Re: [PATCH] i2c: designware: Add support for 16bit register access

[Stefan Roese <sr-ynQEQJNshbs@public.gmane.org>]

Hi Ben,

On Wednesday 14 March 2012 10:05:40 Stefan Roese wrote:
> On Wednesday 14 March 2012 09:19:22 Bhupesh SHARMA wrote:
> > > > > +++ b/drivers/i2c/busses/i2c-designware-core.c
> > > > > @@ -164,7 +164,14 @@ static char *abort_sources[] = {
> > > > > 
> > > > >  u32 dw_readl(struct dw_i2c_dev *dev, int offset)
> > > > >  {
> > > > > 
> > > > > -	u32 value = readl(dev->base + offset);
> > > > > +	u32 value;
> > > > > +
> > > > > +	if (dev->access_16bit) {
> > > > > +		value = readw(dev->base + offset) |
> > > > > +			(readw(dev->base + offset + 2) << 16);
> > > > > +	} else {
> > > > > +		value = readl(dev->base + offset);
> > > > > +	}
> > > > 
> > > > We can do away with '{' parenthesis as these are single line
> > > > of code inside both the 'if-else' blocks.
> > > 
> > > It's not a single-line statement. The first block extends spans 2
> > > lines. At
> > > least that how I interpret this CodingStyle recommendation.
> > 
> > ???
> > Breaking a single line of code into two for readability does not
> > make them two separate executable statements.
> > 
> > As per CodingStyle recommendation:
> > Do not unnecessarily use braces where a single statement will do.
> > 
> > if (condition)
> > 
> > 	action();
> > 
> > So, please modify your patch as braces here waste precious screen
> > space and reduce readability.
> 
> I have no strong preferences here. I it helps getting this patch accepted,
> I'll remove those braces.
> 
> > > > >  	if (dev->swab)
> > > > >  	
> > > > >  		return swab32(value);
> > > > > 
> > > > > @@ -177,7 +184,12 @@ void dw_writel(struct dw_i2c_dev *dev, u32 b,
> > > 
> > > int
> > > 
> > > > > offset)
> > > > > 
> > > > >  	if (dev->swab)
> > > > >  	
> > > > >  		b = swab32(b);
> > > > > 
> > > > > -	writel(b, dev->base + offset);
> > > > > +	if (dev->access_16bit) {
> > > > > +		writew((u16)b, dev->base + offset);
> > > > > +		writew((u16)(b >> 16), dev->base + offset + 2);
> > > > > +	} else {
> > > > > +		writel(b, dev->base + offset);
> > > > > +	}
> > > > > 
> > > > >  }
> > > > >  
> > > > >  static u32
> > > > > 
> > > > > @@ -258,6 +270,12 @@ int i2c_dw_init(struct dw_i2c_dev *dev)
> > > > > 
> > > > >  		reg = DW_IC_COMP_TYPE_VALUE;
> > > > >  	
> > > > >  	}
> > > > > 
> > > > > +	/* Configure register access mode 16bit */
> > > > > +	if (reg == (DW_IC_COMP_TYPE_VALUE & 0x0000ffff)) {
> > > > > +		dev->access_16bit = 1;
> > > > 
> > > > Can we use a suitable macro for 0x0000ffff?
> > > 
> > > Hmmm. Wouldn't that make it more complex? 0x0000ffff is easy to
> > > understand. A
> > > marco would "hide" this value. I would prefer to keep the value.
> > 
> > Using a macro doesn't make things 'more complex', but more readable.
> > Magic numbers must be avoided at all cost. A better
> > named MACRO is always better (for anyone else reading the code)
> > than a magic number like 0x0000ffff.
> 
> I really don't share your point of view here. I feel that in this case, the
> number is much better readable than a macro.
> 
> > > > Also, if dev->access_16bit is bool we can simply set:
> > > > 		dev->access_16bit = true;
> > > > 
> > > > more on that below...
> > > > 
> > > > > +		reg = DW_IC_COMP_TYPE_VALUE;
> > > > > +	}
> > > > > +
> > > > > 
> > > > >  	if (reg != DW_IC_COMP_TYPE_VALUE) {
> > > > >  	
> > > > >  		dev_err(dev->dev, "Unknown Synopsys component type: "
> > > > >  		
> > > > >  			"0x%08x\n", reg);
> > > > > 
> > > > > diff --git a/drivers/i2c/busses/i2c-designware-core.h
> > > > > b/drivers/i2c/busses/i2c-designware-core.h
> > > > > index 02d1a2d..f5af101 100644
> > > > > --- a/drivers/i2c/busses/i2c-designware-core.h
> > > > > +++ b/drivers/i2c/busses/i2c-designware-core.h
> > > > > @@ -83,6 +83,7 @@ struct dw_i2c_dev {
> > > > > 
> > > > >  	u32			abort_source;
> > > > >  	int			irq;
> > > > >  	int			swab;
> > > > > 
> > > > > +	int			access_16bit;
> > > > 
> > > > ...
> > > > int?? Probably we are better off with making this as bool.
> > > 
> > > I'm not a big fan of bool's. But I have no strong preference here. My
> > > reasoning here was consistency. As we already have "int swab" for a
> > > similar
> > > issue.
> > 
> > If we have not done it earlier, doesn't mean that we repeat the same
> > mistake again. There is no reason to take access_16bit as an int when a
> > bool will suffice.
> > 
> > This wastes storage and on some platforms (which have real crunch of
> > memory), such saving is critical.
> 
> Again, I have no big problem changing this to bool.
> 
> > > So basically, I would prefer to not make the changes you suggested. But
> > > in the
> > > end its the decision of the maintainer(s).
> > 
> > Linux is a collaborative world and patches can be reviewed by
> > literally anyone :)
> 
> Sure.
> 
> > I am sure the maintainer(s) will find my comments worth adding in your
> > patch..
> 
> Might be. But who is the maintainer of this driver?

Ben, will you accept this patch as is, or should I make some of the suggested 
changes (braces, int->bool) and send a patch-v2 out?

Thanks,
Stefan