On Wed, Dec 05, 2018 at 11:40:51AM -0800, Anthony Yznaga wrote:
> On 12/04/2018 05:25 PM, Matthew Wilcox wrote:
> > On Tue, Dec 04, 2018 at 05:18:32PM -0800, anthony.yznaga@xxxxxxxxxx wrote:
> >> On 12/04/2018 04:48 PM, Matthew Wilcox wrote:
> >>> On Tue, Dec 04, 2018 at 02:45:26PM -0800, Anthony Yznaga wrote:
> >>>> +static inline int page_has_type(struct page *page)
> >>>> +{
> >>>> + return (PageType(page, 0) &&
> >>>> + ((page->page_type & PAGE_TYPE_ALL) != PAGE_TYPE_ALL));
> >>>> +}
> >>>> +
> >>> I think this is a bit complex, and a bit of a pain to update as we add
> >>> new page types. How about this?
> >>>
> >>> return (int)page_type < -128;
> >>>
> >>> (I'm open to appropriate #defines to make this more obvious that it's ~0x7F)
> >> I thought about having this:
> >>
> >> #define PAGE_TYPE_END 0xffffff80
> >>
> >> static int inline page_has_type(struct page *page)
> >> {
> >> return page->page_type > PAGE_TYPE_BASE &&
> >> page->page_type < PAGE_TYPE_END;
> >> }
> >>
> >> But I opted for the additional complexity to avoid more false-positives from
> >> possibly corrupted values. I'm certainly fine with a simple approach, though.
> > The way I'm thinking about this field is that usually it's _mapcount
> > which is 0xffffffff to represent 0. We allow a certain small amount
> > of underflow and still treat it as a mapcount. We also allow for some
> > amount of overflow. So to be utterly precise, what you had there would
> > have been fine, but for simplicity, I'd rather just do a signed compare
> > against -128.
> The signed compare does not allow for mapcount overflow. Is that acceptable?
> False-positives would be benign for /proc/kpagecount though from a debug
> perspective it could be helpful to see overflowed mapcounts. Some future
> caller would need separate consideration.
Nobody seems terribly interested in mapcount overflows. I got no response
to https://lkml.org/lkml/2018/3/2/991
