Re: [PATCH 07/50] staging: omap-thermal: introduce RMW_BITS macro

[Eduardo Valentin]

Hello Dan,

On 15-03-2013 17:09, Dan Carpenter wrote:
> On Fri, Mar 15, 2013 at 08:59:55AM -0400, Eduardo Valentin wrote:
> > This patch introduce a macro to read, update, write bitfields.
> > It will be specific to bandgap data structures.
> >
> > Signed-off-by: Eduardo Valentin <eduardo.valentin@xxxxxx>
> > ---
> >   drivers/staging/omap-thermal/omap-bandgap.c |  178 +++++++--------------------
> >   1 files changed, 46 insertions(+), 132 deletions(-)
> >
> > diff --git a/drivers/staging/omap-thermal/omap-bandgap.c b/drivers/staging/omap-thermal/omap-bandgap.c
> > index 9f2d7cc..1c1b905 100644
> > --- a/drivers/staging/omap-thermal/omap-bandgap.c
> > +++ b/drivers/staging/omap-thermal/omap-bandgap.c
> > @@ -52,25 +52,29 @@ static void omap_bandgap_writel(struct omap_bandgap *bg_ptr, u32 val, u32 reg)
> >   	writel(val, bg_ptr->base + reg);
> >   }
> >
> > +/* update bits, value will be shifted */
> > +#define RMW_BITS(bg_ptr, id, reg, mask, val)			\
> > +do {								\
> > +	struct temp_sensor_registers *t;			\
> > +	u32 r;							\
> > +								\
> > +	t = bg_ptr->conf->sensors[(id)].registers;		\
> > +	r = omap_bandgap_readl(bg_ptr, t->reg);			\
> > +	r &= ~t->mask;						\
> > +	r |= (val) << __ffs(t->mask);				\
> > +	omap_bandgap_writel(bg_ptr, r, t->reg);			\
> > +} while (0)
> > +
> >   static int omap_bandgap_power(struct omap_bandgap *bg_ptr, bool on)
> >   {
> > -	struct temp_sensor_registers *tsr;
> >   	int i;
> > -	u32 ctrl;
> >
> >   	if (!OMAP_BANDGAP_HAS(bg_ptr, POWER_SWITCH))
> >   		return 0;
> >
> > -	for (i = 0; i < bg_ptr->conf->sensor_count; i++) {
> > -		tsr = bg_ptr->conf->sensors[i].registers;
> > -		ctrl = omap_bandgap_readl(bg_ptr, tsr->temp_sensor_ctrl);
> > -		ctrl &= ~tsr->bgap_tempsoff_mask;
> > +	for (i = 0; i < bg_ptr->conf->sensor_count; i++)
> >   		/* active on 0 */
> > -		ctrl |= !on << __ffs(tsr->bgap_tempsoff_mask);
> > -
> > -		/* write BGAP_TEMPSOFF should be reset to 0 */
> > -		omap_bandgap_writel(bg_ptr, ctrl, tsr->temp_sensor_ctrl);
> > -	}
> > +		RMW_BITS(bg_ptr, i, temp_sensor_ctrl, bgap_tempsoff_mask, !on);
> >
> >   	return 0;
> >   }
> > @@ -78,15 +82,13 @@ static int omap_bandgap_power(struct omap_bandgap *bg_ptr, bool on)
> >   static u32 omap_bandgap_read_temp(struct omap_bandgap *bg_ptr, int id)
> >   {
>
> This patch is fine and it makes it cleaner than before.
>
> But that said, I don't care for the RMW_BITS() very much as a long
> term thing.  If we just used pointers instead of passing the offset
> into the bg_ptr->conf->sensors[] array then everything would be a
> lot cleaner.
>
> In other words, instead of this:
>
> static u32 omap_bandgap_read_temp(struct omap_bandgap *bg_ptr, int id)
>
> We would have:
>
> static u32 omap_bandgap_read_temp(struct omap_bandgap *bg_ptr,
> 				  struct temp_sensor_registers *tsr)

I see. That will require a change in the whole driver though. If you 
see, the driver as of today uses the former approach, not only for 
read_temp or rmw_bits.

> If you have the pointer then it's easy write RMW_BITS() as a
> function.
>
> static void rmw_bits(struct omap_bandgap *bg_ptr, u32 reg, u32 mask, u32 val)
> {
> 	u32 r;
>
> 	r = omap_bandgap_readl(bg_ptr, reg);
> 	r &= ~mask;
> 	r |= val << __ffs(mask);
> 	omap_bandgap_writel(bg_ptr, r, reg);
> }
>
> It's called like:
>
> 	rmw_bits(bg_ptr, tsr->bgap_mask_ctrl, tsr->mask_freeze_mask, 1);

This is nice, but it will require fetching tsr from .conf before every 
call o rmw_bits. And for that you need the sensor index.

> regards,
> dan carpenter

Thanks for your time reviewing this patch and suggesting improvements.

>

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/