Re: cryptoapi highmem bug

[Andrew Morton <akpm@osdl.org>]

James Morris <jmorris@redhat.com> wrote:
>
> On Tue, 24 Feb 2004, Andrew Morton wrote:
> 
> > Jean-Luc Cooke <jlcooke@certainkey.com> wrote:
> > >
> > > How do I check for equal real addresses from two virtual ones?
> > 
> > I don't think there is a practical way of doing this.  It would involve
> > comparing the virtual address with the kmap and atomic kmap regions,
> > performing a pagetable walk, extracting the pageframe.  If the page is not
> > in a kmap area generate the pageframe directly.  Make that work on all
> > architectures.  Very yuk.
> > 
> > If practical this API should have been defined in terms of
> > (page/offset/len) and it should have kmapped the pages itself.  I guess
> > it's too late for that.
> 
> Do you mean that the crypt() function should do kmapping?

Passing the page* down one more level would permit cbc_process() to get at
the pageframes themselves, so it can sanely determine if src and dest
overlap.  (I assume this is for encryption-in-place).

Or maybe it's sufficient for crypt() to pass a simple boolean down to the
prfn() callout which says "this is in-place encryption".