llc: do not use sk_eat_skb()
authorEric Dumazet <>
Mon, 22 Oct 2018 16:24:27 +0000 (09:24 -0700)
committerDavid S. Miller <>
Tue, 23 Oct 2018 02:59:20 +0000 (19:59 -0700)
syzkaller triggered a use-after-free [1], caused by a combination of
skb_get() in llc_conn_state_process() and usage of sk_eat_skb()

sk_eat_skb() is assuming the skb about to be freed is only used by
the current thread. TCP/DCCP stacks enforce this because current
thread holds the socket lock.

llc_conn_state_process() wants to make sure skb does not disappear,
and holds a reference on the skb it manipulates. But as soon as this
skb is added to socket receive queue, another thread can consume it.

This means that llc must use regular skb_unlink() and kfree_skb()
so that both producer and consumer can safely work on the same skb.

BUG: KASAN: use-after-free in atomic_read include/asm-generic/atomic-instrumented.h:21 [inline]
BUG: KASAN: use-after-free in refcount_read include/linux/refcount.h:43 [inline]
BUG: KASAN: use-after-free in skb_unref include/linux/skbuff.h:967 [inline]
BUG: KASAN: use-after-free in kfree_skb+0xb7/0x580 net/core/skbuff.c:655
Read of size 4 at addr ffff8801d1f6fba4 by task ksoftirqd/1/18

CPU: 1 PID: 18 Comm: ksoftirqd/1 Not tainted 4.19.0-rc8+ #295
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+0x1c4/0x2b6 lib/dump_stack.c:113
 print_address_description.cold.8+0x9/0x1ff mm/kasan/report.c:256
 kasan_report_error mm/kasan/report.c:354 [inline]
 kasan_report.cold.9+0x242/0x309 mm/kasan/report.c:412
 check_memory_region_inline mm/kasan/kasan.c:260 [inline]
 check_memory_region+0x13e/0x1b0 mm/kasan/kasan.c:267
 kasan_check_read+0x11/0x20 mm/kasan/kasan.c:272
 atomic_read include/asm-generic/atomic-instrumented.h:21 [inline]
 refcount_read include/linux/refcount.h:43 [inline]
 skb_unref include/linux/skbuff.h:967 [inline]
 kfree_skb+0xb7/0x580 net/core/skbuff.c:655
 llc_sap_state_process+0x9b/0x550 net/llc/llc_sap.c:224
 llc_sap_rcv+0x156/0x1f0 net/llc/llc_sap.c:297
 llc_sap_handler+0x65e/0xf80 net/llc/llc_sap.c:438
 llc_rcv+0x79e/0xe20 net/llc/llc_input.c:208
 __netif_receive_skb_one_core+0x14d/0x200 net/core/dev.c:4913
 __netif_receive_skb+0x2c/0x1e0 net/core/dev.c:5023
 process_backlog+0x218/0x6f0 net/core/dev.c:5829
 napi_poll net/core/dev.c:6249 [inline]
 net_rx_action+0x7c5/0x1950 net/core/dev.c:6315
 __do_softirq+0x30c/0xb03 kernel/softirq.c:292
 run_ksoftirqd+0x94/0x100 kernel/softirq.c:653
 smpboot_thread_fn+0x68b/0xa00 kernel/smpboot.c:164
 kthread+0x35a/0x420 kernel/kthread.c:246
 ret_from_fork+0x3a/0x50 arch/x86/entry/entry_64.S:413

Allocated by task 18:
 save_stack+0x43/0xd0 mm/kasan/kasan.c:448
 set_track mm/kasan/kasan.c:460 [inline]
 kasan_kmalloc+0xc7/0xe0 mm/kasan/kasan.c:553
 kasan_slab_alloc+0x12/0x20 mm/kasan/kasan.c:490
 kmem_cache_alloc_node+0x144/0x730 mm/slab.c:3644
 __alloc_skb+0x119/0x770 net/core/skbuff.c:193
 alloc_skb include/linux/skbuff.h:995 [inline]
 llc_alloc_frame+0xbc/0x370 net/llc/llc_sap.c:54
 llc_station_ac_send_xid_r net/llc/llc_station.c:52 [inline]
 llc_station_rcv+0x1dc/0x1420 net/llc/llc_station.c:111
 llc_rcv+0xc32/0xe20 net/llc/llc_input.c:220
 __netif_receive_skb_one_core+0x14d/0x200 net/core/dev.c:4913
 __netif_receive_skb+0x2c/0x1e0 net/core/dev.c:5023
 process_backlog+0x218/0x6f0 net/core/dev.c:5829
 napi_poll net/core/dev.c:6249 [inline]
 net_rx_action+0x7c5/0x1950 net/core/dev.c:6315
 __do_softirq+0x30c/0xb03 kernel/softirq.c:292

Freed by task 16383:
 save_stack+0x43/0xd0 mm/kasan/kasan.c:448
 set_track mm/kasan/kasan.c:460 [inline]
 __kasan_slab_free+0x102/0x150 mm/kasan/kasan.c:521
 kasan_slab_free+0xe/0x10 mm/kasan/kasan.c:528
 __cache_free mm/slab.c:3498 [inline]
 kmem_cache_free+0x83/0x290 mm/slab.c:3756
 kfree_skbmem+0x154/0x230 net/core/skbuff.c:582
 __kfree_skb+0x1d/0x20 net/core/skbuff.c:642
 sk_eat_skb include/net/sock.h:2366 [inline]
 llc_ui_recvmsg+0xec2/0x1610 net/llc/af_llc.c:882
 sock_recvmsg_nosec net/socket.c:794 [inline]
 sock_recvmsg+0xd0/0x110 net/socket.c:801
 ___sys_recvmsg+0x2b6/0x680 net/socket.c:2278
 __sys_recvmmsg+0x303/0xb90 net/socket.c:2390
 do_sys_recvmmsg+0x181/0x1a0 net/socket.c:2466
 __do_sys_recvmmsg net/socket.c:2484 [inline]
 __se_sys_recvmmsg net/socket.c:2480 [inline]
 __x64_sys_recvmmsg+0xbe/0x150 net/socket.c:2480
 do_syscall_64+0x1b9/0x820 arch/x86/entry/common.c:290

The buggy address belongs to the object at ffff8801d1f6fac0
 which belongs to the cache skbuff_head_cache of size 232
The buggy address is located 228 bytes inside of
 232-byte region [ffff8801d1f6fac0ffff8801d1f6fba8)
The buggy address belongs to the page:
page:ffffea000747dbc0 count:1 mapcount:0 mapping:ffff8801d9be7680 index:0xffff8801d1f6fe80
flags: 0x2fffc0000000100(slab)
raw: 02fffc0000000100 ffffea0007346e88 ffffea000705b108 ffff8801d9be7680
raw: ffff8801d1f6fe80 ffff8801d1f6f0c0 000000010000000b 0000000000000000
page dumped because: kasan: bad access detected

Memory state around the buggy address:
 ffff8801d1f6fa80: fc fc fc fc fc fc fc fc fb fb fb fb fb fb fb fb
 ffff8801d1f6fb00: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
>ffff8801d1f6fb80: fb fb fb fb fb fc fc fc fc fc fc fc fc fc fc fc
 ffff8801d1f6fc00: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
 ffff8801d1f6fc80: fb fb fb fb fb fb fb fb fb fb fb fb fb fc fc fc

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Eric Dumazet <>
Reported-by: syzbot <>
Signed-off-by: David S. Miller <>

index 1beeea9549fa6ec1f7b0e5f9af8ff3250a316f59..b99e73a7e7e0f2b4959b279e3aecbadf29667d55 100644 (file)
@@ -730,7 +730,6 @@ static int llc_ui_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
        struct sk_buff *skb = NULL;
        struct sock *sk = sock->sk;
        struct llc_sock *llc = llc_sk(sk);
-       unsigned long cpu_flags;
        size_t copied = 0;
        u32 peek_seq = 0;
        u32 *seq, skb_len;
@@ -855,9 +854,8 @@ static int llc_ui_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
                        goto copy_uaddr;
                if (!(flags & MSG_PEEK)) {
-                       spin_lock_irqsave(&sk->sk_receive_queue.lock, cpu_flags);
-                       sk_eat_skb(sk, skb);
-                       spin_unlock_irqrestore(&sk->sk_receive_queue.lock, cpu_flags);
+                       skb_unlink(skb, &sk->sk_receive_queue);
+                       kfree_skb(skb);
                        *seq = 0;
@@ -878,9 +876,8 @@ copy_uaddr:
                llc_cmsg_rcv(msg, skb);
        if (!(flags & MSG_PEEK)) {
-               spin_lock_irqsave(&sk->sk_receive_queue.lock, cpu_flags);
-               sk_eat_skb(sk, skb);
-               spin_unlock_irqrestore(&sk->sk_receive_queue.lock, cpu_flags);
+               skb_unlink(skb, &sk->sk_receive_queue);
+               kfree_skb(skb);
                *seq = 0;