Re: [PATCH] x86: mcheck: call put_device on device_register failure

[Chen, Gong]

On Sat, Nov 30, 2013 at 12:12:14PM +0100, Borislav Petkov wrote:
> Date:	Sat, 30 Nov 2013 12:12:14 +0100
> From: Borislav Petkov <bp@xxxxxxxxx>
> To: Levente Kurusa <levex@xxxxxxxxx>
> Cc: Ingo Molnar <mingo@xxxxxxxxxx>, Thomas Gleixner <tglx@xxxxxxxxxxxxx>,
>  Tony Luck <tony.luck@xxxxxxxxx>, "H. Peter Anvin" <hpa@xxxxxxxxx>,
>  x86@xxxxxxxxxx, EDAC <linux-edac@xxxxxxxxxxxxxxx>, LKML
>  <linux-kernel@xxxxxxxxxxxxxxx>
> Subject: Re: [PATCH] x86: mcheck: call put_device on device_register failure
> User-Agent: Mutt/1.5.21 (2010-09-15)
> 
> On Sat, Nov 30, 2013 at 08:30:33AM +0100, Levente Kurusa wrote:
> > No, if the call to put_device gives up the last reference to the
> > device, then device_release gets called which in turn frees the memory
> > associated with it. In this case, mce_device_release() will get
> > called, which is just a simple kfree call.
> 
> Aah, that's that delayed freeing the driver core does, right. Now you
> made me go and look into detail:
> 
> device_unregister
> |->put_device
>   |->kobject_put
>      |->kref_put(&kobj->kref, kobject_release)
> 	|->kref_sub(kref, 1, release)
> 	   |->release
> 	   |->kobject_release
> 	      |->kobject_cleanup
> 	         |->t->release
> 		 |->device_release
> 		    |->mce_device_release
> 
> 
> Ok, I see it now. :-) :-)
> 
> Thanks, I'll take your patch as-is.
> 

I have some concerns about it. if device_register is failed, it will
backtraces all kinds of conditions automatically, including put_device
definately. So do we really need an extra put_device when it returns
failure?

Attachment: signature.asc
Description: Digital signature