On 10/14/2021 3:57 PM, Ralph Campbell wrote:

On 10/14/21 11:01 AM, Jason Gunthorpe wrote:

quoted

On Thu, Oct 14, 2021 at 10:35:27AM -0700, Ralph Campbell wrote:

quoted

I ran xfstests-dev using the kernel boot option to "fake" a pmem device
when I first posted this patch. The tests ran OK (or at least the same
tests passed with and without my patch).

Hmm. I know nothing of xfstests but

tests/generic/413

Looks kind of like it might cover this situation?

Did it run for you?

Jason

I don't remember. I'll have to rerun the test which might take a day 
or two
to set up again.

I just ran this generic/413 on my side using pmem fake device. It does fail.
I remember we proposed a fix on this patch before try_get_page was removed.

@@ -1186,7 +1153,7 @@ bool __must_check try_grab_page(struct page *page, 

unsigned int flags);
  static inline __must_check bool try_get_page(struct page *page)
  {
         page = compound_head(page);
-       if (WARN_ON_ONCE(page_ref_count(page) <= 0))
+       if (WARN_ON_ONCE(page_ref_count(page) < 
(int)!is_zone_device_page(page)))
                 return false;
         page_ref_inc(page);
         return true;

Alex