Re: KASAN: null-ptr-deref Read in rds_ib_get_mr

From: Yanjun Zhu
Date: Fri May 11 2018 - 20:48:18 EST




On 2018/5/12 0:58, Santosh Shilimkar wrote:
On 5/11/2018 12:48 AM, Yanjun Zhu wrote:


On 2018/5/11 13:20, DaeRyong Jeong wrote:
We report the crash: KASAN: null-ptr-deref Read in rds_ib_get_mr

Note that this bug is previously reported by syzkaller.
https://syzkaller.appspot.com/bug?id=0bb56a5a48b000b52aa2b0d8dd20b1f545214d91
Nonetheless, this bug has not fixed yet, and we hope that this report and our
analysis, which gets help by the RaceFuzzer's feature, will helpful to fix the
crash.

This crash has been found in v4.17-rc1 using RaceFuzzer (a modified
version of Syzkaller), which we describe more at the end of this
report. Our analysis shows that the race occurs when invoking two
syscalls concurrently, bind$rds and setsockopt$RDS_GET_MR.


Analysis:
We think the concurrent execution of __rds_rdma_map() and rds_bind()
causes the problem. __rds_rdma_map() checks whether rs->rs_bound_addr is 0
or not. But the concurrent execution with rds_bind() can by-pass this
check. Therefore, __rds_rdmap_map() calls rs->rs_transport->get_mr() and
rds_ib_get_mr() causes the null deref at ib_rdma.c:544 in v4.17-rc1, when
dereferencing rs_conn.


Thread interleaving:
CPU0 (__rds_rdma_map)ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ CPU1 (rds_bind)
ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ // rds_add_bound() sets rs->bound_addr as none 0
ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ ret = rds_add_bound(rs, sin->sin_addr.s_addr, &sin->sin_port);
if (rs->rs_bound_addr == 0 || !rs->rs_transport) {
ÂÂÂÂret = -ENOTCONN; /* XXX not a great errno */
ÂÂÂÂgoto out;
}
ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ if (rs->rs_transport) { /* previously bound */
ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ trans = rs->rs_transport;
ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ if (trans->laddr_check(sock_net(sock->sk),
sin->sin_addr.s_addr) != 0) {
ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ ret = -ENOPROTOOPT;
ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ // rds_remove_bound() sets rs->bound_addr as 0
ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ rds_remove_bound(rs);
...
trans_private = rs->rs_transport->get_mr(sg, nents, rs,
ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ &mr->r_key);
(in rds_ib_get_mr())
struct rds_ib_connection *ic = rs->rs_conn->c_transport_data;


Call sequence (v4.17-rc1):
CPU0
rds_setsockopt
ÂÂÂÂrds_get_mr
ÂÂÂÂÂÂÂ __rds_rdma_map
ÂÂÂÂÂÂÂÂÂÂÂ rds_ib_get_mr


CPU1
rds_bind
ÂÂÂÂrds_add_bound
ÂÂÂÂ...
ÂÂÂÂrds_remove_bound


Crash log:
==================================================================
BUG: KASAN: null-ptr-deref in rds_ib_get_mr+0x3a/0x150 net/rds/ib_rdma.c:544
Read of size 8 at addr 0000000000000068 by task syz-executor0/32067

CPU: 0 PID: 32067 Comm: syz-executor0 Not tainted 4.17.0-rc1 #1
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.8.2-0-g33fbe13 by qemu-project.org 04/01/2014
Call Trace:
 __dump_stack lib/dump_stack.c:77 [inline]
 dump_stack+0x166/0x21c lib/dump_stack.c:113
 kasan_report_error mm/kasan/report.c:352 [inline]
 kasan_report+0x140/0x360 mm/kasan/report.c:412
 check_memory_region_inline mm/kasan/kasan.c:260 [inline]
 __asan_load8+0x54/0x90 mm/kasan/kasan.c:699
 rds_ib_get_mr+0x3a/0x150 net/rds/ib_rdma.c:544
 __rds_rdma_map+0x521/0x9d0 net/rds/rdma.c:271
 rds_get_mr+0xad/0xf0 net/rds/rdma.c:333
 rds_setsockopt+0x57f/0x720 net/rds/af_rds.c:347
 __sys_setsockopt+0x147/0x230 net/socket.c:1903
 __do_sys_setsockopt net/socket.c:1914 [inline]
 __se_sys_setsockopt net/socket.c:1911 [inline]
 __x64_sys_setsockopt+0x67/0x80 net/socket.c:1911
 do_syscall_64+0x15f/0x4a0 arch/x86/entry/common.c:287
 entry_SYSCALL_64_after_hwframe+0x49/0xbe
RIP: 0033:0x4563f9
RSP: 002b:00007f6a2b3c2b28 EFLAGS: 00000246 ORIG_RAX: 0000000000000036
RAX: ffffffffffffffda RBX: 000000000072bee0 RCX: 00000000004563f9
RDX: 0000000000000002 RSI: 0000000000000114 RDI: 0000000000000015
RBP: 0000000000000575 R08: 0000000000000020 R09: 0000000000000000
R10: 0000000020000140 R11: 0000000000000246 R12: 00007f6a2b3c36d4
R13: 00000000ffffffff R14: 00000000006fd398 R15: 0000000000000000
==================================================================
diff --git a/net/rds/ib_rdma.c b/net/rds/ib_rdma.c
index e678699..2228b50 100644
--- a/net/rds/ib_rdma.c
+++ b/net/rds/ib_rdma.c
@@ -539,11 +539,17 @@ void rds_ib_flush_mrs(void)
ÂÂvoid *rds_ib_get_mr(struct scatterlist *sg, unsigned long nents,
ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ struct rds_sock *rs, u32 *key_ret)
ÂÂ{
-ÂÂÂÂÂÂ struct rds_ib_device *rds_ibdev;
+ÂÂÂÂÂÂ struct rds_ib_device *rds_ibdev = NULL;
ÂÂÂÂÂÂÂÂ struct rds_ib_mr *ibmr = NULL;
-ÂÂÂÂÂÂ struct rds_ib_connection *ic = rs->rs_conn->c_transport_data;
+ÂÂÂÂÂÂ struct rds_ib_connection *ic = NULL;
ÂÂÂÂÂÂÂÂ int ret;

+ÂÂÂÂÂÂ if (rs->rs_bound_addr == 0) {
+ÂÂÂÂÂÂÂÂÂÂÂÂÂÂ ret = -EPERM;
+ÂÂÂÂÂÂÂÂÂÂÂÂÂÂ goto out;
+ÂÂÂÂÂÂ }
+
No you can't return such error for this API and the
socket related checks needs to be done at core layer.
I remember fixing this race but probably never pushed
fix upstream.
OK. Wait for your patch. :-)

The MR code is due for update with optimized FRWR code
which now stable enough. We will address this issue as
well as part of that patchset.

Thanks for looking into it.

Regards,
Santosh