[PATCH 1/1] serial: ns16550: move iomem/ioport init after clock init

[PATCH 1/1] serial: ns16550: move iomem/ioport init after clock init

[Maxim Kochetkov]

While probing ns16550 clock provider may be unavailable and
clk_get() returns -EPROBE_DEFER. Next ns16550_probe() attempts
will be failed too because of dev_request_mem_resource() has already
acquired this resource for the device and returns -EBUSY.
So move resource allocation just after clock init. It will let proper
probe for defered clocks.

Signed-off-by: Maxim Kochetkov <fido_max@inbox.ru>
Signed-off-by: Maxim Kochetkov <m.kochetkov@yadro.com>
---
 drivers/serial/serial_ns16550.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/drivers/serial/serial_ns16550.c b/drivers/serial/serial_ns16550.c
index 16f3576645..6eeb67ca81 100644
--- a/drivers/serial/serial_ns16550.c
+++ b/drivers/serial/serial_ns16550.c
@@ -479,10 +479,6 @@ static int ns16550_probe(struct device *dev)
 
 	priv = xzalloc(sizeof(*priv));
 
-	ret = ns16550_init_iomem(dev, priv);
-	if (ret)
-		ret = ns16550_init_ioport(dev, priv);
-
 	if (ret)
 		return ret;
 
@@ -511,6 +507,10 @@ static int ns16550_probe(struct device *dev)
 		goto err;
 	}
 
+	ret = ns16550_init_iomem(dev, priv);
+	if (ret)
+		ret = ns16550_init_ioport(dev, priv);
+
 	cdev = &priv->cdev;
 	cdev->dev = dev;
 	cdev->tstc = ns16550_tstc;
-- 
2.38.1

Re: [PATCH 1/1] serial: ns16550: move iomem/ioport init after clock init

[Ahmad Fatoum]

Hello Maxim,

On 17.02.23 08:31, Maxim Kochetkov wrote:
> -	ret = ns16550_init_iomem(dev, priv);
> -	if (ret)
> -		ret = ns16550_init_ioport(dev, priv);
> -
>  	if (ret)
>  		return ret;

ret is now uninitialized here.

>  
> @@ -511,6 +507,10 @@ static int ns16550_probe(struct device *dev)
>  		goto err;
>  	}
>  
> +	ret = ns16550_init_iomem(dev, priv);
> +	if (ret)
> +		ret = ns16550_init_ioport(dev, priv);

second ret is never checked.

If you move the second if (ret), patch looks ok to me.
jfyi, If you enable deep probe for your board/platform, clk_get()
would automatically probe the clock driver if unavailable, so you
shouldn't ever see -EPROBE_DEFER.

Cheers,
Ahmad



> +
>  	cdev = &priv->cdev;
>  	cdev->dev = dev;
>  	cdev->tstc = ns16550_tstc;

-- 
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

Re: [PATCH 1/1] serial: ns16550: move iomem/ioport init after clock init

[Maxim Kochetkov]

On 17.02.2023 12:02, Ahmad Fatoum wrote:
> Hello Maxim,
> 
> On 17.02.23 08:31, Maxim Kochetkov wrote:
>> -	ret = ns16550_init_iomem(dev, priv);
>> -	if (ret)
>> -		ret = ns16550_init_ioport(dev, priv);
>> -
>>   	if (ret)
>>   		return ret;
> 
> ret is now uninitialized here.
> 
>>   
>> @@ -511,6 +507,10 @@ static int ns16550_probe(struct device *dev)
>>   		goto err;
>>   	}
>>   
>> +	ret = ns16550_init_iomem(dev, priv);
>> +	if (ret)
>> +		ret = ns16550_init_ioport(dev, priv);
> 
> second ret is never checked.

Sorry about that. I will fix it in v2.
> 
> If you move the second if (ret), patch looks ok to me.
> jfyi, If you enable deep probe for your board/platform, clk_get()
> would automatically probe the clock driver if unavailable, so you
> shouldn't ever see -EPROBE_DEFER.

I tried deep probe and it works fine without any -EPROBE_DEFER. Thanks a 
lot for the tip!

Re: [PATCH 1/1] serial: ns16550: move iomem/ioport init after clock init

[Ahmad Fatoum]

Hello Maxim,

On 17.02.23 15:03, Maxim Kochetkov wrote:
> 
> 
> On 17.02.2023 12:02, Ahmad Fatoum wrote:
>> Hello Maxim,
>>
>> On 17.02.23 08:31, Maxim Kochetkov wrote:
>>> -    ret = ns16550_init_iomem(dev, priv);
>>> -    if (ret)
>>> -        ret = ns16550_init_ioport(dev, priv);
>>> -
>>>       if (ret)
>>>           return ret;
>>
>> ret is now uninitialized here.
>>
>>>   @@ -511,6 +507,10 @@ static int ns16550_probe(struct device *dev)
>>>           goto err;
>>>       }
>>>   +    ret = ns16550_init_iomem(dev, priv);
>>> +    if (ret)
>>> +        ret = ns16550_init_ioport(dev, priv);
>>
>> second ret is never checked.
> 
> Sorry about that. I will fix it in v2.
>>
>> If you move the second if (ret), patch looks ok to me.
>> jfyi, If you enable deep probe for your board/platform, clk_get()
>> would automatically probe the clock driver if unavailable, so you
>> shouldn't ever see -EPROBE_DEFER.
> 
> I tried deep probe and it works fine without any -EPROBE_DEFER. Thanks a lot for the tip!

Nice! It's not enabled globally as some existing boards may need rework
(e.g. you need to place of_device_ensure_probed into board code), but new
boards should perferably always use it and just save the time needed to do
EPROBE_DEFER altogether.

Cheers,
Ahmad

> 

-- 
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |