From: William Lee Irwin III The smp_mb() is becaus sync_page() doesn't have PG_locked while it accesses page_mapping(page). The comments in the patch (the entire patch is the addition of this comment) try to explain further how and why smp_mb() is used. Signed-off-by: Andrew Morton --- mm/filemap.c | 20 +++++++++++++++++++- 1 files changed, 19 insertions(+), 1 deletion(-) diff -puN mm/filemap.c~sync_page-smp_mb-comment mm/filemap.c --- 25/mm/filemap.c~sync_page-smp_mb-comment Fri Apr 29 16:13:55 2005 +++ 25-akpm/mm/filemap.c Fri Apr 29 16:13:55 2005 @@ -139,7 +139,25 @@ static int sync_page(void *word) page = container_of((page_flags_t *)word, struct page, flags); /* - * FIXME, fercrissake. What is this barrier here for? + * page_mapping() is being called without PG_locked held. + * Some knowledge of the state and use of the page is used to + * reduce the requirements down to a memory barrier. + * The danger here is of a stale page_mapping() return value + * indicating a struct address_space different from the one it's + * associated with when it is associated with one. + * After smp_mb(), it's either the correct page_mapping() for + * the page, or an old page_mapping() and the page's own + * page_mapping() has gone NULL. + * The ->sync_page() address_space operation must tolerate + * page_mapping() going NULL. By an amazing coincidence, + * this comes about because none of the users of the page + * in the ->sync_page() methods make essential use of the + * page_mapping(), merely passing the page down to the backing + * device's unplug functions when it's non-NULL, which in turn + * ignore it for all cases but swap, where only page->private is + * of interest. When page_mapping() does go NULL, the entire + * call stack gracefully ignores the page and returns. + * -- wli */ smp_mb(); mapping = page_mapping(page); _