On Fri, 2018-03-23 at 11:38 +0100, Alexandre Belloni wrote:

On 23/03/2018 at 17:15:05 +0800, sean.wang@mediatek.com wrote:

quoted

From: Sean Wang <sean.wang@mediatek.com>

It's unnecessary doing irq_dispose_mapping as a reverse operation for
platform_get_irq.

Ususally, irq_dispose_mapping should be called in error path or module
removal to release the resources for irq_of_parse_and_map requested.

Signed-off-by: Sean Wang <sean.wang@mediatek.com>
---
 drivers/rtc/rtc-mt6397.c | 7 ++-----
 1 file changed, 2 insertions(+), 5 deletions(-)

diff --git a/drivers/rtc/rtc-mt6397.c b/drivers/rtc/rtc-mt6397.c
index b62eaa8..cefb83b 100644
--- a/drivers/rtc/rtc-mt6397.c
+++ b/drivers/rtc/rtc-mt6397.c

@@ -17,7 +17,6 @@
 #include <linux/module.h>
 #include <linux/regmap.h>
 #include <linux/rtc.h>
-#include <linux/irqdomain.h>
 #include <linux/jiffies.h>
 #include <linux/platform_device.h>
 #include <linux/of_address.h>

@@ -336,7 +335,7 @@ static int mtk_rtc_probe(struct platform_device *pdev)
 	if (ret) {
 		dev_err(&pdev->dev, "Failed to request alarm IRQ: %d: %d\n",
 			rtc->irq, ret);
-		goto out_dispose_irq;
+		return ret;
 	}
 
 	device_init_wakeup(&pdev->dev, 1);

@@ -353,8 +352,7 @@ static int mtk_rtc_probe(struct platform_device *pdev)
 
 out_free_irq:
 	free_irq(rtc->irq, rtc->rtc_dev);
-out_dispose_irq:
-	irq_dispose_mapping(rtc->irq);
+

Don't you still have a irq_create_mapping?

Sorry for that I didn't mention in the beginning that the series must
depend on another patch [1]. With the patch, the job irq_create_mapping
had been moved from rtc to mfd, so here it should be better to cleanup
up irq_dispose_mapping in all paths.

[1] https://patchwork.kernel.org/patch/9954643/

quoted

 	return ret;
 }
 

@@ -364,7 +362,6 @@ static int mtk_rtc_remove(struct platform_device *pdev)
 
 	rtc_device_unregister(rtc->rtc_dev);
 	free_irq(rtc->irq, rtc->rtc_dev);
-	irq_dispose_mapping(rtc->irq);
 
 	return 0;
 }

-- 
2.7.4