Re: [PATCH v5] x86: UV uv_tlb.c cleanup

[Ingo Molnar <mingo@elte.hu>]

Well, so you have changed this current sequence in uv_init_per_cpu():

>	...
>		return 1;

>  	kfree(uvhub_descs);
>  	kfree(uvhub_mask);
>       [ ...]
>  	return 0;

to:

> +       if (summarize_uvhub_sockets(nuvhubs, uvhub_descs, uvhub_mask))
> +               goto fail;


>  	kfree(uvhub_descs);
>  	kfree(uvhub_mask);
>       [ ...]
> +	init_per_cpu_tunables();
>  	return 0;
> +
> +fail:
> +	kfree(uvhub_descs);
> +	kfree(uvhub_mask);
> +	return 1;

Don't the kfree()s get changed - i.e. the patch has a side-effect?

And yes, while this might be a kmem leak fix, it is a fix and a *SIDE EFFECT* 
which i explicitly asked to be queued in a separate patch(es) from the large 
cleanup patch ...

The cleanup patch should be a cleanup of existing code - even if that code has 
bugs. Nothing more. Bug fixes are for different patches.

Thanks,

	Ingo