Re: [PATCH] x86/mce: Improve mcheck_init_device() error handling.

[Borislav Petkov]

On Fri, Apr 25, 2014 at 07:40:04PM +0200, Mathieu Souchaud wrote:
> Check return code of every function called.
> 
> Signed-off-by: Mathieu Souchaud <mattieu.souchaud@xxxxxxx>
> ---
>  arch/x86/kernel/cpu/mcheck/mce.c |   30 +++++++++++++++++++++++-------
>  1 file changed, 23 insertions(+), 7 deletions(-)
> 
> diff --git a/arch/x86/kernel/cpu/mcheck/mce.c b/arch/x86/kernel/cpu/mcheck/mce.c
> index 68317c8..b865285 100644
> --- a/arch/x86/kernel/cpu/mcheck/mce.c
> +++ b/arch/x86/kernel/cpu/mcheck/mce.c
> @@ -2437,33 +2437,49 @@ static __init int mcheck_init_device(void)
>  	int err;
>  	int i = 0;
>  
> -	if (!mce_available(&boot_cpu_data))
> -		return -EIO;
> +	if (!mce_available(&boot_cpu_data)) {
> +		err = -EIO;
> +		goto err_out;
> +	}
>  
> -	zalloc_cpumask_var(&mce_device_initialized, GFP_KERNEL);
> +	if (!zalloc_cpumask_var(&mce_device_initialized, GFP_KERNEL)) {
> +		err = -ENOMEM;
> +		goto err_out;
> +	}
>  
>  	mce_init_banks();
>  
>  	err = subsys_system_register(&mce_subsys, NULL);
>  	if (err)
> -		return err;
> +		goto err_out_mem;
>  
>  	cpu_notifier_register_begin();
>  	for_each_online_cpu(i) {
>  		err = mce_device_create(i);
>  		if (err) {
>  			cpu_notifier_register_done();
> -			return err;
> +			goto err_out_mem;

Actually, this should partially unwind what has been created already,
like this:

		if (err)
			goto err_device_create;

and then at that label, you can do:

err_device_create:
	while (--i > 0)
		mce_device_remove(i);
	cpu_notifier_register_done();

And then here come your other labels:

err_out_mem:
 ....

And so on. Makes sense?

>  	register_syscore_ops(&mce_syscore_ops);
> -	__register_hotcpu_notifier(&mce_cpu_notifier);
> +	err = __register_hotcpu_notifier(&mce_cpu_notifier);
>  	cpu_notifier_register_done();
> +	if (err)
> +		goto err_out_mem;

This needs to be a label which goes before err_device_create above and
which does

err_reg_hotcpu:
	unregister_syscore_ops(&mce_syscore_ops);

>  
>  	/* register character device /dev/mcelog */
> -	misc_register(&mce_chrdev_device);
> +	err = misc_register(&mce_chrdev_device);
> +	if (err)
> +		goto err_out_mem;

And this has to go to the top-label which unwinds all the work, along
with the mce percpu devices like:

	for_each_online_cpu(i)
		mce_device_remove(i);

	unregister_hotcpu_notifier(&mce_cpu_notifier);

	...

and so on.

Ok, am I making sense?

Don't be afraid to give it a try, even if it is wrong - we'll get it
right eventually :-)

For testing, you can use a kvm guest and "inject" errors in the
registration code just to see whether it works as expected.

And feel free to ask questions if something's not clear.

Thanks!

-- 
Regards/Gruss,
    Boris.

Sent from a fat crate under my desk. Formatting is fine.
--
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/