Re: [Patch 2/3] Export get_one_pte_map.

[Hugh Dickins <hugh@veritas.com>]

On Mon, 17 Oct 2005, Robin Holt wrote:
> On Mon, Oct 17, 2005 at 01:41:52PM +0200, Dave Hansen wrote:
> > On Mon, 2005-10-17 at 06:31 -0500, Robin Holt wrote:
> > > On Fri, Oct 14, 2005 at 02:30:38PM -0700, Greg KH wrote:
> > > > On Fri, Oct 14, 2005 at 02:22:25PM -0500, Robin Holt wrote:
> > > > > +EXPORT_SYMBOL(get_one_pte_map);
> 
> I got a little push from our internal incident tracking system for
> this being a module.  _GPL it will be.

Sorry, Robin, I've not been following your patches.  But if you look
at 2.6.14-rc4-mm1, you'll find that there isn't even a get_one_pte_map
there.  Though there's no certainty yet that my pt locking changes, or
Nick's PageReserved changes, will actually go forward, there's a lot of
work queued up in -mm that is likely to affect your code.  And I don't
think exporting internal functions from mremap.c, _GPL or otherwise,
is the way to go.

Moving useful functions to a more central location might be.  But I'm
very dubious about your doing this kind of pte stuff deep down in an
architecture-specific driver.  You're not the only one interested in
this kind of functionality: we were thinking of providing it via an
alternative to the ->nopage method, which deals in pfns rather than
struct pages (I think that was wli's suggestion originally; Carsten
has an interest in it on s390, and I bet there are others).  There
may be excellent reasons why that wouldn't be good enough for you,
and your retcode method may be a better idea: I don't know yet.

Please rebase your work to 2.6.14-rc4-mm1 (but I won't get to look
at the result for a few days: perhaps others will).

The big question has to be: what are you expecting to happen for
PROT_WRITE, MAP_PRIVATE?

Hugh

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>