On Mon, 14 Feb 2011, Marek Vasut wrote:

On Sunday 13 February 2011 13:12:12 Julia Lawall wrote:

quoted

Request_region should be used with release_region, not release_resource.

The result of request_mem_region is no longer stored.  Instead the field
ioarea is used to store a pointer to the resource structure that contains
the start address.  This is the information that is needed later in
nuc900_i2c_remove to release the region.  The field ioarea is also printed
in some debugging code.

The semantic match that finds this problem is as follows:
(http://coccinelle.lip6.fr/)

// <smpl>
@@
expression x,E;
@@
(
*x = request_region(...)

*x = request_mem_region(...)
)
... when != release_region(x)
    when != x = E
* release_resource(x);
// </smpl>

Signed-off-by: Julia Lawall <redacted>

---
Not compiled due to incompatible architecture.

 drivers/i2c/busses/i2c-nuc900.c |   13 +++++--------
 1 file changed, 5 insertions(+), 8 deletions(-)

diff --git a/drivers/i2c/busses/i2c-nuc900.c

b/drivers/i2c/busses/i2c-nuc900.c index 7243426..97b16b7 100644

--- a/drivers/i2c/busses/i2c-nuc900.c
+++ b/drivers/i2c/busses/i2c-nuc900.c

@@ -568,15 +568,13 @@ static int __devinit nuc900_i2c_probe(struct

platform_device *pdev) goto err_clk;
 	}

-	i2c->ioarea = request_mem_region(res->start, resource_size(res),
-					 pdev->name);
-
-	if (i2c->ioarea == NULL) {
+	if (!request_mem_region(res->start, resource_size(res), pdev->name)) {

I prefer the original version here -- it's clearer and also, doesn't the new 
version violate line-over-80 principle ?

The line ends before column 80.

I'm not sure that the result of request_mem_region is the proper argment 
to give to release_region.  In any case, all of the other calls to 
request_mem_region that do use release_region do not use the return value.  
They just test it directly as I have done here.

julia

quoted

 		dev_err(&pdev->dev, "cannot request IO\n");
 		ret = -ENXIO;
 		goto err_clk;
 	}

+	i2c->ioarea = res;
 	i2c->regs = ioremap(res->start, resource_size(res));

 	if (i2c->regs == NULL) {

@@ -645,8 +643,7 @@ static int __devinit nuc900_i2c_probe(struct

platform_device *pdev) iounmap(i2c->regs);

  err_ioarea:
-	release_resource(i2c->ioarea);
-	kfree(i2c->ioarea);
+	release_mem_region(res->start, resource_size(res));

  err_clk:
 	clk_disable(i2c->clk);

@@ -665,6 +662,7 @@ static int __devinit nuc900_i2c_probe(struct

platform_device *pdev) static int __devexit nuc900_i2c_remove(struct
platform_device *pdev) {
 	struct nuc900_i2c *i2c = platform_get_drvdata(pdev);
+	struct resource *res = i2c->ioarea;

No need for this ...

quoted

 	i2c_del_adapter(&i2c->adap);
 	free_irq(i2c->irq, i2c);

@@ -674,8 +672,7 @@ static int __devexit nuc900_i2c_remove(struct

platform_device *pdev)

 	iounmap(i2c->regs);

-	release_resource(i2c->ioarea);
-	kfree(i2c->ioarea);

... the i2c pointer is still available here, why introducing struct resource 
*res. release_mem_region() looks like a sane change to me though.

quoted

+	release_mem_region(res->start, resource_size(res));
 	kfree(i2c);

 	return 0;

--
To unsubscribe from this list: send the line "unsubscribe kernel-janitors" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html