diff mbox series

[v7,14/33] crypto: rockchip: handle reset also in PM

Message ID 20220508185957.3629088-15-clabbe@baylibre.com (mailing list archive)
State New, archived
Headers show
Series crypto: rockchip: permit to pass self-tests | expand

Commit Message

Corentin LABBE May 8, 2022, 6:59 p.m. UTC
reset could be handled by PM functions.

Signed-off-by: Corentin Labbe <clabbe@baylibre.com>
---
 drivers/crypto/rockchip/rk3288_crypto.c | 19 ++++---------------
 1 file changed, 4 insertions(+), 15 deletions(-)

Comments

John Keeping June 20, 2022, 11:04 a.m. UTC | #1
On Sun, May 08, 2022 at 06:59:38PM +0000, Corentin Labbe wrote:
> reset could be handled by PM functions.

Is there any further rationale for this?

After this change there is no longer a guaranteed reset pulse on probe
since the reset control may already be de-asserted.  This is normally
the most important case for a reset as it's the only time when the state
of the hardware is unknown.

The original use of devm_add_action_or_reset() seems a bit weird already
since there doesn't seem to be any need to assert reset when the driver
is unloaded.

> Signed-off-by: Corentin Labbe <clabbe@baylibre.com>
> ---
>  drivers/crypto/rockchip/rk3288_crypto.c | 19 ++++---------------
>  1 file changed, 4 insertions(+), 15 deletions(-)
> 
> diff --git a/drivers/crypto/rockchip/rk3288_crypto.c b/drivers/crypto/rockchip/rk3288_crypto.c
> index d9258b9e71b3..a11a92e1f3fd 100644
> --- a/drivers/crypto/rockchip/rk3288_crypto.c
> +++ b/drivers/crypto/rockchip/rk3288_crypto.c
> @@ -73,6 +73,8 @@ static int rk_crypto_pm_suspend(struct device *dev)
>  {
>  	struct rk_crypto_info *rkdev = dev_get_drvdata(dev);
>  
> +	reset_control_assert(rkdev->rst);
> +
>  	rk_crypto_disable_clk(rkdev);
>  	return 0;
>  }
> @@ -81,6 +83,8 @@ static int rk_crypto_pm_resume(struct device *dev)
>  {
>  	struct rk_crypto_info *rkdev = dev_get_drvdata(dev);
>  
> +	reset_control_deassert(rkdev->rst);
> +
>  	return rk_crypto_enable_clk(rkdev);
>  }
>  
> @@ -222,13 +226,6 @@ static void rk_crypto_unregister(void)
>  	}
>  }
>  
> -static void rk_crypto_action(void *data)
> -{
> -	struct rk_crypto_info *crypto_info = data;
> -
> -	reset_control_assert(crypto_info->rst);
> -}
> -
>  static const struct of_device_id crypto_of_id_table[] = {
>  	{ .compatible = "rockchip,rk3288-crypto" },
>  	{}
> @@ -254,14 +251,6 @@ static int rk_crypto_probe(struct platform_device *pdev)
>  		goto err_crypto;
>  	}
>  
> -	reset_control_assert(crypto_info->rst);
> -	usleep_range(10, 20);
> -	reset_control_deassert(crypto_info->rst);
> -
> -	err = devm_add_action_or_reset(dev, rk_crypto_action, crypto_info);
> -	if (err)
> -		goto err_crypto;
> -
>  	crypto_info->reg = devm_platform_ioremap_resource(pdev, 0);
>  	if (IS_ERR(crypto_info->reg)) {
>  		err = PTR_ERR(crypto_info->reg);
> -- 
> 2.35.1
Corentin LABBE June 21, 2022, 8:05 a.m. UTC | #2
Le Mon, Jun 20, 2022 at 12:04:24PM +0100, John Keeping a écrit :
> On Sun, May 08, 2022 at 06:59:38PM +0000, Corentin Labbe wrote:
> > reset could be handled by PM functions.
> 
> Is there any further rationale for this?
> 
> After this change there is no longer a guaranteed reset pulse on probe
> since the reset control may already be de-asserted.  This is normally
> the most important case for a reset as it's the only time when the state
> of the hardware is unknown.
> 
> The original use of devm_add_action_or_reset() seems a bit weird already
> since there doesn't seem to be any need to assert reset when the driver
> is unloaded.
> 

I am not an hw engineer, so my knowledge on reset is low.
So why not having a reset pulse on probe is a problem ?

Do you mean I must put reset asserted on probe ?
John Keeping June 21, 2022, 1:37 p.m. UTC | #3
On Tue, Jun 21, 2022 at 10:05:54AM +0200, LABBE Corentin wrote:
> Le Mon, Jun 20, 2022 at 12:04:24PM +0100, John Keeping a écrit :
> > On Sun, May 08, 2022 at 06:59:38PM +0000, Corentin Labbe wrote:
> > > reset could be handled by PM functions.
> > 
> > Is there any further rationale for this?
> > 
> > After this change there is no longer a guaranteed reset pulse on probe
> > since the reset control may already be de-asserted.  This is normally
> > the most important case for a reset as it's the only time when the state
> > of the hardware is unknown.
> > 
> > The original use of devm_add_action_or_reset() seems a bit weird already
> > since there doesn't seem to be any need to assert reset when the driver
> > is unloaded.
> > 
> 
> I am not an hw engineer, so my knowledge on reset is low.
> So why not having a reset pulse on probe is a problem ?

The point of the reset is to bring the hardware back to a known state.
Since we don't know what state the hardware will be in following the
bootloader or previous OS, I think the reset in probe is the only place
that it is important.

If this patch isn't fixing anything, I suggest just dropping it.
Corentin LABBE June 30, 2022, 2:51 p.m. UTC | #4
Le Tue, Jun 21, 2022 at 02:37:35PM +0100, John Keeping a écrit :
> On Tue, Jun 21, 2022 at 10:05:54AM +0200, LABBE Corentin wrote:
> > Le Mon, Jun 20, 2022 at 12:04:24PM +0100, John Keeping a écrit :
> > > On Sun, May 08, 2022 at 06:59:38PM +0000, Corentin Labbe wrote:
> > > > reset could be handled by PM functions.
> > > 
> > > Is there any further rationale for this?
> > > 
> > > After this change there is no longer a guaranteed reset pulse on probe
> > > since the reset control may already be de-asserted.  This is normally
> > > the most important case for a reset as it's the only time when the state
> > > of the hardware is unknown.
> > > 
> > > The original use of devm_add_action_or_reset() seems a bit weird already
> > > since there doesn't seem to be any need to assert reset when the driver
> > > is unloaded.
> > > 
> > 
> > I am not an hw engineer, so my knowledge on reset is low.
> > So why not having a reset pulse on probe is a problem ?
> 
> The point of the reset is to bring the hardware back to a known state.
> Since we don't know what state the hardware will be in following the
> bootloader or previous OS, I think the reset in probe is the only place
> that it is important.
> 
> If this patch isn't fixing anything, I suggest just dropping it.

Thanks for the explanation, I will re-add the reset at probe.
diff mbox series

Patch

diff --git a/drivers/crypto/rockchip/rk3288_crypto.c b/drivers/crypto/rockchip/rk3288_crypto.c
index d9258b9e71b3..a11a92e1f3fd 100644
--- a/drivers/crypto/rockchip/rk3288_crypto.c
+++ b/drivers/crypto/rockchip/rk3288_crypto.c
@@ -73,6 +73,8 @@  static int rk_crypto_pm_suspend(struct device *dev)
 {
 	struct rk_crypto_info *rkdev = dev_get_drvdata(dev);
 
+	reset_control_assert(rkdev->rst);
+
 	rk_crypto_disable_clk(rkdev);
 	return 0;
 }
@@ -81,6 +83,8 @@  static int rk_crypto_pm_resume(struct device *dev)
 {
 	struct rk_crypto_info *rkdev = dev_get_drvdata(dev);
 
+	reset_control_deassert(rkdev->rst);
+
 	return rk_crypto_enable_clk(rkdev);
 }
 
@@ -222,13 +226,6 @@  static void rk_crypto_unregister(void)
 	}
 }
 
-static void rk_crypto_action(void *data)
-{
-	struct rk_crypto_info *crypto_info = data;
-
-	reset_control_assert(crypto_info->rst);
-}
-
 static const struct of_device_id crypto_of_id_table[] = {
 	{ .compatible = "rockchip,rk3288-crypto" },
 	{}
@@ -254,14 +251,6 @@  static int rk_crypto_probe(struct platform_device *pdev)
 		goto err_crypto;
 	}
 
-	reset_control_assert(crypto_info->rst);
-	usleep_range(10, 20);
-	reset_control_deassert(crypto_info->rst);
-
-	err = devm_add_action_or_reset(dev, rk_crypto_action, crypto_info);
-	if (err)
-		goto err_crypto;
-
 	crypto_info->reg = devm_platform_ioremap_resource(pdev, 0);
 	if (IS_ERR(crypto_info->reg)) {
 		err = PTR_ERR(crypto_info->reg);