[PATCH] clk: rockchip: fix deadlock possibility in cpuclk

[PATCH] clk: rockchip: fix deadlock possibility in cpuclk

[Heiko Stübner]

Lockdep reported a possible deadlock between the cpuclk lock and for example
the i2c driver.

       CPU0                    CPU1
       ----                    ----
  lock(clk_lock);
                               local_irq_disable();
                               lock(&(&i2c->lock)->rlock);
                               lock(clk_lock);
  <Interrupt>
    lock(&(&i2c->lock)->rlock);

 *** DEADLOCK ***

The generic clock-types of the core ccf already use spin_lock_irqsave when
touching clock registers, so do the same for the cpuclk.

Signed-off-by: Heiko Stuebner <heiko@sntech.de>
---
 drivers/clk/rockchip/clk-cpu.c | 10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)

diff --git a/drivers/clk/rockchip/clk-cpu.c b/drivers/clk/rockchip/clk-cpu.c
index 75c8c45..3ecdf7d 100644
--- a/drivers/clk/rockchip/clk-cpu.c
+++ b/drivers/clk/rockchip/clk-cpu.c
@@ -124,10 +124,11 @@ static int rockchip_cpuclk_pre_rate_change(struct rockchip_cpuclk *cpuclk,
 {
 	const struct rockchip_cpuclk_reg_data *reg_data = cpuclk->reg_data;
 	unsigned long alt_prate, alt_div;
+	unsigned long flags = 0;
 
 	alt_prate = clk_get_rate(cpuclk->alt_parent);
 
-	spin_lock(cpuclk->lock);
+	spin_lock_irqsave(cpuclk->lock, flags);
 
 	/*
 	 * If the old parent clock speed is less than the clock speed
@@ -164,7 +165,7 @@ static int rockchip_cpuclk_pre_rate_change(struct rockchip_cpuclk *cpuclk,
 			cpuclk->reg_base + reg_data->core_reg);
 	}
 
-	spin_unlock(cpuclk->lock);
+	spin_unlock_irqrestore(cpuclk->lock, flags);
 	return 0;
 }
 
@@ -173,6 +174,7 @@ static int rockchip_cpuclk_post_rate_change(struct rockchip_cpuclk *cpuclk,
 {
 	const struct rockchip_cpuclk_reg_data *reg_data = cpuclk->reg_data;
 	const struct rockchip_cpuclk_rate_table *rate;
+	unsigned long flags = 0;
 
 	rate = rockchip_get_cpuclk_settings(cpuclk, ndata->new_rate);
 	if (!rate) {
@@ -181,7 +183,7 @@ static int rockchip_cpuclk_post_rate_change(struct rockchip_cpuclk *cpuclk,
 		return -EINVAL;
 	}
 
-	spin_lock(cpuclk->lock);
+	spin_lock_irqsave(cpuclk->lock, flags);
 
 	if (ndata->old_rate < ndata->new_rate)
 		rockchip_cpuclk_set_dividers(cpuclk, rate);
@@ -201,7 +203,7 @@ static int rockchip_cpuclk_post_rate_change(struct rockchip_cpuclk *cpuclk,
 	if (ndata->old_rate > ndata->new_rate)
 		rockchip_cpuclk_set_dividers(cpuclk, rate);
 
-	spin_unlock(cpuclk->lock);
+	spin_unlock_irqrestore(cpuclk->lock, flags);
 	return 0;
 }
 
-- 
2.1.1

[PATCH] clk: rockchip: fix deadlock possibility in cpuclk

[Doug Anderson]

Heiko,

On Fri, Jan 16, 2015 at 8:52 AM, Heiko St?bner <heiko@sntech.de> wrote:
> Lockdep reported a possible deadlock between the cpuclk lock and for example
> the i2c driver.
>
>        CPU0                    CPU1
>        ----                    ----
>   lock(clk_lock);
>                                local_irq_disable();
>                                lock(&(&i2c->lock)->rlock);
>                                lock(clk_lock);
>   <Interrupt>
>     lock(&(&i2c->lock)->rlock);
>
>  *** DEADLOCK ***
>
> The generic clock-types of the core ccf already use spin_lock_irqsave when
> touching clock registers, so do the same for the cpuclk.
>
> Signed-off-by: Heiko Stuebner <heiko@sntech.de>
> ---
>  drivers/clk/rockchip/clk-cpu.c | 10 ++++++----
>  1 file changed, 6 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/clk/rockchip/clk-cpu.c b/drivers/clk/rockchip/clk-cpu.c
> index 75c8c45..3ecdf7d 100644
> --- a/drivers/clk/rockchip/clk-cpu.c
> +++ b/drivers/clk/rockchip/clk-cpu.c
> @@ -124,10 +124,11 @@ static int rockchip_cpuclk_pre_rate_change(struct rockchip_cpuclk *cpuclk,
>  {
>         const struct rockchip_cpuclk_reg_data *reg_data = cpuclk->reg_data;
>         unsigned long alt_prate, alt_div;
> +       unsigned long flags = 0;

nit: I don't often see flags initted to 0 here when using
spin_lock_irqsave().  I don't think it's needed...

I doubt it really matters though, and this looks fine to me.

Reviewed-by: Doug Anderson <dianders@chromium.org>

-Doug

[PATCH] clk: rockchip: fix deadlock possibility in cpuclk

[Mike Turquette]

Quoting Doug Anderson (2015-01-16 09:11:06)
> Heiko,
> 
> On Fri, Jan 16, 2015 at 8:52 AM, Heiko St?bner <heiko@sntech.de> wrote:
> > Lockdep reported a possible deadlock between the cpuclk lock and for example
> > the i2c driver.
> >
> >        CPU0                    CPU1
> >        ----                    ----
> >   lock(clk_lock);
> >                                local_irq_disable();
> >                                lock(&(&i2c->lock)->rlock);
> >                                lock(clk_lock);
> >   <Interrupt>
> >     lock(&(&i2c->lock)->rlock);
> >
> >  *** DEADLOCK ***
> >
> > The generic clock-types of the core ccf already use spin_lock_irqsave when
> > touching clock registers, so do the same for the cpuclk.
> >
> > Signed-off-by: Heiko Stuebner <heiko@sntech.de>
> > ---
> >  drivers/clk/rockchip/clk-cpu.c | 10 ++++++----
> >  1 file changed, 6 insertions(+), 4 deletions(-)
> >
> > diff --git a/drivers/clk/rockchip/clk-cpu.c b/drivers/clk/rockchip/clk-cpu.c
> > index 75c8c45..3ecdf7d 100644
> > --- a/drivers/clk/rockchip/clk-cpu.c
> > +++ b/drivers/clk/rockchip/clk-cpu.c
> > @@ -124,10 +124,11 @@ static int rockchip_cpuclk_pre_rate_change(struct rockchip_cpuclk *cpuclk,
> >  {
> >         const struct rockchip_cpuclk_reg_data *reg_data = cpuclk->reg_data;
> >         unsigned long alt_prate, alt_div;
> > +       unsigned long flags = 0;
> 
> nit: I don't often see flags initted to 0 here when using
> spin_lock_irqsave().  I don't think it's needed...
> 
> I doubt it really matters though, and this looks fine to me.
> 
> Reviewed-by: Doug Anderson <dianders@chromium.org>

Applied to clk-fixes and removed flags initialization locally.

Regards,
Mike

> 
> -Doug