On Thu, Mar 29, 2007 at 10:23:05AM +0100, Derek Murray wrote:
> On 29 Mar 2007, at 07:26, Isaku Yamahata wrote:
> >On Wed, Mar 28, 2007 at 01:56:16PM +0100, Derek Murray wrote:
> >
> >>+static int gntdev_mmap (struct file *flip, struct vm_area_struct
> >>*vma)
> >>+{
> >...
> >>+undo_map_out:
> >
> >This doesn't undo vm_insert_page() when auto trasnalted physmap mode.
> >Probably just calling zap_page_range() would be OK instead
> >of undo_count_kmap loop.
>
> Hmm, in fact, this whole section is redundant, because a failing
> gntdev_mmap() will be cleaned up in do_mmap_pgoff(), and,
> subsequently, the gntdev_clear_pte() hook will be called. This does a
> pte_clear_full() on the user-space PTE in auto translated physmap
> mode, so would this be enough?
Yes. Understood.
--
yamahata
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel
|
Home