioned problems of this kind before. Without
> detailed knowledge of the CPM internals I feel much safer using my old
> force_reinit(), which does a complete re-init.

Good, however I think we should try to it simple, the code that is, and avoid
unneded bloat. I would like to see the shorter version in the kernel.

 Jocke