2014-07-04 11:10 GMT+02:00 Emil Goode <emilgoode@xxxxxxxxx>:
> Hello Rickard,
>
> Since this is a probe function there is also no need to release the devm_*
> resources in the i2c_pxa_remove function, this leads to double free.
>
> Also I have a few nit-pick comments below.
>
> On Thu, Jul 03, 2014 at 10:19:16PM +0200, Rickard Strandqvist wrote:
>> Fix for possible null pointer dereferenc, and there is a risk for memory leak if something unexpected happens and the function returns.
>> It now use Managed Device Resource instead.
>>
>> Signed-off-by: Rickard Strandqvist <rickard_strandqvist@xxxxxxxxxxxxxxxxxx>
>> ---
>> drivers/i2c/busses/i2c-pxa.c | 37 ++++++++++++++++---------------------
>> 1 file changed, 16 insertions(+), 21 deletions(-)
>>
>> diff --git a/drivers/i2c/busses/i2c-pxa.c b/drivers/i2c/busses/i2c-pxa.c
>> index be671f7..2edb633 100644
>> --- a/drivers/i2c/busses/i2c-pxa.c
>> +++ b/drivers/i2c/busses/i2c-pxa.c
>> @@ -1141,10 +1141,10 @@ static int i2c_pxa_probe(struct platform_device *dev)
>> struct resource *res = NULL;
>> int ret, irq;
>>
>> - i2c = kzalloc(sizeof(struct pxa_i2c), GFP_KERNEL);
>> + i2c = devm_kzalloc(&dev->dev, sizeof(struct pxa_i2c), GFP_KERNEL);
>> if (!i2c) {
>> ret = -ENOMEM;
>> - goto emalloc;
>> + goto err_nothing_to_release;
>
> Perhaps its a good idea to return directly here, then the label
> err_nothing_to_release can be removed.
>
>> }
>>
>> /* Default adapter num to device id; i2c_pxa_probe_dt can override. */
>> @@ -1154,18 +1154,19 @@ static int i2c_pxa_probe(struct platform_device *dev)
>> if (ret > 0)
>> ret = i2c_pxa_probe_pdata(dev, i2c, &i2c_type);
>> if (ret < 0)
>> - goto eclk;
>> + goto err_nothing_to_release;
>
> Same here.
>
>>
>> res = platform_get_resource(dev, IORESOURCE_MEM, 0);
>> irq = platform_get_irq(dev, 0);
>> if (res == NULL || irq < 0) {
>> ret = -ENODEV;
>> - goto eclk;
>> + goto err_nothing_to_release;
>
> Same here.
>
>> }
>>
>> - if (!request_mem_region(res->start, resource_size(res), res->name)) {
>> + if (!devm_request_mem_region(&dev->dev, res->start,
>> + resource_size(res), res->name)) {
>> ret = -ENOMEM;
>> - goto eclk;
>> + goto emalloc;
>
> We could also return directly here since the release_mem_region call should
> be removed, as mentioned in Jingoos reply.
>
>> }
>>
>> i2c->adap.owner = THIS_MODULE;
>> @@ -1176,16 +1177,16 @@ static int i2c_pxa_probe(struct platform_device *dev)
>>
>> strlcpy(i2c->adap.name, "pxa_i2c-i2c", sizeof(i2c->adap.name));
>>
>> - i2c->clk = clk_get(&dev->dev, NULL);
>> + i2c->clk = devm_clk_get(&dev->dev, NULL);
>> if (IS_ERR(i2c->clk)) {
>> ret = PTR_ERR(i2c->clk);
>> - goto eclk;
>> + goto emalloc;
>
> Same here.
>
>> }
>>
>> - i2c->reg_base = ioremap(res->start, resource_size(res));
>> - if (!i2c->reg_base) {
>> + i2c->reg_base = devm_ioremap_resource(&adev->dev, res));
>> + if (IS_ERR(i2c->reg_base)) {
>> ret = -EIO;
>> - goto eremap;
>> + goto emalloc;
>
> Same here.
>
>> }
>>
>> i2c->reg_ibmr = i2c->reg_base + pxa_reg_layout[i2c_type].ibmr;
>> @@ -1227,10 +1228,10 @@ static int i2c_pxa_probe(struct platform_device *dev)
>> i2c->adap.algo = &i2c_pxa_pio_algorithm;
>> } else {
>> i2c->adap.algo = &i2c_pxa_algorithm;
>> - ret = request_irq(irq, i2c_pxa_handler, IRQF_SHARED,
>> - dev_name(&dev->dev), i2c);
>> + ret = devm_request_irq(&dev->dev, irq, i2c_pxa_handler,
>> + IRQF_SHARED, dev_name(&dev->dev), i2c);
>> if (ret)
>> - goto ereqirq;
>> + goto emalloc;
>
> Same here.
>
>> }
>>
>> i2c_pxa_reset(i2c);
>> @@ -1261,15 +1262,9 @@ static int i2c_pxa_probe(struct platform_device *dev)
>> eadapt:
>> if (!i2c->use_pio)
>> free_irq(irq, i2c);
>
> The free_irq() call should also be removed since devm_request_irq() was used.
>
>> -ereqirq:
>> - clk_disable_unprepare(i2c->clk);
>> - iounmap(i2c->reg_base);
>> -eremap:
>> - clk_put(i2c->clk);
>> -eclk:
>> - kfree(i2c);
>> emalloc:
>> release_mem_region(res->start, resource_size(res));
>> +err_nothing_to_release:
>> return ret;
>> }
>>
>
> Best regards,
>
> Emil Goode
Hi All!
Ok, so there is literally nothing to do release anymore.
A good system Devres, I understand you want everyone to use it.
I also think it is also better to return directly then.
But who decides, Wolfram?
Would appreciate if I did not have to do more than one patch more.
And thanks Jingoo for spelling, etc.
Kind regards
Rickard Strandqvist
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
