[PATCH V2 35/63] clocksource/drivers/time-lpc32xx: Convert init function to return error

The init functions do not return any error. They behave as the following:

  - panic, thus leading to a kernel crash while another timer may work and
       make the system boot up correctly

  or

  - print an error and let the caller unaware if the state of the system

Change that by converting the init functions to return an error conforming
to the CLOCKSOURCE_OF_RET prototype.

Proper error handling (rollback, errno value) will be changed later case
by case, thus this change just return back an error or success in the init
function.

Signed-off-by: Daniel Lezcano <redacted>
---
 drivers/clocksource/time-efm32.c | 20 +++++++++++++-------
 1 file changed, 13 insertions(+), 7 deletions(-)

diff --git a/drivers/clocksource/time-efm32.c b/drivers/clocksource/time-efm32.c
index b06e4c2..b71ffc6 100644
--- a/drivers/clocksource/time-efm32.c
+++ b/drivers/clocksource/time-efm32.c

@@ -233,7 +233,11 @@ static int __init efm32_clockevent_init(struct device_node *np)
 					DIV_ROUND_CLOSEST(rate, 1024),
 					0xf, 0xffff);
 
-	setup_irq(irq, &efm32_clock_event_irq);
+	ret = setup_irq(irq, &efm32_clock_event_irq);
+	if (ret) {
+		pr_err("Failed setup irq");
+		goto err_get_irq;
+	}
 
 	return 0;
 

@@ -255,16 +259,16 @@ err_clk_get:
  * This function asserts that we have exactly one clocksource and one
  * clock_event_device in the end.
  */
-static void __init efm32_timer_init(struct device_node *np)
+static int __init efm32_timer_init(struct device_node *np)
 {
 	static int has_clocksource, has_clockevent;
-	int ret;
+	int ret = 0;
 
 	if (!has_clocksource) {
 		ret = efm32_clocksource_init(np);
 		if (!ret) {
 			has_clocksource = 1;
-			return;
+			return 0;
 		}
 	}
 

@@ -272,9 +276,11 @@ static void __init efm32_timer_init(struct device_node *np)
 		ret = efm32_clockevent_init(np);
 		if (!ret) {
 			has_clockevent = 1;
-			return;
+			return 0;
 		}
 	}
+
+	return ret;
 }
-CLOCKSOURCE_OF_DECLARE(efm32compat, "efm32,timer", efm32_timer_init);
-CLOCKSOURCE_OF_DECLARE(efm32, "energymicro,efm32-timer", efm32_timer_init);
+CLOCKSOURCE_OF_DECLARE_RET(efm32compat, "efm32,timer", efm32_timer_init);
+CLOCKSOURCE_OF_DECLARE_RET(efm32, "energymicro,efm32-timer", efm32_timer_init);

-- 
1.9.1

Hello Daniel,

$Subject ~= s/lpc32xx/efm32/

On Thu, Jun 16, 2016 at 11:26:54PM +0200, Daniel Lezcano wrote:

The init functions do not return any error. They behave as the following:

  - panic, thus leading to a kernel crash while another timer may work and
       make the system boot up correctly

  or

  - print an error and let the caller unaware if the state of the system

These are both wrong for efm32. It doesn't panic and doesn't print an
error message (obviously the "let the caller unaware" part is true).

Change that by converting the init functions to return an error conforming
to the CLOCKSOURCE_OF_RET prototype.

Proper error handling (rollback, errno value) will be changed later case
by case, thus this change just return back an error or success in the init
function.

Apart from the comment below the error handling is already fine in this
driver.

I would prefer to introduce another label "err_setup_irq" at the same
place as err_get_irq and use it here.

+	}
 
 	return 0;
 

Best regards
Uwe


-- 
Pengutronix e.K.                           | Uwe Kleine-K?nig            |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |

The init functions do not return any error and let the caller unaware of
the state of the system.

Change that by converting the init functions to return an error conforming
to the CLOCKSOURCE_OF_RET prototype.

Proper error handling (rollback, errno value) will be changed later case
by case, thus this change just return back an error or success in the init
function.

Signed-off-by: Daniel Lezcano <redacted>
---
 drivers/clocksource/time-efm32.c | 21 ++++++++++++++-------
 1 file changed, 14 insertions(+), 7 deletions(-)

diff --git a/drivers/clocksource/time-efm32.c b/drivers/clocksource/time-efm32.c
index b06e4c2..6e2f79f 100644
--- a/drivers/clocksource/time-efm32.c
+++ b/drivers/clocksource/time-efm32.c

@@ -233,10 +233,15 @@ static int __init efm32_clockevent_init(struct device_node *np)
 					DIV_ROUND_CLOSEST(rate, 1024),
 					0xf, 0xffff);
 
-	setup_irq(irq, &efm32_clock_event_irq);
+	ret = setup_irq(irq, &efm32_clock_event_irq);
+	if (ret) {
+		pr_err("Failed setup irq");
+		goto err_setup_irq;
+	}
 
 	return 0;
 
+err_setup_irq:
 err_get_irq:
 
 	iounmap(base);

@@ -255,16 +260,16 @@ err_clk_get:
  * This function asserts that we have exactly one clocksource and one
  * clock_event_device in the end.
  */
-static void __init efm32_timer_init(struct device_node *np)
+static int __init efm32_timer_init(struct device_node *np)
 {
 	static int has_clocksource, has_clockevent;
-	int ret;
+	int ret = 0;
 
 	if (!has_clocksource) {
 		ret = efm32_clocksource_init(np);
 		if (!ret) {
 			has_clocksource = 1;
-			return;
+			return 0;
 		}
 	}
 

@@ -272,9 +277,11 @@ static void __init efm32_timer_init(struct device_node *np)
 		ret = efm32_clockevent_init(np);
 		if (!ret) {
 			has_clockevent = 1;
-			return;
+			return 0;
 		}
 	}
+
+	return ret;
 }
-CLOCKSOURCE_OF_DECLARE(efm32compat, "efm32,timer", efm32_timer_init);
-CLOCKSOURCE_OF_DECLARE(efm32, "energymicro,efm32-timer", efm32_timer_init);
+CLOCKSOURCE_OF_DECLARE_RET(efm32compat, "efm32,timer", efm32_timer_init);
+CLOCKSOURCE_OF_DECLARE_RET(efm32, "energymicro,efm32-timer", efm32_timer_init);

-- 
1.9.1

On Fri, Jun 17, 2016 at 10:44:19AM +0200, Daniel Lezcano wrote:

The init functions do not return any error and let the caller unaware of
the state of the system.

Change that by converting the init functions to return an error conforming
to the CLOCKSOURCE_OF_RET prototype.

Proper error handling (rollback, errno value) will be changed later case
by case, thus this change just return back an error or success in the init
function.

Signed-off-by: Daniel Lezcano <redacted>

Acked-by: Uwe Kleine-K?nig <redacted>

Thanks
Uwe

-- 
Pengutronix e.K.                           | Uwe Kleine-K?nig            |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |