tcp: fix slab-out-of-bounds in tcp_zerocopy_receive()
authorEric Dumazet <edumazet@google.com>
Thu, 3 Oct 2019 03:19:59 +0000 (20:19 -0700)
committerDavid S. Miller <davem@davemloft.net>
Thu, 3 Oct 2019 19:05:34 +0000 (12:05 -0700)
Apparently a refactoring patch brought a bug, that was caught
by syzbot [1]

Original code was correct, do not try to be smarter than the
compiler :/

[1]
BUG: KASAN: slab-out-of-bounds in tcp_zerocopy_receive net/ipv4/tcp.c:1807 [inline]
BUG: KASAN: slab-out-of-bounds in do_tcp_getsockopt.isra.0+0x2c6c/0x3120 net/ipv4/tcp.c:3654
Read of size 4 at addr ffff8880943cf188 by task syz-executor.2/17508

CPU: 0 PID: 17508 Comm: syz-executor.2 Not tainted 5.3.0-rc7+ #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
Call Trace:
 __dump_stack lib/dump_stack.c:77 [inline]
 dump_stack+0x172/0x1f0 lib/dump_stack.c:113
 print_address_description.cold+0xd4/0x306 mm/kasan/report.c:351
 __kasan_report.cold+0x1b/0x36 mm/kasan/report.c:482
 kasan_report+0x12/0x17 mm/kasan/common.c:618
 __asan_report_load4_noabort+0x14/0x20 mm/kasan/generic_report.c:131
 tcp_zerocopy_receive net/ipv4/tcp.c:1807 [inline]
 do_tcp_getsockopt.isra.0+0x2c6c/0x3120 net/ipv4/tcp.c:3654
 tcp_getsockopt+0xbf/0xe0 net/ipv4/tcp.c:3680
 sock_common_getsockopt+0x94/0xd0 net/core/sock.c:3098
 __sys_getsockopt+0x16d/0x310 net/socket.c:2129
 __do_sys_getsockopt net/socket.c:2144 [inline]
 __se_sys_getsockopt net/socket.c:2141 [inline]
 __x64_sys_getsockopt+0xbe/0x150 net/socket.c:2141
 do_syscall_64+0xfd/0x6a0 arch/x86/entry/common.c:296

Fixes: d8e18a516f8f ("net: Use skb accessors in network core")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/ipv4/tcp.c

index 79c325a07ba5dc7cfad0a846d1f03bf1787f840b..f98a1882e537dca0102e829cb349be50302d83ab 100644 (file)
@@ -1798,13 +1798,11 @@ static int tcp_zerocopy_receive(struct sock *sk,
                }
                if (skb_frag_size(frags) != PAGE_SIZE || skb_frag_off(frags)) {
                        int remaining = zc->recv_skip_hint;
-                       int size = skb_frag_size(frags);
 
-                       while (remaining && (size != PAGE_SIZE ||
+                       while (remaining && (skb_frag_size(frags) != PAGE_SIZE ||
                                             skb_frag_off(frags))) {
-                               remaining -= size;
+                               remaining -= skb_frag_size(frags);
                                frags++;
-                               size = skb_frag_size(frags);
                        }
                        zc->recv_skip_hint -= remaining;
                        break;