[PATCH] [OOPS] apbuart.c: Two problems related to grlib_apbuart_configure()

[PATCH] [OOPS] apbuart.c: Two problems related to grlib_apbuart_configure()

[Miguel Ojeda]

Hi,

I found two problems using the Simics' sunfire target, so maybe the
following does not occur in real machines.

1. At grlib_apbuart_configure() in apbuart.c, prop can be NULL if
"clock-frequency" doesn't exist in the OF tree:

        /* Get bus frequency */
        rp = of_find_node_by_path("/");
        rp = of_get_next_child(rp, NULL);
        prop = of_get_property(rp, "clock-frequency", NULL);
        freq_khz = *prop;

2. In addition, apbuart.c does not check if there aren't any ports
configured after calls to grlib_apbuart_configure(), so other oops will
occur if no port was configured (e.g. uart_set_options() because of
port->ops).

In order to solve that, I added a check after both of the calls to
grlib_apbuart_configure().

The patch that I provide below prevents both problems in the Simics'
sunfire target.

Tested against 2.6.33.3. Please review.

Signed-off-by: Miguel Ojeda <miguel.ojeda.sandonis@gmail.com>
---
--- drivers/serial/apbuart.c.orig	2010-04-26 16:48:30.000000000 +0200
+++ drivers/serial/apbuart.c	2010-05-08 23:50:36.786148571 +0200
@@ -525,6 +525,9 @@ static void grlib_apbuart_configure(void
 static int __init apbuart_console_init(void)
 {
 	grlib_apbuart_configure();
+	if (grlib_apbuart_port_nr == 0)
+		return 0;
+
 	register_console(&grlib_apbuart_console);
 	return 0;
 }
@@ -612,6 +615,8 @@ static void grlib_apbuart_configure(void
 	rp = of_find_node_by_path("/");
 	rp = of_get_next_child(rp, NULL);
 	prop = of_get_property(rp, "clock-frequency", NULL);
+	if (prop == NULL)
+		return;
 	freq_khz = *prop;
 
 	line = 0;
@@ -666,6 +671,11 @@ static int __init grlib_apbuart_init(voi
 
 	/* Find all APBUARTS in device the tree and initialize their ports */
 	grlib_apbuart_configure();
+	if (grlib_apbuart_port_nr == 0) {
+		printk(KERN_INFO "Serial: GRLIB APBUART: No ports found.\n",
+			__FILE__);
+		return 0;
+	}
 
 	printk(KERN_INFO "Serial: GRLIB APBUART driver\n");
 

Re: [PATCH] [OOPS] apbuart.c: Two problems related to grlib_apbuart_configure()

[Miguel Ojeda]

On Sat, May 8, 2010 at 11:58 PM, Miguel Ojeda
<miguel.ojeda.sandonis@gmail.com> wrote:
> Hi,
>
> I found two problems using the Simics' sunfire target, so maybe the
> following does not occur in real machines.
>
> 1. At grlib_apbuart_configure() in apbuart.c, prop can be NULL if
> "clock-frequency" doesn't exist in the OF tree:
>
>        /* Get bus frequency */
>        rp = of_find_node_by_path("/");
>        rp = of_get_next_child(rp, NULL);
>        prop = of_get_property(rp, "clock-frequency", NULL);
>        freq_khz = *prop;
>
> 2. In addition, apbuart.c does not check if there aren't any ports
> configured after calls to grlib_apbuart_configure(), so other oops will
> occur if no port was configured (e.g. uart_set_options() because of
> port->ops).
>
> In order to solve that, I added a check after both of the calls to
> grlib_apbuart_configure().
>
> The patch that I provide below prevents both problems in the Simics'
> sunfire target.
>
> Tested against 2.6.33.3. Please review.
>
> Signed-off-by: Miguel Ojeda <miguel.ojeda.sandonis@gmail.com>
> ---
> --- drivers/serial/apbuart.c.orig       2010-04-26 16:48:30.000000000 +0200
> +++ drivers/serial/apbuart.c    2010-05-08 23:50:36.786148571 +0200
> @@ -525,6 +525,9 @@ static void grlib_apbuart_configure(void
>  static int __init apbuart_console_init(void)
>  {
>        grlib_apbuart_configure();
> +       if (grlib_apbuart_port_nr == 0)
> +               return 0;
> +
>        register_console(&grlib_apbuart_console);
>        return 0;
>  }
> @@ -612,6 +615,8 @@ static void grlib_apbuart_configure(void
>        rp = of_find_node_by_path("/");
>        rp = of_get_next_child(rp, NULL);
>        prop = of_get_property(rp, "clock-frequency", NULL);
> +       if (prop == NULL)

grlib_apbuart_port_nr = 0; should be added here or in apbuart.h better.

> +               return;
>        freq_khz = *prop;
>
>        line = 0;
> @@ -666,6 +671,11 @@ static int __init grlib_apbuart_init(voi
>
>        /* Find all APBUARTS in device the tree and initialize their ports */
>        grlib_apbuart_configure();
> +       if (grlib_apbuart_port_nr == 0) {
> +               printk(KERN_INFO "Serial: GRLIB APBUART: No ports found.\n",
> +                       __FILE__);
> +               return 0;
> +       }
>
>        printk(KERN_INFO "Serial: GRLIB APBUART driver\n");
>
>
>

Re: [PATCH] [OOPS] apbuart.c: Two problems related to grlib_apbuart_configure()

[Kristoffer Glembo]

Miguel Ojeda wrote:
> On Sat, May 8, 2010 at 11:58 PM, Miguel Ojeda
> <miguel.ojeda.sandonis@gmail.com> wrote:
>> Hi,
>>
>> I found two problems using the Simics' sunfire target, so maybe the
>> following does not occur in real machines.
>>
>> 1. At grlib_apbuart_configure() in apbuart.c, prop can be NULL if
>> "clock-frequency" doesn't exist in the OF tree:
>>
>>        /* Get bus frequency */
>>        rp = of_find_node_by_path("/");
>>        rp = of_get_next_child(rp, NULL);
>>        prop = of_get_property(rp, "clock-frequency", NULL);
>>        freq_khz = *prop;
>>
>> 2. In addition, apbuart.c does not check if there aren't any ports
>> configured after calls to grlib_apbuart_configure(), so other oops will
>> occur if no port was configured (e.g. uart_set_options() because of
>> port->ops).
>>
>> In order to solve that, I added a check after both of the calls to
>> grlib_apbuart_configure().
>>
>> The patch that I provide below prevents both problems in the Simics'
>> sunfire target.
>>
>> Tested against 2.6.33.3. Please review.
>>
>> Signed-off-by: Miguel Ojeda <miguel.ojeda.sandonis@gmail.com>


Thanks for the patch, it looks fine to me. 

Acked-by: Kristoffer Glembo <kristoffer@gaisler.com>

Re: [PATCH] [OOPS] apbuart.c: Two problems related to grlib_apbuart_configure()

[Miguel Ojeda]

Hi,

Kristoffer, please check this one instead.

Changes:
1. Added the grlib_apbuart_port_nr = 0 line.
2. Returned -ENODEV instead of 0 in the "no ports found branch" in
init(), because if compiled as a module, exit() will try to unregister
things that were not registered. In addition, this will alert the users
that modprobe'd.
3. Deleted the unused __FILE__ argument of the previous patch (oops
sorry :).

Andrew, if Kristoffer agrees, please drop the older patch from -mm and
merge this one instead.

Signed-off-by: Miguel Ojeda <miguel.ojeda.sandonis@gmail.com>
---
--- drivers/serial/apbuart.c.orig	2010-04-26 16:48:30.000000000 +0200
+++ drivers/serial/apbuart.c	2010-05-11 15:21:28.984666230 +0200
@@ -525,6 +525,9 @@ static void grlib_apbuart_configure(void
 static int __init apbuart_console_init(void)
 {
 	grlib_apbuart_configure();
+	if (grlib_apbuart_port_nr == 0)
+		return 0;
+
 	register_console(&grlib_apbuart_console);
 	return 0;
 }
@@ -612,6 +615,10 @@ static void grlib_apbuart_configure(void
 	rp = of_find_node_by_path("/");
 	rp = of_get_next_child(rp, NULL);
 	prop = of_get_property(rp, "clock-frequency", NULL);
+	if (prop == NULL) {
+		grlib_apbuart_port_nr = 0;
+		return;
+	}
 	freq_khz = *prop;
 
 	line = 0;
@@ -666,6 +673,10 @@ static int __init grlib_apbuart_init(voi
 
 	/* Find all APBUARTS in device the tree and initialize their ports */
 	grlib_apbuart_configure();
+	if (grlib_apbuart_port_nr == 0) {
+		printk(KERN_INFO "Serial: GRLIB APBUART: No ports found.\n");
+		return -ENODEV;
+	}
 
 	printk(KERN_INFO "Serial: GRLIB APBUART driver\n");
 

Re: [PATCH] [OOPS] apbuart.c: Two problems related to grlib_apbuart_configure()

[Kristoffer Glembo]

Hi,

Looks finer :-)

Acked-by: Kristoffer Glembo <kristoffer@gaisler.com>


Miguel Ojeda wrote:
> Hi,
> 
> Kristoffer, please check this one instead.
> 
> Changes:
> 1. Added the grlib_apbuart_port_nr = 0 line.
> 2. Returned -ENODEV instead of 0 in the "no ports found branch" in
> init(), because if compiled as a module, exit() will try to unregister
> things that were not registered. In addition, this will alert the users
> that modprobe'd.
> 3. Deleted the unused __FILE__ argument of the previous patch (oops
> sorry :).
> 
> Andrew, if Kristoffer agrees, please drop the older patch from -mm and
> merge this one instead.
> 
> Signed-off-by: Miguel Ojeda <miguel.ojeda.sandonis@gmail.com>
> ---
> --- drivers/serial/apbuart.c.orig	2010-04-26 16:48:30.000000000 +0200
> +++ drivers/serial/apbuart.c	2010-05-11 15:21:28.984666230 +0200
> @@ -525,6 +525,9 @@ static void grlib_apbuart_configure(void
>  static int __init apbuart_console_init(void)
>  {
>  	grlib_apbuart_configure();
> +	if (grlib_apbuart_port_nr == 0)
> +		return 0;
> +
>  	register_console(&grlib_apbuart_console);
>  	return 0;
>  }
> @@ -612,6 +615,10 @@ static void grlib_apbuart_configure(void
>  	rp = of_find_node_by_path("/");
>  	rp = of_get_next_child(rp, NULL);
>  	prop = of_get_property(rp, "clock-frequency", NULL);
> +	if (prop == NULL) {
> +		grlib_apbuart_port_nr = 0;
> +		return;
> +	}
>  	freq_khz = *prop;
>  
>  	line = 0;
> @@ -666,6 +673,10 @@ static int __init grlib_apbuart_init(voi
>  
>  	/* Find all APBUARTS in device the tree and initialize their ports */
>  	grlib_apbuart_configure();
> +	if (grlib_apbuart_port_nr == 0) {
> +		printk(KERN_INFO "Serial: GRLIB APBUART: No ports found.\n");
> +		return -ENODEV;
> +	}
>  
>  	printk(KERN_INFO "Serial: GRLIB APBUART driver\n");
>  
> 
> 

Re: [alternative-merged] serial-apbuartc-fix-two-problems-related-to-grlib_apbuart_configure.patch removed from -mm tree

[Miguel Ojeda]

Hi,

I found two problems using the Simics' sunfire target. It occurs as well on real machines.

1. At grlib_apbuart_configure() in apbuart.c, prop can be NULL if
"clock-frequency" doesn't exist in the OF tree:

       /* Get bus frequency */
       rp = of_find_node_by_path("/");
       rp = of_get_next_child(rp, NULL);
       prop = of_get_property(rp, "clock-frequency", NULL);
       freq_khz = *prop;

2. In addition, apbuart.c does not check if there aren't any ports
configured after calls to grlib_apbuart_configure(), so other oops will
occur if no port was configured (e.g. uart_set_options() because of
port->ops).

In order to solve that, I added a check after both of the calls to
grlib_apbuart_configure().

The patch that I provide below prevents both problems in the Simics'
sunfire target. In addition, it warns the users if no ports were found.

Signed-off-by: Miguel Ojeda <miguel.ojeda.sandonis@gmail.com>
Acked-by: Kristoffer Glembo <kristoffer@gaisler.com>
---
--- drivers/serial/apbuart.c.orig	2010-04-26 16:48:30.000000000 +0200
+++ drivers/serial/apbuart.c	2010-05-11 15:21:28.984666230 +0200
@@ -525,6 +525,9 @@ static void grlib_apbuart_configure(void
 static int __init apbuart_console_init(void)
 {
 	grlib_apbuart_configure();
+	if (grlib_apbuart_port_nr == 0)
+		return 0;
+
 	register_console(&grlib_apbuart_console);
 	return 0;
 }
@@ -612,6 +615,10 @@ static void grlib_apbuart_configure(void
 	rp = of_find_node_by_path("/");
 	rp = of_get_next_child(rp, NULL);
 	prop = of_get_property(rp, "clock-frequency", NULL);
+	if (prop == NULL) {
+		grlib_apbuart_port_nr = 0;
+		return;
+	}
 	freq_khz = *prop;
 
 	line = 0;
@@ -666,6 +673,10 @@ static int __init grlib_apbuart_init(voi
 
 	/* Find all APBUARTS in device the tree and initialize their ports */
 	grlib_apbuart_configure();
+	if (grlib_apbuart_port_nr == 0) {
+		printk(KERN_INFO "Serial: GRLIB APBUART: No ports found.\n");
+		return -ENODEV;
+	}
 
 	printk(KERN_INFO "Serial: GRLIB APBUART driver\n");
 

Re: [alternative-merged] serial-apbuartc-fix-two-problems-related-to-grlib_apbuart_configure.patch removed from -mm tree

[Miguel Ojeda]

On Thu, May 20, 2010 at 9:52 PM, Miguel Ojeda
<miguel.ojeda.sandonis@gmail.com> wrote:
> Hi,
>
> I found two problems using the Simics' sunfire target. It occurs as well on real machines.
>
> 1. At grlib_apbuart_configure() in apbuart.c, prop can be NULL if
> "clock-frequency" doesn't exist in the OF tree:
>
>       /* Get bus frequency */
>       rp = of_find_node_by_path("/");
>       rp = of_get_next_child(rp, NULL);
>       prop = of_get_property(rp, "clock-frequency", NULL);
>       freq_khz = *prop;
>
> 2. In addition, apbuart.c does not check if there aren't any ports
> configured after calls to grlib_apbuart_configure(), so other oops will
> occur if no port was configured (e.g. uart_set_options() because of
> port->ops).
>
> In order to solve that, I added a check after both of the calls to
> grlib_apbuart_configure().
>
> The patch that I provide below prevents both problems in the Simics'
> sunfire target. In addition, it warns the users if no ports were found.

CC stable@kernel.org, as .33 is "stable" now and also has apbuart.c.

>
> Signed-off-by: Miguel Ojeda <miguel.ojeda.sandonis@gmail.com>
> Acked-by: Kristoffer Glembo <kristoffer@gaisler.com>
> ---
> --- drivers/serial/apbuart.c.orig       2010-04-26 16:48:30.000000000 +0200
> +++ drivers/serial/apbuart.c    2010-05-11 15:21:28.984666230 +0200
> @@ -525,6 +525,9 @@ static void grlib_apbuart_configure(void
>  static int __init apbuart_console_init(void)
>  {
>        grlib_apbuart_configure();
> +       if (grlib_apbuart_port_nr == 0)
> +               return 0;
> +
>        register_console(&grlib_apbuart_console);
>        return 0;
>  }
> @@ -612,6 +615,10 @@ static void grlib_apbuart_configure(void
>        rp = of_find_node_by_path("/");
>        rp = of_get_next_child(rp, NULL);
>        prop = of_get_property(rp, "clock-frequency", NULL);
> +       if (prop == NULL) {
> +               grlib_apbuart_port_nr = 0;
> +               return;
> +       }
>        freq_khz = *prop;
>
>        line = 0;
> @@ -666,6 +673,10 @@ static int __init grlib_apbuart_init(voi
>
>        /* Find all APBUARTS in device the tree and initialize their ports */
>        grlib_apbuart_configure();
> +       if (grlib_apbuart_port_nr == 0) {
> +               printk(KERN_INFO "Serial: GRLIB APBUART: No ports found.\n");
> +               return -ENODEV;
> +       }
>
>        printk(KERN_INFO "Serial: GRLIB APBUART driver\n");
>
>
>
>