On Fri, 4 Dec 2015, Thierry Reding wrote:
> On Tue, Nov 03, 2015 at 05:01:46PM -0500, Nicolas Pitre wrote:
> > do_div() must only be used with a u64 dividend.
> >
> > Signed-off-by: Nicolas Pitre <nico at linaro.org>
> >
> > diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/clk/gk20a.c
> > b/drivers/gpu/drm/nouveau/nvkm/subdev/clk/gk20a.c
> > index 254094ab7f..5da2aa8cc3 100644
> > --- a/drivers/gpu/drm/nouveau/nvkm/subdev/clk/gk20a.c
> > +++ b/drivers/gpu/drm/nouveau/nvkm/subdev/clk/gk20a.c
> > @@ -141,9 +141,8 @@ gk20a_pllg_calc_rate(struct gk20a_clk *clk)
> >
> > rate = clk->parent_rate * clk->n;
> > divider = clk->m * pl_to_div[clk->pl];
> > - do_div(rate, divider);
> >
> > - return rate / 2;
> > + return rate / divider / 2;
> > }
> >
> > static int
>
> This causes build breakage on 32-bit ARM.
Funny, because the above _fixes_ build warnings on 32-bit ARM for me.
> I'm also confused by the commit message because the code that I'm
> looking at has u64 rate and u32 divider,
If I look at drivers/gpu/drm/nouveau/nvkm/subdev/clk/gk20a.c
in v4.4-rc3 between line 137 and line 147, I see this:
gk20a_pllg_calc_rate(struct gk20a_clk *clk)
{
u32 rate;
u32 divider;
rate = clk->parent_rate * clk->n;
divider = clk->m * pl_to_div[clk->pl];
do_div(rate, divider);
return rate / 2;
}
So clearly both rate and divider are u32 variables and do_div() should
not be used.
Here's the warning:
drivers/gpu/drm/nouveau/nvkm/subdev/clk/gk20a.c: In function
'gk20a_pllg_calc_rate':
include/asm-generic/div64.h:207:28: warning: comparison of distinct pointer
types lacks a cast
(void)(((typeof((n)) *)0) == ((uint64_t *)0)); \
^
drivers/gpu/drm/nouveau/nvkm/subdev/clk/gk20a.c:144:2: note: in expansion of
macro 'do_div'
do_div(rate, divider);
^
The corresponding line number in vanilla v4.4-rc3 for
include/asm-generic/div64.h is line 43.
Nicolas
