[openib-general] Re: [PATCH][RFC][0/4] InfiniBand userspace verbs implementation

[Troy Benjegerdes]

On Mon, Apr 11, 2005 at 09:56:53AM -0700, Roland Dreier wrote:
>     Troy> Is there a check in the kernel that the memory is actually
>     Troy> mlock()ed?
> 
> No.
> 
>     Troy> What if a malicious (or broken) application does
>     Troy> ibv_reg_mr() but doesn't lock the memory? Does the IB card
>     Troy> get a physical address for a page that might get swapped
>     Troy> out?
> 
> No, the kernel does get_user_pages().  So the pages that the HCA gets
> will not be swapped or used for anything else.  The only thing a
> malicious userspace app can do is screw itself up.
> 
>  - R.

Do we even need the mlock in userspace then?