On Thu, Apr 22, 2021 at 2:18 PM Peter Xu [off-list ref] wrote:

Axel,

On Thu, Apr 22, 2021 at 01:22:02PM -0700, Axel Rasmussen wrote:

quoted

quoted

+       if (page_in_cache)
+               page_add_file_rmap(page, false);
+       else
+               page_add_new_anon_rmap(page, dst_vma, dst_addr, false);
+
+       /*
+        * Must happen after rmap, as mm_counter() checks mapping (via
+        * PageAnon()), which is set by __page_set_anon_rmap().
+        */
+       inc_mm_counter(dst_mm, mm_counter(page));

Actually, I've noticed that this is still slightly incorrect.

As Hugh pointed out, this works for the anon case, because
page_add_new_anon_rmap() sets page->mapping.

But for the page_in_cache case, it doesn't work: unlike its anon
counterpart, page_add_file_rmap() *does not* set page->mapping.

If it's already in the page cache, shouldn't it be set already in e.g. one
previous call to shmem_add_to_page_cache()?  Thanks,

Ah, of course. Sorry for the noise. This should have been obvious to
me from how page_in_cache is defined.

I had run into the same "Bad rss-counter state" warning while applying
my patches to an earlier kernel version, and got concerned about this
line after looking at page_add_file_rmap().

But, you're right that this ought to work, and indeed I can't
reproduce the warning when the patches are based on the mm snapshot
mentioned in the cover letter. So, it seems the problem lies with this
other unrelated merge I'm doing, not the series itself. :)

--
Peter Xu