linux-kernel - [PATCH 1/1] mm: compaction: avoid fast_isolate_around() to set pageblock

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]

Message-Id: <20201121194506.13464-2-aarcange@redhat.com>
Date:   Sat, 21 Nov 2020 14:45:06 -0500
From:   Andrea Arcangeli <aarcange@...hat.com>
To:     Mel Gorman <mgorman@...e.de>,
        Andrew Morton <akpm@...ux-foundation.org>, linux-mm@...ck.org,
        Qian Cai <cai@....pw>
Cc:     Vlastimil Babka <vbabka@...e.cz>, Michal Hocko <mhocko@...nel.org>,
        David Hildenbrand <david@...hat.com>,
        linux-kernel@...r.kernel.org, Mike Rapoport <rppt@...ux.ibm.com>,
        Baoquan He <bhe@...hat.com>
Subject: [PATCH 1/1] mm: compaction: avoid fast_isolate_around() to set pageblock_skip on reserved pages

A corollary issue was fixed in
e577c8b64d58fe307ea4d5149d31615df2d90861. A second issue remained in
v5.7:

	https://lkml.kernel.org/r/8C537EB7-85EE-4DCF-943E-3CC0ED0DF56D@lca.pw

==
page:ffffea0000aa0000 refcount:1 mapcount:0 mapping:000000002243743b index:0x0
flags: 0x1fffe000001000(reserved)
==

73a6e474cb376921a311786652782155eac2fdf0 was applied to supposedly the
second issue, but I still reproduced it twice with v5.9 on two
different systems:

==
page:0000000062b3e92f refcount:1 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0x39800
flags: 0x1000(reserved)
==
page:000000002a7114f8 refcount:1 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0x7a200
flags: 0x1fff000000001000(reserved)
==

I actually never reproduced it until v5.9, but it's still the same bug
as it was reported first for v5.7.

See the page is "reserved" in all 3 cases. In the last two crashes
with the pfn:

pfn 0x39800 -> 0x39800000 min_pfn hit non-RAM:

39639000-39814fff : Unknown E820 type

pfn 0x7a200 -> 0x7a200000 min_pfn hit non-RAM:

7a17b000-7a216fff : Unknown E820 type

This actually seems a false positive bugcheck, the page structures are
valid and the zones are correct, just it's non-RAM but setting
pageblockskip should do no harm. However it's possible to solve the
crash without lifting the bugcheck, by enforcing the invariant that
the free_pfn cursor doesn't point to reserved pages (which would be
otherwise implicitly achieved through the PageBuddy check, except in
the new fast_isolate_around() path).

Fixes: 5a811889de10 ("mm, compaction: use free lists to quickly locate a migration target")
Signed-off-by: Andrea Arcangeli <aarcange@...hat.com>
---
 mm/compaction.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/mm/compaction.c b/mm/compaction.c
index 13cb7a961b31..d17e69549d34 100644
--- a/mm/compaction.c
+++ b/mm/compaction.c
@@ -1433,7 +1433,10 @@ fast_isolate_freepages(struct compact_control *cc)
 					page = pageblock_pfn_to_page(min_pfn,
 						pageblock_end_pfn(min_pfn),
 						cc->zone);
-					cc->free_pfn = min_pfn;
+					if (likely(!PageReserved(page)))
+						cc->free_pfn = min_pfn;
+					else
+						page = NULL;
 				}
 			}
 		}