Re: [PATCH] TWL4030: add function to send PB messages

["Peter 'p2' De Schrijver" <peter.de-schrijver@nokia.com>]

> > @@ -81,6 +82,69 @@ twl4030reg_write(struct twlreg_info *info, unsigned offset, u8 value)
> >  			value, info->base + offset);
> >  }
> >  
> > +static int twl4030_wait_pb_ready(void)
> > +{
> > +
> > +	u8 pb_status;
> > +	int status, timeout = 10;
> > +
> > +	do {
> > +		status = twl4030_i2c_read_u8(TWL4030_MODULE_PM_MASTER,
> > +						&pb_status, 0x14);
> > +		if (status < 0)
> > +			return status;
> > +
> 
> Worth a comment that PB_CFG.BIT(0) == PB_I2C_BUSY ... true if there's
> a word queued for the power bus, but not yet sent.  And that we assume
> no other I2C master is sending such events...
> 

The multi master situation seems inherently racy to me anyway as you
need to write 2 bytes to 2 different registers to send 1 message. Or is there
a way to keep mastership of the bus between transactions ?

> > +	/* Enable I2C access to powerbus */
> > +	status = twl4030_i2c_write_u8(TWL4030_MODULE_PM_MASTER,
> > +					pb_state | (1<<1), 0x14);
> > +	if (status < 0)
> > +		return status;
> > +
> > +	status = twl4030_wait_pb_ready();
> 
> I'd probably combine wait_pb_ready() with this; not that
> this is wrong, but you should only need to set BIT(1) once,
> and there's no need to re-read that byte to test BIT(0).
> 
> Minor point ... I hate needless I/O.  This isn't a critical
> path though.
> 

Indeed. Exactly why I didn't try to optimize on I/O here :)

Cheers,

Peter.

-- 
goa is a state of mind