On Tue, 21 Aug 2012, Lukas Czerner wrote:
> On Sun, 19 Aug 2012, Hugh Dickins wrote:
> > > --- a/include/linux/fs.h
> > > +++ b/include/linux/fs.h
> > > @@ -620,6 +620,8 @@ struct address_space_operations {
> > > /* Unfortunately this kludge is needed for FIBMAP. Don't use it */
> > > sector_t (*bmap)(struct address_space *, sector_t);
> > > void (*invalidatepage) (struct page *, unsigned long);
> > > + void (*invalidatepage_range) (struct page *, unsigned long,
> > > + unsigned long);
> >
> > It may turn out to be bad advice, given how invalidatepage() already
> > takes an unsigned long, but I'd be tempted to make both of these args
> > unsigned int, since that helps to make it clearer that they're intended
> > to be offsets within a page, in the range 0..PAGE_CACHE_SIZE.
> >
> > (partial_start, partial_end and top in truncate_inode_pages_range()
> > are all unsigned int.)
>
> Hmm, this does not seem right. I can see that PAGE_CACHE_SIZE
> (PAGE_SIZE) can be defined as unsigned long, or am I missing
> something ?
They would be defined as unsigned long so that they can be used in
masks like ~(PAGE_SIZE - 1), and behave as expected on addresses,
without needing casts to be added all over.
We do not (currently!) expect PAGE_SIZE or PAGE_CACHE_SIZE to grow
beyond an unsigned int - but indeed they can be larger than what's
held in an unsigned short (look no further than ia64 or ppc64).
For more reassurance, see include/linux/highmem.h, which declares
zero_user_segments() and others: unsigned int (well, unsigned with
the int implicit) for offsets within a page.
Hugh
> >
> > Andrew is very keen on naming arguments in prototypes,
> > and I think there is an especially strong case for it here.
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>