icmp: guard against too small mtu
authorEric Dumazet <edumazet@google.com>
Thu, 30 Mar 2023 17:45:02 +0000 (17:45 +0000)
committerJakub Kicinski <kuba@kernel.org>
Sat, 1 Apr 2023 04:37:06 +0000 (21:37 -0700)
syzbot was able to trigger a panic [1] in icmp_glue_bits(), or
more exactly in skb_copy_and_csum_bits()

There is no repro yet, but I think the issue is that syzbot
manages to lower device mtu to a small value, fooling __icmp_send()

__icmp_send() must make sure there is enough room for the
packet to include at least the headers.

We might in the future refactor skb_copy_and_csum_bits() and its
callers to no longer crash when something bad happens.

[1]
kernel BUG at net/core/skbuff.c:3343 !
invalid opcode: 0000 [#1] PREEMPT SMP KASAN
CPU: 0 PID: 15766 Comm: syz-executor.0 Not tainted 6.3.0-rc4-syzkaller-00039-gffe78bbd5121 #0
Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.14.0-2 04/01/2014
RIP: 0010:skb_copy_and_csum_bits+0x798/0x860 net/core/skbuff.c:3343
Code: f0 c1 c8 08 41 89 c6 e9 73 ff ff ff e8 61 48 d4 f9 e9 41 fd ff ff 48 8b 7c 24 48 e8 52 48 d4 f9 e9 c3 fc ff ff e8 c8 27 84 f9 <0f> 0b 48 89 44 24 28 e8 3c 48 d4 f9 48 8b 44 24 28 e9 9d fb ff ff
RSP: 0018:ffffc90000007620 EFLAGS: 00010246
RAX: 0000000000000000 RBX: 00000000000001e8 RCX: 0000000000000100
RDX: ffff8880276f6280 RSI: ffffffff87fdd138 RDI: 0000000000000005
RBP: 0000000000000000 R08: 0000000000000005 R09: 0000000000000000
R10: 00000000000001e8 R11: 0000000000000001 R12: 000000000000003c
R13: 0000000000000000 R14: ffff888028244868 R15: 0000000000000b0e
FS: 00007fbc81f1c700(0000) GS:ffff88802ca00000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000001b2df43000 CR3: 00000000744db000 CR4: 0000000000150ef0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
Call Trace:
<IRQ>
icmp_glue_bits+0x7b/0x210 net/ipv4/icmp.c:353
__ip_append_data+0x1d1b/0x39f0 net/ipv4/ip_output.c:1161
ip_append_data net/ipv4/ip_output.c:1343 [inline]
ip_append_data+0x115/0x1a0 net/ipv4/ip_output.c:1322
icmp_push_reply+0xa8/0x440 net/ipv4/icmp.c:370
__icmp_send+0xb80/0x1430 net/ipv4/icmp.c:765
ipv4_send_dest_unreach net/ipv4/route.c:1239 [inline]
ipv4_link_failure+0x5a9/0x9e0 net/ipv4/route.c:1246
dst_link_failure include/net/dst.h:423 [inline]
arp_error_report+0xcb/0x1c0 net/ipv4/arp.c:296
neigh_invalidate+0x20d/0x560 net/core/neighbour.c:1079
neigh_timer_handler+0xc77/0xff0 net/core/neighbour.c:1166
call_timer_fn+0x1a0/0x580 kernel/time/timer.c:1700
expire_timers+0x29b/0x4b0 kernel/time/timer.c:1751
__run_timers kernel/time/timer.c:2022 [inline]

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Reported-by: syzbot+d373d60fddbdc915e666@syzkaller.appspotmail.com
Signed-off-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/20230330174502.1915328-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
net/ipv4/icmp.c

index 8cebb47..b860776 100644 (file)
@@ -749,6 +749,11 @@ void __icmp_send(struct sk_buff *skb_in, int type, int code, __be32 info,
                room = 576;
        room -= sizeof(struct iphdr) + icmp_param.replyopts.opt.opt.optlen;
        room -= sizeof(struct icmphdr);
+       /* Guard against tiny mtu. We need to include at least one
+        * IP network header for this message to make any sense.
+        */
+       if (room <= (int)sizeof(struct iphdr))
+               goto ende;
 
        icmp_param.data_len = skb_in->len - icmp_param.offset;
        if (icmp_param.data_len > room)