[PATCH v3] clk_mux: Fix set_parent doing the wrong thing when INDEX_BIT && index >= 3

Mike Turquette mturquette at linaro.org
Wed Nov 19 14:04:49 PST 2014


Quoting Hans de Goede (2014-11-19 05:48:59)
> If CLK_MUX_INDEX_BIT is set, then each bit turns on / off a single parent,
> so theoretically multiple parents could be enabled at the same time, but in
> practice only one bit should ever be 1. So to select parent 0, set
> the register (*) to 0x01, to select parent 1 set it 0x02, parent 2, 0x04,
> parent 3, 0x08, etc.
> 
> But the current code does:
> 
>                 if (mux->flags & CLK_MUX_INDEX_BIT)
>                         index = (1 << ffs(index));
> 
> Which means that:
> 
> For an input index of 0, ffs returns 0, so we set the register
> to 0x01, ok.
> 
> For an input index of 1, ffs returns 1, so we set the register
> to 0x02, ok.
> 
> For an input index of 2, ffs returns 2, so we set the register
> to 0x04, ok.
> 
> For an input index of 3, ffs returns 1, so we set the register
> to 0x02, not good!
> 
> The code should simply be:
> 
>                 if (mux->flags & CLK_MUX_INDEX_BIT)
>                         index = 1 << index;
> 
> Which always does the right thing, this commit fixes this.
> 
> Signed-off-by: Hans de Goede <hdegoede at redhat.com>

Applied to clk-next towards 3.19. I don't think any merged platforms are
hitting this bug in the field so I've not put it into -fixes.

Thanks,
Mike

> ---
>  drivers/clk/clk-mux.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/clk/clk-mux.c b/drivers/clk/clk-mux.c
> index 4f96ff3..6e1ecf9 100644
> --- a/drivers/clk/clk-mux.c
> +++ b/drivers/clk/clk-mux.c
> @@ -77,7 +77,7 @@ static int clk_mux_set_parent(struct clk_hw *hw, u8 index)
>  
>         else {
>                 if (mux->flags & CLK_MUX_INDEX_BIT)
> -                       index = (1 << ffs(index));
> +                       index = 1 << index;
>  
>                 if (mux->flags & CLK_MUX_INDEX_ONE)
>                         index++;
> -- 
> 2.1.0
> 



More information about the linux-arm-kernel mailing list