linux-kernel - [RFC][PATCH 1/2] mm/migrate: remove extra page

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Fri, 01 May 2020 14:05:18 -0700
From:   Dave Hansen <dave.hansen@...ux.intel.com>
To:     linux-kernel@...r.kernel.org
Cc:     Dave Hansen <dave.hansen@...ux.intel.com>, npiggin@...il.com,
        akpm@...ux-foundation.org, willy@...radead.org,
        yang.shi@...ux.alibaba.com, linux-mm@...ck.org
Subject: [RFC][PATCH 1/2] mm/migrate: remove extra page_count() check


From: Dave Hansen <dave.hansen@...ux.intel.com>

This is not a bug fix.  It was found by inspection, but I believe
that it is confusing as it stands.

First, page_ref_freeze() is implemented internally with:

	atomic_cmpxchg(&page->_refcount, expected, 0) == expected

The "cmp" part of cmpxchg is making sure that _refcount==expected
which means that there's an implicit check here, equivalent to:

	page_count(page) == expected_count

This appears to have originated in "e286781: mm: speculative page
references", which is pretty ancient.  This check is also somewhat
dangerous to have here because it might lead someone to think that
page_ref_freeze() *doesn't* do its own page_count() checking.

Remove the unnecessary check.

Signed-off-by: Dave Hansen <dave.hansen@...ux.intel.com>
Cc: Nicholas Piggin <npiggin@...il.com>
Cc: Andrew Morton <akpm@...ux-foundation.org>
Cc: Matthew Wilcox (Oracle) <willy@...radead.org>
Cc: Yang Shi <yang.shi@...ux.alibaba.com>
Cc: linux-mm@...ck.org
Cc: linux-kernel@...r.kernel.org
---

 b/mm/migrate.c |    3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff -puN mm/migrate.c~remove_extra_page_count_check mm/migrate.c
--- a/mm/migrate.c~remove_extra_page_count_check	2020-05-01 14:00:42.331525924 -0700
+++ b/mm/migrate.c	2020-05-01 14:00:42.336525924 -0700
@@ -425,11 +425,12 @@ int migrate_page_move_mapping(struct add
 	newzone = page_zone(newpage);
 
 	xas_lock_irq(&xas);
-	if (page_count(page) != expected_count || xas_load(&xas) != page) {
+	if (xas_load(&xas) != page) {
 		xas_unlock_irq(&xas);
 		return -EAGAIN;
 	}
 
+	/* Freezing will fail if page_count()!=expected_count */
 	if (!page_ref_freeze(page, expected_count)) {
 		xas_unlock_irq(&xas);
 		return -EAGAIN;
_