should failed calls to device_register() always call put_device()?

Belisko Marek marek.belisko at gmail.com
Sat May 28 14:56:34 EDT 2011


On Sat, May 28, 2011 at 6:29 PM, Robert P. J. Day <rpjday at crashcourse.ca> wrote:
> On Sat, 28 May 2011, Belisko Marek wrote:
>
>> Hi Robert,
>>
>> On Sat, May 28, 2011 at 5:15 PM, Robert P. J. Day <rpjday at crashcourse.ca> wrote:
>> >
>> >  from drivers/base/core.c, we have the fairly unambiguous advice:
>> >
>> > * NOTE: _Never_ directly free @dev after calling this function, even
>> > * if it returned an error! Always use put_device() to give up the
>> > * reference initialized in this function instead.
>> > */
>> > int device_register(struct device *dev)
>> > {
>> >        device_initialize(dev);
>> >        return device_add(dev);
>> > }
>> >
>> >  and yet, there appears to be driver code that does exactly that,
>> > such as this snippet from drivers/w1/w1_int.c (line 86):
>> >
>> >        ... snip ...
>> >        err = device_register(&dev->dev);
>> >        if (err) {
>> >                printk(KERN_ERR "Failed to register master device. err=%d\n", err);
>> >                memset(dev, 0, sizeof(struct w1_master));
>> >                kfree(dev);
>> >                dev = NULL;
>> >        }
>> Free is for allocated dev not for struct device so it is OK. IMO thi
>> snippet should look like:
>> err = device_register(&dev->dev);
>> if (err) {
>>     printk(KERN_ERR "Failed to register master device. err=%d\n", err);
>>     put_device(&dev->dev);
>>     memset(dev, 0, sizeof(struct w1_master));
>>     kfree(dev);
>>     dev = NULL;
>> }
>
>  i agree that there should be a "put_device(&dev->dev);" statement as
> you show above.  however, i still don't see how this can be just a
> stylistic improvement as you seem to suggest.  based on the warning
> from the kernel source file, it would seem that you *must* do a
> put_device() in that situation -- it's not optional.
Sure you're right. You can send a patch to fix this problem. Good catch.
>
> rday
>
> p.s.  i would also never do a memset() to zero, followed by a kfree(),
> when a kzfree() is so much more concise.
>
> --
>
> ========================================================================
> Robert P. J. Day                                 Ottawa, Ontario, CANADA
>                        http://crashcourse.ca
>
> Twitter:                                       http://twitter.com/rpjday
> LinkedIn:                               http://ca.linkedin.com/in/rpjday
> ========================================================================

regards,

marek

-- 
as simple and primitive as possible
-------------------------------------------------
Marek Belisko - OPEN-NANDRA
Freelance Developer

Ruska Nova Ves 219 | Presov, 08005 Slovak Republic
Tel: +421 915 052 184
skype: marekwhite
icq: 290551086
web: http://open-nandra.com



More information about the Kernelnewbies mailing list