clk: tegra: Fix an infinite loop when clock rate is zero
diff mbox series

Message ID 20180921220037.16862-1-decatf@gmail.com
State Deferred
Headers show
Series
  • clk: tegra: Fix an infinite loop when clock rate is zero
Related show

Commit Message

Robert Yang Sept. 21, 2018, 10 p.m. UTC
Calling clk_set_rate or clk_round_rate will lock up the kernel when the
rate is zero. This avoids the infinite loop and uses a slightly more
optimized p divider calculation.

Signed-off-by: ryang <decatf@gmail.com>
---
 drivers/clk/tegra/clk-pll.c | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

Comments

Peter De Schrijver Sept. 24, 2018, 8:06 a.m. UTC | #1
On Fri, Sep 21, 2018 at 06:00:37PM -0400, ryang wrote:
> Calling clk_set_rate or clk_round_rate will lock up the kernel when the
> rate is zero. This avoids the infinite loop and uses a slightly more
> optimized p divider calculation.
> 

Acked-By: Peter De Schrijver <pdeschrijver@nvidia.com>

At some point we should also limit pdiv to its maximum possible value, but
that's not so obvious as we need to take into account PLLs where pdiv is
non-linear.

Peter.

> Signed-off-by: ryang <decatf@gmail.com>
> ---
>  drivers/clk/tegra/clk-pll.c | 5 ++---
>  1 file changed, 2 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/clk/tegra/clk-pll.c b/drivers/clk/tegra/clk-pll.c
> index 830d1c87fa7c..17a058c3bbc1 100644
> --- a/drivers/clk/tegra/clk-pll.c
> +++ b/drivers/clk/tegra/clk-pll.c
> @@ -582,9 +582,8 @@ static int _calc_rate(struct clk_hw *hw, struct tegra_clk_pll_freq_table *cfg,
>  	}
>  
>  	/* Raise VCO to guarantee 0.5% accuracy */
> -	for (cfg->output_rate = rate; cfg->output_rate < 200 * cfreq;
> -	     cfg->output_rate <<= 1)
> -		p_div++;
> +	p_div = rate ? fls((200 * cfreq) / rate) : 0;
> +	cfg->output_rate = rate << p_div;
>  
>  	cfg->m = parent_rate / cfreq;
>  	cfg->n = cfg->output_rate / cfreq;
> -- 
> 2.17.1
>
Thierry Reding Sept. 24, 2018, 11:40 a.m. UTC | #2
On Fri, Sep 21, 2018 at 06:00:37PM -0400, ryang wrote:
> Calling clk_set_rate or clk_round_rate will lock up the kernel when the
> rate is zero. This avoids the infinite loop and uses a slightly more
> optimized p divider calculation.
> 
> Signed-off-by: ryang <decatf@gmail.com>
> ---
>  drivers/clk/tegra/clk-pll.c | 5 ++---
>  1 file changed, 2 insertions(+), 3 deletions(-)

Acked-by: Thierry Reding <treding@nvidia.com>
Stephen Boyd Oct. 1, 2018, 10:08 p.m. UTC | #3
Quoting ryang (2018-09-21 15:00:37)
> Calling clk_set_rate or clk_round_rate will lock up the kernel when the
> rate is zero. This avoids the infinite loop and uses a slightly more
> optimized p divider calculation.
> 
> Signed-off-by: ryang <decatf@gmail.com>

Do you have a more proper name? Or you want the author name to show up as
"ryang"?
Stephen Boyd Oct. 2, 2018, 7:35 a.m. UTC | #4
Quoting Stephen Boyd (2018-10-01 15:08:46)
> Quoting ryang (2018-09-21 15:00:37)
> > Calling clk_set_rate or clk_round_rate will lock up the kernel when the
> > rate is zero. This avoids the infinite loop and uses a slightly more
> > optimized p divider calculation.
> > 
> > Signed-off-by: ryang <decatf@gmail.com>
> 
> Do you have a more proper name? Or you want the author name to show up as
> "ryang"?
> 

I see a v2 on the list, and it looks similar so I'm going to assume this
is superseded now.

Patch
diff mbox series

diff --git a/drivers/clk/tegra/clk-pll.c b/drivers/clk/tegra/clk-pll.c
index 830d1c87fa7c..17a058c3bbc1 100644
--- a/drivers/clk/tegra/clk-pll.c
+++ b/drivers/clk/tegra/clk-pll.c
@@ -582,9 +582,8 @@  static int _calc_rate(struct clk_hw *hw, struct tegra_clk_pll_freq_table *cfg,
 	}
 
 	/* Raise VCO to guarantee 0.5% accuracy */
-	for (cfg->output_rate = rate; cfg->output_rate < 200 * cfreq;
-	     cfg->output_rate <<= 1)
-		p_div++;
+	p_div = rate ? fls((200 * cfreq) / rate) : 0;
+	cfg->output_rate = rate << p_div;
 
 	cfg->m = parent_rate / cfreq;
 	cfg->n = cfg->output_rate / cfreq;