[PATCH] clk: clk-divider: Fix infinite loop for table divider

[PATCH] clk: clk-divider: Fix infinite loop for table divider

[Fabio Estevam]

From: Fabio Estevam <fabio.estevam@freescale.com>

Since commit e7489693b3 (clk: divider: Optimize clk_divider_bestdiv loop) we 
notice a system hang on mx6qboard.

The cause for the hang is well explained by Sascha Hauer [1]:

"This cannot work. _round_up_table is implemented like this:

static int _round_up_table(const struct clk_div_table *table, int div)
{
        const struct clk_div_table *clkt;
        int up = _get_table_maxdiv(table);

        for (clkt = table; clkt->div; clkt++) {
                if (clkt->div == div)
                        return clkt->div;
                ...
        }
        ...
}

Here when a table entry matches the input div this function will return
exactly the input div. This means _next_div() will always return the
same value and clk_divider_bestdiv() has an infinite loop:

        for (i = 1; i <= maxdiv; i = _next_div(divider, i)) {
                ...
        } 
"
[1] http://marc.info/?l=linux-arm-kernel&m=139940658726253&w=2

Remove the 'return _round_up_table' so that the system can work normally again.

Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
---
 drivers/clk/clk-divider.c | 2 --
 1 file changed, 2 deletions(-)

diff --git a/drivers/clk/clk-divider.c b/drivers/clk/clk-divider.c
index b3c8396..fc3e344 100644
--- a/drivers/clk/clk-divider.c
+++ b/drivers/clk/clk-divider.c
@@ -246,8 +246,6 @@ static int _next_div(struct clk_divider *divider, int div)
 
 	if (divider->flags & CLK_DIVIDER_POWER_OF_TWO)
 		return __roundup_pow_of_two(div);
-	if (divider->table)
-		return _round_up_table(divider->table, div);
 
 	return div;
 }
-- 
1.8.3.2

[PATCH] clk: clk-divider: Fix infinite loop for table divider

[Maxime Coquelin]

Hi Fabio,

On 05/06/2014 10:26 PM, Fabio Estevam wrote:
> From: Fabio Estevam <fabio.estevam@freescale.com>
>
> Since commit e7489693b3 (clk: divider: Optimize clk_divider_bestdiv loop) we
> notice a system hang on mx6qboard.
>
> The cause for the hang is well explained by Sascha Hauer [1]:
>
> "This cannot work. _round_up_table is implemented like this:
>
> static int _round_up_table(const struct clk_div_table *table, int div)
> {
>          const struct clk_div_table *clkt;
>          int up = _get_table_maxdiv(table);
>
>          for (clkt = table; clkt->div; clkt++) {
>                  if (clkt->div == div)
>                          return clkt->div;
>                  ...
>          }
>          ...
> }
>
> Here when a table entry matches the input div this function will return
> exactly the input div. This means _next_div() will always return the
> same value and clk_divider_bestdiv() has an infinite loop:
>
>          for (i = 1; i <= maxdiv; i = _next_div(divider, i)) {
>                  ...
>          }
> "
> [1] http://marc.info/?l=linux-arm-kernel&m=139940658726253&w=2
>
> Remove the 'return _round_up_table' so that the system can work normally again.
>
> Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
> ---
>   drivers/clk/clk-divider.c | 2 --
>   1 file changed, 2 deletions(-)
>
> diff --git a/drivers/clk/clk-divider.c b/drivers/clk/clk-divider.c
> index b3c8396..fc3e344 100644
> --- a/drivers/clk/clk-divider.c
> +++ b/drivers/clk/clk-divider.c
> @@ -246,8 +246,6 @@ static int _next_div(struct clk_divider *divider, int div)
>
>   	if (divider->flags & CLK_DIVIDER_POWER_OF_TWO)
>   		return __roundup_pow_of_two(div);
> -	if (divider->table)
> -		return _round_up_table(divider->table, div);

Couldn't we initialize "up" variable to INT_MAX in _round_up_table() as 
I just proposed in mail thread reporting the regression?

See http://marc.info/?l=linux-arm-kernel&m=139945172905130&w=2

Regards,
Maxime

>
>   	return div;
>   }
>

[PATCH] clk: clk-divider: Fix infinite loop for table divider

[Fabio Estevam]

On Wed, May 7, 2014 at 5:37 AM, Maxime Coquelin <maxime.coquelin@st.com> wrote:

> Couldn't we initialize "up" variable to INT_MAX in _round_up_table() as I
> just proposed in mail thread reporting the regression?

Yes, it works fine. Please submit your version.

Thanks