mm: add comment why mark_page_accessed() would be better than pte_mkyoung() in follow...
KOSAKI Motohiro [Tue, 31 Mar 2009 22:19:37 +0000 (15:19 -0700)]
At first look, mark_page_accessed() in follow_page() seems a bit strange.
It seems pte_mkyoung() would be better consistent with other kernel code.

However, it is intentional. The commit log said:

    ------------------------------------------------
    commit 9e45f61d69be9024a2e6bef3831fb04d90fac7a8
    Author: akpm <akpm>
    Date:   Fri Aug 15 07:24:59 2003 +0000

    [PATCH] Use mark_page_accessed() in follow_page()

    Touching a page via follow_page() counts as a reference so we should be
    either setting the referenced bit in the pte or running mark_page_accessed().

    Altering the pte is tricky because we haven't implemented an atomic
    pte_mkyoung().  And mark_page_accessed() is better anyway because it has more
    aging state: it can move the page onto the active list.

    BKrev: 3f3c8acbplT8FbwBVGtth7QmnqWkIw
    ------------------------------------------------

The atomic issue is still true nowadays. adding comment help to understand
code intention and it would be better.

[akpm@linux-foundation.org: clarify text]
Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: Nick Piggin <npiggin@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/memory.c

index 0017111..5b4ad5e 100644 (file)
@@ -1151,6 +1151,11 @@ struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
                if ((flags & FOLL_WRITE) &&
                    !pte_dirty(pte) && !PageDirty(page))
                        set_page_dirty(page);
+               /*
+                * pte_mkyoung() would be more correct here, but atomic care
+                * is needed to avoid losing the dirty bit: it is easier to use
+                * mark_page_accessed().
+                */
                mark_page_accessed(page);
        }
 unlock: