Patchwork mtd: orion_nand: Improve error handling in orion_nand_probe

login
register
mail settings
Submitter Emil Goode
Date June 9, 2013, 9:08 a.m.
Message ID <1370768902-20239-1-git-send-email-emilgoode@gmail.com>
Download mbox | patch
Permalink /patch/250016/
State New
Headers show

Comments

Emil Goode - June 9, 2013, 9:08 a.m.
This patch fixes some issues in the error handling and simplifies
the code by converting to devm* functions.

If the kzalloc call fails it is unnecessary to use the label no_res
and pass a NULL pointer to kfree. If the devm_kzalloc call fails on
line 110 we forgett to call iounmap for the previous ioremap call.

The following changes are introduced:
- Convert to devm_kzalloc and remove calls to kfree.
- Convert to devm_ioremap_resource that adds a missing call to
  *request_mem_region and remove calls to iounmap.
- The devm_ioremap_resource function checks the passed resource so
  we can remove the NULL check after the platform_get_resource call.

Signed-off-by: Emil Goode <emilgoode@gmail.com>
---
The patch is only build tested

 drivers/mtd/nand/orion_nand.c |   39 +++++++++++----------------------------
 1 file changed, 11 insertions(+), 28 deletions(-)
Andy Shevchenko - June 9, 2013, 5:15 p.m.
On Sun, Jun 9, 2013 at 12:08 PM, Emil Goode <emilgoode@gmail.com> wrote:
> This patch fixes some issues in the error handling and simplifies
> the code by converting to devm* functions.
>
> If the kzalloc call fails it is unnecessary to use the label no_res
> and pass a NULL pointer to kfree. If the devm_kzalloc call fails on
> line 110 we forgett to call iounmap for the previous ioremap call.

"forget" (typo here)

> The following changes are introduced:
> - Convert to devm_kzalloc and remove calls to kfree.
> - Convert to devm_ioremap_resource that adds a missing call to
>   *request_mem_region and remove calls to iounmap.
> - The devm_ioremap_resource function checks the passed resource so
>   we can remove the NULL check after the platform_get_resource call.

What about another patch (that I guess should go first)  that converts
printk to dev_* or pr_*?

--
With Best Regards,
Andy Shevchenko
Dan Carpenter - June 9, 2013, 7:17 p.m.
On Sun, Jun 09, 2013 at 08:15:20PM +0300, Andy Shevchenko wrote:
> On Sun, Jun 9, 2013 at 12:08 PM, Emil Goode <emilgoode@gmail.com> wrote:
> > This patch fixes some issues in the error handling and simplifies
> > the code by converting to devm* functions.
> >
> > If the kzalloc call fails it is unnecessary to use the label no_res
> > and pass a NULL pointer to kfree. If the devm_kzalloc call fails on
> > line 110 we forgett to call iounmap for the previous ioremap call.
> 
> "forget" (typo here)
> 
> > The following changes are introduced:
> > - Convert to devm_kzalloc and remove calls to kfree.
> > - Convert to devm_ioremap_resource that adds a missing call to
> >   *request_mem_region and remove calls to iounmap.
> > - The devm_ioremap_resource function checks the passed resource so
> >   we can remove the NULL check after the platform_get_resource call.
> 
> What about another patch (that I guess should go first)  that converts
> printk to dev_* or pr_*?
> 

Huh, what?  Those are two totally unrelated things.  Emil's patch is
a bug fix.

Patches are applied in first come first serve order anyway, so I'm
not sure what you are saying.

regards,
dan carpenter
Andy Shevchenko - June 9, 2013, 7:27 p.m.
On Sun, Jun 9, 2013 at 10:17 PM, Dan Carpenter <dan.carpenter@oracle.com> wrote:
> On Sun, Jun 09, 2013 at 08:15:20PM +0300, Andy Shevchenko wrote:

[]

>> What about another patch (that I guess should go first)  that converts
>> printk to dev_* or pr_*?

> Huh, what?  Those are two totally unrelated things.  Emil's patch is
> a bug fix.

To be precise his patch is bugfix and cleanup at once.
My proposal is to add a clean up patch.

--
With Best Regards,
Andy Shevchenko
Dan Carpenter - June 9, 2013, 8:57 p.m.
On Sun, Jun 09, 2013 at 10:27:03PM +0300, Andy Shevchenko wrote:
> On Sun, Jun 9, 2013 at 10:17 PM, Dan Carpenter <dan.carpenter@oracle.com> wrote:
> > On Sun, Jun 09, 2013 at 08:15:20PM +0300, Andy Shevchenko wrote:
> 
> []
> 
> >> What about another patch (that I guess should go first)  that converts
> >> printk to dev_* or pr_*?
> 
> > Huh, what?  Those are two totally unrelated things.  Emil's patch is
> > a bug fix.
> 
> To be precise his patch is bugfix and cleanup at once.
> My proposal is to add a clean up patch.

The printk idea isn't a bad one, it's just totally unrelated to what
the patch is doing.  You can't just randomly tell people to do a
bunch of work for no reason.  "Good bugfix, but before we apply it
you have to mow my lawn."

regards,
dan carpenter
Andy Shevchenko - June 9, 2013, 9:15 p.m.
On Sun, Jun 9, 2013 at 11:57 PM, Dan Carpenter <dan.carpenter@oracle.com> wrote:
> On Sun, Jun 09, 2013 at 10:27:03PM +0300, Andy Shevchenko wrote:
>> On Sun, Jun 9, 2013 at 10:17 PM, Dan Carpenter <dan.carpenter@oracle.com> wrote:
>> > On Sun, Jun 09, 2013 at 08:15:20PM +0300, Andy Shevchenko wrote:

>> >> What about another patch (that I guess should go first)  that converts
>> >> printk to dev_* or pr_*?
>>
>> > Huh, what?  Those are two totally unrelated things.  Emil's patch is
>> > a bug fix.
>>
>> To be precise his patch is bugfix and cleanup at once.
>> My proposal is to add a clean up patch.
>
> The printk idea isn't a bad one, it's just totally unrelated to what
> the patch is doing.

I don't agree with word 'totally'. It touches probe function and
shuffles code there.

>  You can't just randomly tell people to do a
> bunch of work for no reason.  "Good bugfix, but before we apply it
> you have to mow my lawn."

In this particular case I didn't think it's a big deal to fix 3 printfs.
Anyway, see patch in the next message.


--
With Best Regards,
Andy Shevchenko
Emil Goode - June 9, 2013, 10:45 p.m.
Hello,

Sorry for not responding earlier, I was away from my computer.
Thanks for the review, I will send a second version that applies
on top of Andy's patch and fix that typo in the change log.

Best regards,

Emil

On Mon, Jun 10, 2013 at 12:15:22AM +0300, Andy Shevchenko wrote:
> On Sun, Jun 9, 2013 at 11:57 PM, Dan Carpenter <dan.carpenter@oracle.com> wrote:
> > On Sun, Jun 09, 2013 at 10:27:03PM +0300, Andy Shevchenko wrote:
> >> On Sun, Jun 9, 2013 at 10:17 PM, Dan Carpenter <dan.carpenter@oracle.com> wrote:
> >> > On Sun, Jun 09, 2013 at 08:15:20PM +0300, Andy Shevchenko wrote:
> 
> >> >> What about another patch (that I guess should go first)  that converts
> >> >> printk to dev_* or pr_*?
> >>
> >> > Huh, what?  Those are two totally unrelated things.  Emil's patch is
> >> > a bug fix.
> >>
> >> To be precise his patch is bugfix and cleanup at once.
> >> My proposal is to add a clean up patch.
> >
> > The printk idea isn't a bad one, it's just totally unrelated to what
> > the patch is doing.
> 
> I don't agree with word 'totally'. It touches probe function and
> shuffles code there.
> 
> >  You can't just randomly tell people to do a
> > bunch of work for no reason.  "Good bugfix, but before we apply it
> > you have to mow my lawn."
> 
> In this particular case I didn't think it's a big deal to fix 3 printfs.
> Anyway, see patch in the next message.
> 
> 
> --
> With Best Regards,
> Andy Shevchenko

Patch

diff --git a/drivers/mtd/nand/orion_nand.c b/drivers/mtd/nand/orion_nand.c
index 8fbd002..cc22f2f 100644
--- a/drivers/mtd/nand/orion_nand.c
+++ b/drivers/mtd/nand/orion_nand.c
@@ -85,34 +85,25 @@  static int __init orion_nand_probe(struct platform_device *pdev)
 	int ret = 0;
 	u32 val = 0;
 
-	nc = kzalloc(sizeof(struct nand_chip) + sizeof(struct mtd_info), GFP_KERNEL);
+	nc = devm_kzalloc(&pdev->dev, sizeof(struct nand_chip) +
+			  sizeof(struct mtd_info), GFP_KERNEL);
 	if (!nc) {
 		printk(KERN_ERR "orion_nand: failed to allocate device structure.\n");
-		ret = -ENOMEM;
-		goto no_res;
+		return -ENOMEM;
 	}
 	mtd = (struct mtd_info *)(nc + 1);
 
 	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
-	if (!res) {
-		ret = -ENODEV;
-		goto no_res;
-	}
-
-	io_base = ioremap(res->start, resource_size(res));
-	if (!io_base) {
-		printk(KERN_ERR "orion_nand: ioremap failed\n");
-		ret = -EIO;
-		goto no_res;
-	}
+	io_base = devm_ioremap_resource(&pdev->dev, res);
+	if (IS_ERR(io_base))
+		return PTR_ERR(io_base);
 
 	if (pdev->dev.of_node) {
 		board = devm_kzalloc(&pdev->dev, sizeof(struct orion_nand_data),
-					GFP_KERNEL);
+				     GFP_KERNEL);
 		if (!board) {
 			printk(KERN_ERR "orion_nand: failed to allocate board structure.\n");
-			ret = -ENOMEM;
-			goto no_res;
+			return -ENOMEM;
 		}
 		if (!of_property_read_u32(pdev->dev.of_node, "cle", &val))
 			board->cle = (u8)val;
@@ -167,7 +158,7 @@  static int __init orion_nand_probe(struct platform_device *pdev)
 
 	if (nand_scan(mtd, 1)) {
 		ret = -ENXIO;
-		goto no_dev;
+		goto disable_clk;
 	}
 
 	mtd->name = "orion_nand";
@@ -176,20 +167,17 @@  static int __init orion_nand_probe(struct platform_device *pdev)
 			board->parts, board->nr_parts);
 	if (ret) {
 		nand_release(mtd);
-		goto no_dev;
+		goto disable_clk;
 	}
 
 	return 0;
 
-no_dev:
+disable_clk:
 	if (!IS_ERR(clk)) {
 		clk_disable_unprepare(clk);
 		clk_put(clk);
 	}
 	platform_set_drvdata(pdev, NULL);
-	iounmap(io_base);
-no_res:
-	kfree(nc);
 
 	return ret;
 }
@@ -197,15 +185,10 @@  no_res:
 static int orion_nand_remove(struct platform_device *pdev)
 {
 	struct mtd_info *mtd = platform_get_drvdata(pdev);
-	struct nand_chip *nc = mtd->priv;
 	struct clk *clk;
 
 	nand_release(mtd);
 
-	iounmap(nc->IO_ADDR_W);
-
-	kfree(nc);
-
 	clk = clk_get(&pdev->dev, NULL);
 	if (!IS_ERR(clk)) {
 		clk_disable_unprepare(clk);