sctp: kernel memory overwrite attempt detected in sctp_getsockopt_assoc_stats

From: Dmitry Vyukov <dvyukov@google.com>
To: Vladislav Yasevich <vyasevich@gmail.com>,
	Neil Horman <nhorman@tuxdriver.com>,
	David Miller <davem@davemloft.net>,
	linux-sctp@vger.kernel.org, netdev <netdev@vger.kernel.org>,
	LKML <linux-kernel@vger.kernel.org>,
	Kees Cook <keescook@google.com>
Cc: syzkaller <syzkaller@googlegroups.com>
Subject: sctp: kernel memory overwrite attempt detected in sctp_getsockopt_assoc_stats
Date: Sun, 15 Jan 2017 18:29:59 +0100	[thread overview]
Message-ID: <CACT4Y+ZDPa+spEh6P_ip9=hsNfeZ8KJfwJu5jTrPKVkGJUaZSg@mail.gmail.com> (raw)

Hello,

I've enabled CONFIG_HARDENED_USERCOPY_PAGESPAN on syzkaller fuzzer and
now I am seeing lots of:

usercopy: kernel memory overwrite attempt detected to ffff8801a74f6f10
(<spans multiple pages>) (256 bytes)

kernel BUG at mm/usercopy.c:75!
invalid opcode: 0000 [#1] SMP KASAN
Dumping ftrace buffer:
   (ftrace buffer empty)
Modules linked in:
CPU: 1 PID: 15686 Comm: syz-executor3 Not tainted 4.9.0 #1
Hardware name: Google Google Compute Engine/Google Compute Engine,
BIOS Google 01/01/2011
task: ffff8801c89b2500 task.stack: ffff8801a74f0000
RIP: 0010:[<ffffffff81a1b041>]  [<ffffffff81a1b041>] report_usercopy
mm/usercopy.c:67 [inline]
RIP: 0010:[<ffffffff81a1b041>]  [<ffffffff81a1b041>]
__check_object_size+0x2d1/0x9ec mm/usercopy.c:278
RSP: 0018:ffff8801a74f6cd0  EFLAGS: 00010286
RAX: 000000000000006b RBX: ffffffff84500120 RCX: 0000000000000000
RDX: 000000000000006b RSI: ffffffff815a7791 RDI: ffffed0034e9ed8c
RBP: ffff8801a74f6e48 R08: 0000000000000001 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000001 R12: ffff8801a74f6f10
R13: 0000000000000100 R14: ffffffff845000e0 R15: ffff8801a74f700f
FS:  00007f80918de700(0000) GS:ffff8801dc100000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000000020058ffc CR3: 00000001cc1cc000 CR4: 00000000001406e0
Stack:
 ffffffff8598fcc8 0000000000000000 000077ff80000000 ffffea0005c99608
 ffffffff844fff40 ffffffff844fff40 0000000041b58ab3 ffffffff84ae0fa0
 ffffffff81a1ad70 ffff8801c89b2500 dead000000000100 ffffffff814d4425
Call Trace:
 [<ffffffff83e4ece9>] check_object_size include/linux/thread_info.h:129 [inline]
 [<ffffffff83e4ece9>] copy_from_user arch/x86/include/asm/uaccess.h:692 [inline]
 [<ffffffff83e4ece9>] sctp_getsockopt_assoc_stats+0x169/0xa10
net/sctp/socket.c:6182
 [<ffffffff83e5cc52>] sctp_getsockopt+0x1af2/0x66a0 net/sctp/socket.c:6556
 [<ffffffff834f92c5>] sock_common_getsockopt+0x95/0xd0 net/core/sock.c:2649
 [<ffffffff834f4910>] SYSC_getsockopt net/socket.c:1788 [inline]
 [<ffffffff834f4910>] SyS_getsockopt+0x240/0x380 net/socket.c:1770
 [<ffffffff81009798>] do_syscall_64+0x2e8/0x930 arch/x86/entry/common.c:280
 [<ffffffff84370a49>] entry_SYSCALL64_slow_path+0x25/0x25
Code: b0 fe ff ff e8 e1 25 ce ff 48 8b 85 b0 fe ff ff 4d 89 e9 4c 89
e1 4c 89 f2 48 89 de 48 c7 c7 a0 01 50 84 49 89 c0 e8 51 d9 e0 ff <0f>
0b e8 b8 25 ce ff 4c 89 f2 4c 89 ee 4c 89 e7 e8 6a 1b fc ff
RIP  [<ffffffff81a1b041>] report_usercopy mm/usercopy.c:67 [inline]
RIP  [<ffffffff81a1b041>] __check_object_size+0x2d1/0x9ec mm/usercopy.c:278
 RSP <ffff8801a74f6cd0>
---[ end trace 5e438996b2c0b35d ]---

I am not sure why check_object_size flags this an a bug,
copy_from_user copies into a stack object:

static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
                                       char __user *optval,
                                       int __user *optlen)
{
        struct sctp_assoc_stats sas;
        len = min_t(size_t, len, sizeof(sas));
        if (copy_from_user(&sas, optval, len))
                return -EFAULT;

Kees, can this be a false positive?

On commit f4d3935e4f4884ba80561db5549394afb8eef8f7.