projects / GitHub / LineageOS / G12 / android_kernel_amlogic_linux-4.9.git / commitdiff
? search:
summary | shortlog | log | commit | commitdiff | tree
raw | patch | inline | side by side (parent: cbdbd2a)
can: raw: raw_setsockopt(): fix raw_rcv panic for sock UAF
authorZiyang Xuan <william.xuanziyang@huawei.com>
Thu, 22 Jul 2021 07:08:19 +0000 (15:08 +0800)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sun, 8 Aug 2021 06:38:54 +0000 (08:38 +0200)
commit 54f93336d000229f72c26d8a3f69dd256b744528 upstream.

We get a bug during ltp can_filter test as following.

===========================================
[60919.264984] BUG: unable to handle kernel NULL pointer dereference at 0000000000000010
[60919.265223] PGD 8000003dda726067 P4D 8000003dda726067 PUD 3dda727067 PMD 0
[60919.265443] Oops: 0000 [#1] SMP PTI
[60919.265550] CPU: 30 PID: 3638365 Comm: can_filter Kdump: loaded Tainted: G        W         4.19.90+ #1
[60919.266068] RIP: 0010:selinux_socket_sock_rcv_skb+0x3e/0x200
[60919.293289] RSP: 0018:ffff8d53bfc03cf8 EFLAGS: 00010246
[60919.307140] RAX: 0000000000000000 RBX: 000000000000001d RCX: 0000000000000007
[60919.320756] RDX: 0000000000000001 RSI: ffff8d5104a8ed00 RDI: ffff8d53bfc03d30
[60919.334319] RBP: ffff8d9338056800 R08: ffff8d53bfc29d80 R09: 0000000000000001
[60919.347969] R10: ffff8d53bfc03ec0 R11: ffffb8526ef47c98 R12: ffff8d53bfc03d30
[60919.350320] perf: interrupt took too long (3063 > 2500), lowering kernel.perf_event_max_sample_rate to 65000
[60919.361148] R13: 0000000000000001 R14: ffff8d53bcf90000 R15: 0000000000000000
[60919.361151] FS:  00007fb78b6b3600(0000) GS:ffff8d53bfc00000(0000) knlGS:0000000000000000
[60919.400812] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[60919.413730] CR2: 0000000000000010 CR3: 0000003e3f784006 CR4: 00000000007606e0
[60919.426479] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[60919.439339] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[60919.451608] PKRU: 55555554
[60919.463622] Call Trace:
[60919.475617]  <IRQ>
[60919.487122]  ? update_load_avg+0x89/0x5d0
[60919.498478]  ? update_load_avg+0x89/0x5d0
[60919.509822]  ? account_entity_enqueue+0xc5/0xf0
[60919.520709]  security_sock_rcv_skb+0x2a/0x40
[60919.531413]  sk_filter_trim_cap+0x47/0x1b0
[60919.542178]  ? kmem_cache_alloc+0x38/0x1b0
[60919.552444]  sock_queue_rcv_skb+0x17/0x30
[60919.562477]  raw_rcv+0x110/0x190 [can_raw]
[60919.572539]  can_rcv_filter+0xbc/0x1b0 [can]
[60919.582173]  can_receive+0x6b/0xb0 [can]
[60919.591595]  can_rcv+0x31/0x70 [can]
[60919.600783]  __netif_receive_skb_one_core+0x5a/0x80
[60919.609864]  process_backlog+0x9b/0x150
[60919.618691]  net_rx_action+0x156/0x400
[60919.627310]  ? sched_clock_cpu+0xc/0xa0
[60919.635714]  __do_softirq+0xe8/0x2e9
[60919.644161]  do_softirq_own_stack+0x2a/0x40
[60919.652154]  </IRQ>
[60919.659899]  do_softirq.part.17+0x4f/0x60
[60919.667475]  __local_bh_enable_ip+0x60/0x70
[60919.675089]  __dev_queue_xmit+0x539/0x920
[60919.682267]  ? finish_wait+0x80/0x80
[60919.689218]  ? finish_wait+0x80/0x80
[60919.695886]  ? sock_alloc_send_pskb+0x211/0x230
[60919.702395]  ? can_send+0xe5/0x1f0 [can]
[60919.708882]  can_send+0xe5/0x1f0 [can]
[60919.715037]  raw_sendmsg+0x16d/0x268 [can_raw]

It's because raw_setsockopt() concurrently with
unregister_netdevice_many(). Concurrent scenario as following.

cpu0 cpu1
raw_bind
raw_setsockopt unregister_netdevice_many
unlist_netdevice
dev_get_by_index raw_notifier
raw_enable_filters ......
can_rx_register
can_rcv_list_find(..., net->can.rx_alldev_list)

......

sock_close
raw_release(sock_a)

......

can_receive
can_rcv_filter(net->can.rx_alldev_list, ...)
raw_rcv(skb, sock_a)
BUG

After unlist_netdevice(), dev_get_by_index() return NULL in
raw_setsockopt(). Function raw_enable_filters() will add sock
and can_filter to net->can.rx_alldev_list. Then the sock is closed.
Followed by, we sock_sendmsg() to a new vcan device use the same
can_filter. Protocol stack match the old receiver whose sock has
been released on net->can.rx_alldev_list in can_rcv_filter().
Function raw_rcv() uses the freed sock. UAF BUG is triggered.

We can find that the key issue is that net_device has not been
protected in raw_setsockopt(). Use rtnl_lock to protect net_device
in raw_setsockopt().

Fixes: c18ce101f2e4 ("[CAN]: Add raw protocol")
Link: https://lore.kernel.org/r/20210722070819.1048263-1-william.xuanziyang@huawei.com
Cc: linux-stable <stable@vger.kernel.org>
Signed-off-by: Ziyang Xuan <william.xuanziyang@huawei.com>
Acked-by: Oliver Hartkopp <socketcan@hartkopp.net>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
net/can/raw.c patch | blob | blame | history

diff --git a/net/can/raw.c b/net/can/raw.c
index 2bb50b1535c2fb7297ece9efb8333acd9d8095c0..082965c8dcaf6522cb1b8d46c95b449824655b34 100644 (file)
--- a/net/can/raw.c
+++ b/net/can/raw.c
@@ -541,10 +541,18 @@ static int raw_setsockopt(struct socket *sock, int level, int optname,
                                return -EFAULT;
                }
 
+               rtnl_lock();
                lock_sock(sk);
 
-               if (ro->bound && ro->ifindex)
+               if (ro->bound && ro->ifindex) {
                        dev = dev_get_by_index(&init_net, ro->ifindex);
+                       if (!dev) {
+                               if (count > 1)
+                                       kfree(filter);
+                               err = -ENODEV;
+                               goto out_fil;
+                       }
+               }
 
                if (ro->bound) {
                        /* (try to) register the new filters */
@@ -581,6 +589,7 @@ static int raw_setsockopt(struct socket *sock, int level, int optname,
                        dev_put(dev);
 
                release_sock(sk);
+               rtnl_unlock();
 
                break;
 
@@ -593,10 +602,16 @@ static int raw_setsockopt(struct socket *sock, int level, int optname,
 
                err_mask &= CAN_ERR_MASK;
 
+               rtnl_lock();
                lock_sock(sk);
 
-               if (ro->bound && ro->ifindex)
+               if (ro->bound && ro->ifindex) {
                        dev = dev_get_by_index(&init_net, ro->ifindex);
+                       if (!dev) {
+                               err = -ENODEV;
+                               goto out_err;
+                       }
+               }
 
                /* remove current error mask */
                if (ro->bound) {
@@ -618,6 +633,7 @@ static int raw_setsockopt(struct socket *sock, int level, int optname,
                        dev_put(dev);
 
                release_sock(sk);
+               rtnl_unlock();
 
                break;