| Message ID | 20230123172233.GA112231@ubuntu (mailing list archive) |
|---|---|
| State | Superseded |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | [v2] netrom: Fix use-after-free caused by accept on already connected socket | expand |
Hi, I think the diff looks good but changelog is not correct. From: Hyunwoo Kim <v4bel@theori.io> Date: Mon, 23 Jan 2023 09:22:33 -0800 > If listen() and accept() are called on an AF_NETROM socket that > has already been connect()ed, accept() succeeds in connecting. > This is because nr_accept() dequeues the skb queued in > `sk->sk_receive_queue` in nr_connect(). This sentence is misleading. The skb sent by nr_connect() is queued up for the peer's sk_receive_queue (socket_2 below), and also the reply is not queued for connect()er (socket_1) because it is NR_CONNACK and just consumed by nr_state1_machine() to transition to NR_STATE_3 and TCP_ESTABLISHED. > > This causes nr_accept() to allocate and return a sock with the > sk of the parent AF_NETROM socket. This happens because the peer (socket_2) sends some data by nr_sendmsg(), whose type is NR_INFO and the data is queued up for the connect()er's (socket_1) sk_receive_queue by - nr_state3_machine - nr_queue_rx_frame - sock_queue_rcv_skb > And here's where use-after-free > can happen through complex race conditions: > ``` > cpu0 cpu1 > 1. socket_2 = socket(AF_NETROM) > listen(socket_2) > accepted_socket = accept(socket_2) // loopback connection with socket_1 > 2. socket_1 = socket(AF_NETROM) > nr_create() // sk refcount : 1 > connect(socket_1) // loopback connection with socket_2 > nr_connect() > nr_establish_data_link() > nr_write_internal() This is NR_CONNREQ for socket_2. > nr_transmit_buffer() > nr_route_frame() > nr_loopback_queue() > nr_loopback_timer() > nr_rx_frame() > nr_process_rx_frame() So, nr_process_rx_frame() is not called from nr_rx_frame() because the kernel finds socket_2 by nr_find_listener(), and creates accepted_socket, and then responds with NR_CONNACK to socket_1. And accepted_socket sends some data with NR_INFO, then, nr_state3_machine() is called for socket_1 and the skb with sk pointing socket_1 is queued for socket_1's sk_receive_queue. I think this part is lacking in this UAF scenario. What do you think ? > nr_state3_machine() > nr_queue_rx_frame() > sock_queue_rcv_skb() > sock_queue_rcv_skb_reason() > __sock_queue_rcv_skb() > __skb_queue_tail(list, skb); // list : sk->sk_receive_queue > > 3. listen(socket_1) > nr_listen() > uaf_socket = accept(socket_1) > nr_accept() > skb_dequeue(&sk->sk_receive_queue); > 4. close(accepted_socket) And I guess this close() is not so important in this case. Even if we don't close the socket cleanly, the UAF will happen, right ? (it might be caused in a different place like timer though) > nr_release() > nr_write_internal(sk, NR_DISCREQ) > nr_transmit_buffer() // NR_DISCREQ > nr_route_frame() > nr_loopback_queue() > nr_loopback_timer() > nr_rx_frame() // sk : socket_1's sk > nr_process_rx_frame() // NR_STATE_3 > nr_state3_machine() // NR_DISCREQ > nr_disconnect() > nr_sk(sk)->state = NR_STATE_0; > 5. close(socket_1) // sk refcount : 3 > nr_release() // NR_STATE_0 > sock_put(sk); // sk refcount : 0 > sk_free(sk); > close(uaf_socket) > nr_release() > sock_hold(sk); // UAF > ``` > > KASAN report by syzbot: > ``` > BUG: KASAN: use-after-free in nr_release+0x66/0x460 net/netrom/af_netrom.c:520 > Write of size 4 at addr ffff8880235d8080 by task syz-executor564/5128 > > Call Trace: > <TASK> > __dump_stack lib/dump_stack.c:88 [inline] > dump_stack_lvl+0xd1/0x138 lib/dump_stack.c:106 > print_address_description mm/kasan/report.c:306 [inline] > print_report+0x15e/0x461 mm/kasan/report.c:417 > kasan_report+0xbf/0x1f0 mm/kasan/report.c:517 > check_region_inline mm/kasan/generic.c:183 [inline] > kasan_check_range+0x141/0x190 mm/kasan/generic.c:189 > instrument_atomic_read_write include/linux/instrumented.h:102 [inline] > atomic_fetch_add_relaxed include/linux/atomic/atomic-instrumented.h:116 [inline] > __refcount_add include/linux/refcount.h:193 [inline] > __refcount_inc include/linux/refcount.h:250 [inline] > refcount_inc include/linux/refcount.h:267 [inline] > sock_hold include/net/sock.h:775 [inline] > nr_release+0x66/0x460 net/netrom/af_netrom.c:520 > __sock_release+0xcd/0x280 net/socket.c:650 > sock_close+0x1c/0x20 net/socket.c:1365 > __fput+0x27c/0xa90 fs/file_table.c:320 > task_work_run+0x16f/0x270 kernel/task_work.c:179 > exit_task_work include/linux/task_work.h:38 [inline] > do_exit+0xaa8/0x2950 kernel/exit.c:867 > do_group_exit+0xd4/0x2a0 kernel/exit.c:1012 > get_signal+0x21c3/0x2450 kernel/signal.c:2859 > arch_do_signal_or_restart+0x79/0x5c0 arch/x86/kernel/signal.c:306 > exit_to_user_mode_loop kernel/entry/common.c:168 [inline] > exit_to_user_mode_prepare+0x15f/0x250 kernel/entry/common.c:203 > __syscall_exit_to_user_mode_work kernel/entry/common.c:285 [inline] > syscall_exit_to_user_mode+0x1d/0x50 kernel/entry/common.c:296 > do_syscall_64+0x46/0xb0 arch/x86/entry/common.c:86 > entry_SYSCALL_64_after_hwframe+0x63/0xcd > RIP: 0033:0x7f6c19e3c9b9 > Code: Unable to access opcode bytes at 0x7f6c19e3c98f. > RSP: 002b:00007fffd4ba2ce8 EFLAGS: 00000246 ORIG_RAX: 0000000000000133 > RAX: 0000000000000116 RBX: 0000000000000003 RCX: 00007f6c19e3c9b9 > RDX: 0000000000000318 RSI: 00000000200bd000 RDI: 0000000000000006 > RBP: 0000000000000003 R08: 000000000000000d R09: 000000000000000d > R10: 0000000000000000 R11: 0000000000000246 R12: 000055555566a2c0 > R13: 0000000000000011 R14: 0000000000000000 R15: 0000000000000000 > </TASK> > > Allocated by task 5128: > kasan_save_stack+0x22/0x40 mm/kasan/common.c:45 > kasan_set_track+0x25/0x30 mm/kasan/common.c:52 > ____kasan_kmalloc mm/kasan/common.c:371 [inline] > ____kasan_kmalloc mm/kasan/common.c:330 [inline] > __kasan_kmalloc+0xa3/0xb0 mm/kasan/common.c:380 > kasan_kmalloc include/linux/kasan.h:211 [inline] > __do_kmalloc_node mm/slab_common.c:968 [inline] > __kmalloc+0x5a/0xd0 mm/slab_common.c:981 > kmalloc include/linux/slab.h:584 [inline] > sk_prot_alloc+0x140/0x290 net/core/sock.c:2038 > sk_alloc+0x3a/0x7a0 net/core/sock.c:2091 > nr_create+0xb6/0x5f0 net/netrom/af_netrom.c:433 > __sock_create+0x359/0x790 net/socket.c:1515 > sock_create net/socket.c:1566 [inline] > __sys_socket_create net/socket.c:1603 [inline] > __sys_socket_create net/socket.c:1588 [inline] > __sys_socket+0x133/0x250 net/socket.c:1636 > __do_sys_socket net/socket.c:1649 [inline] > __se_sys_socket net/socket.c:1647 [inline] > __x64_sys_socket+0x73/0xb0 net/socket.c:1647 > do_syscall_x64 arch/x86/entry/common.c:50 [inline] > do_syscall_64+0x39/0xb0 arch/x86/entry/common.c:80 > entry_SYSCALL_64_after_hwframe+0x63/0xcd > > Freed by task 5128: > kasan_save_stack+0x22/0x40 mm/kasan/common.c:45 > kasan_set_track+0x25/0x30 mm/kasan/common.c:52 > kasan_save_free_info+0x2b/0x40 mm/kasan/generic.c:518 > ____kasan_slab_free mm/kasan/common.c:236 [inline] > ____kasan_slab_free+0x13b/0x1a0 mm/kasan/common.c:200 > kasan_slab_free include/linux/kasan.h:177 [inline] > __cache_free mm/slab.c:3394 [inline] > __do_kmem_cache_free mm/slab.c:3580 [inline] > __kmem_cache_free+0xcd/0x3b0 mm/slab.c:3587 > sk_prot_free net/core/sock.c:2074 [inline] > __sk_destruct+0x5df/0x750 net/core/sock.c:2166 > sk_destruct net/core/sock.c:2181 [inline] > __sk_free+0x175/0x460 net/core/sock.c:2192 > sk_free+0x7c/0xa0 net/core/sock.c:2203 > sock_put include/net/sock.h:1991 [inline] > nr_release+0x39e/0x460 net/netrom/af_netrom.c:554 > __sock_release+0xcd/0x280 net/socket.c:650 > sock_close+0x1c/0x20 net/socket.c:1365 > __fput+0x27c/0xa90 fs/file_table.c:320 > task_work_run+0x16f/0x270 kernel/task_work.c:179 > exit_task_work include/linux/task_work.h:38 [inline] > do_exit+0xaa8/0x2950 kernel/exit.c:867 > do_group_exit+0xd4/0x2a0 kernel/exit.c:1012 > get_signal+0x21c3/0x2450 kernel/signal.c:2859 > arch_do_signal_or_restart+0x79/0x5c0 arch/x86/kernel/signal.c:306 > exit_to_user_mode_loop kernel/entry/common.c:168 [inline] > exit_to_user_mode_prepare+0x15f/0x250 kernel/entry/common.c:203 > __syscall_exit_to_user_mode_work kernel/entry/common.c:285 [inline] > syscall_exit_to_user_mode+0x1d/0x50 kernel/entry/common.c:296 > do_syscall_64+0x46/0xb0 arch/x86/entry/common.c:86 > entry_SYSCALL_64_after_hwframe+0x63/0xcd > ``` > > To fix this problem, nr_listen() returns -EINVAL for sockets that successfully nr_connect(). > > Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2") > Reported-by: syzbot+caa188bdfc1eeafeb418@syzkaller.appspotmail.com > Signed-off-by: Hyunwoo Kim <v4bel@theori.io> > --- > net/netrom/af_netrom.c | 5 +++++ > 1 file changed, 5 insertions(+) > > diff --git a/net/netrom/af_netrom.c b/net/netrom/af_netrom.c > index 6f7f4392cffb..5a4cb796150f 100644 > --- a/net/netrom/af_netrom.c > +++ b/net/netrom/af_netrom.c > @@ -400,6 +400,11 @@ static int nr_listen(struct socket *sock, int backlog) > struct sock *sk = sock->sk; > > lock_sock(sk); > + if (sock->state != SS_UNCONNECTED) { > + release_sock(sk); > + return -EINVAL; > + } > + > if (sk->sk_state != TCP_LISTEN) { > memset(&nr_sk(sk)->user_addr, 0, AX25_ADDR_LEN); > sk->sk_max_ack_backlog = backlog; > -- > 2.25.1
On Tue, Jan 24, 2023 at 05:43:47PM -0800, Kuniyuki Iwashima wrote: > Hi, > > I think the diff looks good but changelog is not correct. > > From: Hyunwoo Kim <v4bel@theori.io> > Date: Mon, 23 Jan 2023 09:22:33 -0800 > > If listen() and accept() are called on an AF_NETROM socket that > > has already been connect()ed, accept() succeeds in connecting. > > This is because nr_accept() dequeues the skb queued in > > `sk->sk_receive_queue` in nr_connect(). > > This sentence is misleading. The skb sent by nr_connect() is queued > up for the peer's sk_receive_queue (socket_2 below), and also the reply > is not queued for connect()er (socket_1) because it is NR_CONNACK and > just consumed by nr_state1_machine() to transition to NR_STATE_3 and > TCP_ESTABLISHED. > > > > > > This causes nr_accept() to allocate and return a sock with the > > sk of the parent AF_NETROM socket. > > This happens because the peer (socket_2) sends some data by nr_sendmsg(), > whose type is NR_INFO and the data is queued up for the connect()er's > (socket_1) sk_receive_queue by > > - nr_state3_machine > - nr_queue_rx_frame > - sock_queue_rcv_skb you're right. I misanalyzed nr_rx_frame(). sorry. > > > > And here's where use-after-free > > can happen through complex race conditions: > > ``` > > cpu0 cpu1 > > 1. socket_2 = socket(AF_NETROM) > > listen(socket_2) > > accepted_socket = accept(socket_2) // loopback connection with socket_1 > > 2. socket_1 = socket(AF_NETROM) > > nr_create() // sk refcount : 1 > > connect(socket_1) // loopback connection with socket_2 > > nr_connect() > > nr_establish_data_link() > > nr_write_internal() > > This is NR_CONNREQ for socket_2. > > > > nr_transmit_buffer() > > nr_route_frame() > > nr_loopback_queue() > > nr_loopback_timer() > > nr_rx_frame() > > nr_process_rx_frame() > > So, nr_process_rx_frame() is not called from nr_rx_frame() because the > kernel finds socket_2 by nr_find_listener(), and creates accepted_socket, > and then responds with NR_CONNACK to socket_1. > > And accepted_socket sends some data with NR_INFO, then, nr_state3_machine() > is called for socket_1 and the skb with sk pointing socket_1 is queued for > socket_1's sk_receive_queue. > > I think this part is lacking in this UAF scenario. > > What do you think ? > > > > nr_state3_machine() > > nr_queue_rx_frame() > > sock_queue_rcv_skb() > > sock_queue_rcv_skb_reason() > > __sock_queue_rcv_skb() > > __skb_queue_tail(list, skb); // list : sk->sk_receive_queue > > > > 3. listen(socket_1) > > nr_listen() > > uaf_socket = accept(socket_1) > > nr_accept() > > skb_dequeue(&sk->sk_receive_queue); > > 4. close(accepted_socket) > > And I guess this close() is not so important in this case. > Even if we don't close the socket cleanly, the UAF will happen, right ? > (it might be caused in a different place like timer though) IMHO, I think this `close(accepted_socket)` is a necessary procedure. This is because `case NR_STATE_3:` in nr_release() is executed here, and nr_write_internal(sk, NR_DISCREQ) ~ nr_disconnect() sets socket_1's sk to NR_STATE_0. If sk of socket_1 is not set to NR_STATE_0, `case NR_STATE_3:` is executed instead of `case NR_STATE_0:` in `close(socket_1)` and sk is not released. This means you will not be able to trigger UAF on close(uaf_socket). What do you think? So the modified flow chart looks like this: ``` cpu0 cpu1 1. socket_2 = socket(AF_NETROM) . . listen(socket_2) accepted_socket = accept(socket_2) 2. socket_1 = socket(AF_NETROM) nr_create() // sk refcount : 1 connect(socket_1) nr_connect() nr_establish_data_link() nr_write_internal() nr_transmit_buffer() nr_route_frame() nr_loopback_queue() nr_loopback_timer() nr_rx_frame() nr_process_rx_frame() nr_state1_machine() nr->state = NR_STATE_3; 3. write(accepted_socket) nr_sendmsg() nr_output() nr_kick() nr_send_iframe() nr_transmit_buffer() nr_route_frame() nr_loopback_queue() nr_loopback_timer() nr_process_rx_frame(sk, skb); // sk : socket_1's sk nr_state3_machine() nr_queue_rx_frame() sock_queue_rcv_skb() sock_queue_rcv_skb_reason() __sock_queue_rcv_skb() __skb_queue_tail(list, skb); // list : socket_1's sk->sk_receive_queue 4. listen(socket_1) nr_listen() uaf_socket = accept(socket_1) nr_accept() skb_dequeue(&sk->sk_receive_queue); 5. close(accepted_socket) nr_release() nr_write_internal(sk, NR_DISCREQ) nr_transmit_buffer() // NR_DISCREQ nr_route_frame() nr_loopback_queue() nr_loopback_timer() nr_rx_frame() // sk : socket_1's sk nr_process_rx_frame() // NR_STATE_3 nr_state3_machine() // NR_DISCREQ nr_disconnect() nr_sk(sk)->state = NR_STATE_0; 6. close(socket_1) // sk refcount : 3 nr_release() // NR_STATE_0 sock_put(sk); // sk refcount : 0 sk_free(sk); close(uaf_socket) nr_release() sock_hold(sk); // UAF ``` Anything wrong? > > > nr_release() > > nr_write_internal(sk, NR_DISCREQ) > > nr_transmit_buffer() // NR_DISCREQ > > nr_route_frame() > > nr_loopback_queue() > > nr_loopback_timer() > > nr_rx_frame() // sk : socket_1's sk > > nr_process_rx_frame() // NR_STATE_3 > > nr_state3_machine() // NR_DISCREQ > > nr_disconnect() > > nr_sk(sk)->state = NR_STATE_0; > > 5. close(socket_1) // sk refcount : 3 > > nr_release() // NR_STATE_0 > > sock_put(sk); // sk refcount : 0 > > sk_free(sk); > > close(uaf_socket) > > nr_release() > > sock_hold(sk); // UAF > > ``` > > > > KASAN report by syzbot: > > ``` > > BUG: KASAN: use-after-free in nr_release+0x66/0x460 net/netrom/af_netrom.c:520 > > Write of size 4 at addr ffff8880235d8080 by task syz-executor564/5128 > > > > Call Trace: > > <TASK> > > __dump_stack lib/dump_stack.c:88 [inline] > > dump_stack_lvl+0xd1/0x138 lib/dump_stack.c:106 > > print_address_description mm/kasan/report.c:306 [inline] > > print_report+0x15e/0x461 mm/kasan/report.c:417 > > kasan_report+0xbf/0x1f0 mm/kasan/report.c:517 > > check_region_inline mm/kasan/generic.c:183 [inline] > > kasan_check_range+0x141/0x190 mm/kasan/generic.c:189 > > instrument_atomic_read_write include/linux/instrumented.h:102 [inline] > > atomic_fetch_add_relaxed include/linux/atomic/atomic-instrumented.h:116 [inline] > > __refcount_add include/linux/refcount.h:193 [inline] > > __refcount_inc include/linux/refcount.h:250 [inline] > > refcount_inc include/linux/refcount.h:267 [inline] > > sock_hold include/net/sock.h:775 [inline] > > nr_release+0x66/0x460 net/netrom/af_netrom.c:520 > > __sock_release+0xcd/0x280 net/socket.c:650 > > sock_close+0x1c/0x20 net/socket.c:1365 > > __fput+0x27c/0xa90 fs/file_table.c:320 > > task_work_run+0x16f/0x270 kernel/task_work.c:179 > > exit_task_work include/linux/task_work.h:38 [inline] > > do_exit+0xaa8/0x2950 kernel/exit.c:867 > > do_group_exit+0xd4/0x2a0 kernel/exit.c:1012 > > get_signal+0x21c3/0x2450 kernel/signal.c:2859 > > arch_do_signal_or_restart+0x79/0x5c0 arch/x86/kernel/signal.c:306 > > exit_to_user_mode_loop kernel/entry/common.c:168 [inline] > > exit_to_user_mode_prepare+0x15f/0x250 kernel/entry/common.c:203 > > __syscall_exit_to_user_mode_work kernel/entry/common.c:285 [inline] > > syscall_exit_to_user_mode+0x1d/0x50 kernel/entry/common.c:296 > > do_syscall_64+0x46/0xb0 arch/x86/entry/common.c:86 > > entry_SYSCALL_64_after_hwframe+0x63/0xcd > > RIP: 0033:0x7f6c19e3c9b9 > > Code: Unable to access opcode bytes at 0x7f6c19e3c98f. > > RSP: 002b:00007fffd4ba2ce8 EFLAGS: 00000246 ORIG_RAX: 0000000000000133 > > RAX: 0000000000000116 RBX: 0000000000000003 RCX: 00007f6c19e3c9b9 > > RDX: 0000000000000318 RSI: 00000000200bd000 RDI: 0000000000000006 > > RBP: 0000000000000003 R08: 000000000000000d R09: 000000000000000d > > R10: 0000000000000000 R11: 0000000000000246 R12: 000055555566a2c0 > > R13: 0000000000000011 R14: 0000000000000000 R15: 0000000000000000 > > </TASK> > > > > Allocated by task 5128: > > kasan_save_stack+0x22/0x40 mm/kasan/common.c:45 > > kasan_set_track+0x25/0x30 mm/kasan/common.c:52 > > ____kasan_kmalloc mm/kasan/common.c:371 [inline] > > ____kasan_kmalloc mm/kasan/common.c:330 [inline] > > __kasan_kmalloc+0xa3/0xb0 mm/kasan/common.c:380 > > kasan_kmalloc include/linux/kasan.h:211 [inline] > > __do_kmalloc_node mm/slab_common.c:968 [inline] > > __kmalloc+0x5a/0xd0 mm/slab_common.c:981 > > kmalloc include/linux/slab.h:584 [inline] > > sk_prot_alloc+0x140/0x290 net/core/sock.c:2038 > > sk_alloc+0x3a/0x7a0 net/core/sock.c:2091 > > nr_create+0xb6/0x5f0 net/netrom/af_netrom.c:433 > > __sock_create+0x359/0x790 net/socket.c:1515 > > sock_create net/socket.c:1566 [inline] > > __sys_socket_create net/socket.c:1603 [inline] > > __sys_socket_create net/socket.c:1588 [inline] > > __sys_socket+0x133/0x250 net/socket.c:1636 > > __do_sys_socket net/socket.c:1649 [inline] > > __se_sys_socket net/socket.c:1647 [inline] > > __x64_sys_socket+0x73/0xb0 net/socket.c:1647 > > do_syscall_x64 arch/x86/entry/common.c:50 [inline] > > do_syscall_64+0x39/0xb0 arch/x86/entry/common.c:80 > > entry_SYSCALL_64_after_hwframe+0x63/0xcd > > > > Freed by task 5128: > > kasan_save_stack+0x22/0x40 mm/kasan/common.c:45 > > kasan_set_track+0x25/0x30 mm/kasan/common.c:52 > > kasan_save_free_info+0x2b/0x40 mm/kasan/generic.c:518 > > ____kasan_slab_free mm/kasan/common.c:236 [inline] > > ____kasan_slab_free+0x13b/0x1a0 mm/kasan/common.c:200 > > kasan_slab_free include/linux/kasan.h:177 [inline] > > __cache_free mm/slab.c:3394 [inline] > > __do_kmem_cache_free mm/slab.c:3580 [inline] > > __kmem_cache_free+0xcd/0x3b0 mm/slab.c:3587 > > sk_prot_free net/core/sock.c:2074 [inline] > > __sk_destruct+0x5df/0x750 net/core/sock.c:2166 > > sk_destruct net/core/sock.c:2181 [inline] > > __sk_free+0x175/0x460 net/core/sock.c:2192 > > sk_free+0x7c/0xa0 net/core/sock.c:2203 > > sock_put include/net/sock.h:1991 [inline] > > nr_release+0x39e/0x460 net/netrom/af_netrom.c:554 > > __sock_release+0xcd/0x280 net/socket.c:650 > > sock_close+0x1c/0x20 net/socket.c:1365 > > __fput+0x27c/0xa90 fs/file_table.c:320 > > task_work_run+0x16f/0x270 kernel/task_work.c:179 > > exit_task_work include/linux/task_work.h:38 [inline] > > do_exit+0xaa8/0x2950 kernel/exit.c:867 > > do_group_exit+0xd4/0x2a0 kernel/exit.c:1012 > > get_signal+0x21c3/0x2450 kernel/signal.c:2859 > > arch_do_signal_or_restart+0x79/0x5c0 arch/x86/kernel/signal.c:306 > > exit_to_user_mode_loop kernel/entry/common.c:168 [inline] > > exit_to_user_mode_prepare+0x15f/0x250 kernel/entry/common.c:203 > > __syscall_exit_to_user_mode_work kernel/entry/common.c:285 [inline] > > syscall_exit_to_user_mode+0x1d/0x50 kernel/entry/common.c:296 > > do_syscall_64+0x46/0xb0 arch/x86/entry/common.c:86 > > entry_SYSCALL_64_after_hwframe+0x63/0xcd > > ``` > > > > To fix this problem, nr_listen() returns -EINVAL for sockets that successfully nr_connect(). > > > > Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2") > > Reported-by: syzbot+caa188bdfc1eeafeb418@syzkaller.appspotmail.com > > Signed-off-by: Hyunwoo Kim <v4bel@theori.io> > > --- > > net/netrom/af_netrom.c | 5 +++++ > > 1 file changed, 5 insertions(+) > > > > diff --git a/net/netrom/af_netrom.c b/net/netrom/af_netrom.c > > index 6f7f4392cffb..5a4cb796150f 100644 > > --- a/net/netrom/af_netrom.c > > +++ b/net/netrom/af_netrom.c > > @@ -400,6 +400,11 @@ static int nr_listen(struct socket *sock, int backlog) > > struct sock *sk = sock->sk; > > > > lock_sock(sk); > > + if (sock->state != SS_UNCONNECTED) { > > + release_sock(sk); > > + return -EINVAL; > > + } > > + > > if (sk->sk_state != TCP_LISTEN) { > > memset(&nr_sk(sk)->user_addr, 0, AX25_ADDR_LEN); > > sk->sk_max_ack_backlog = backlog; > > -- > > 2.25.1
From: Hyunwoo Kim <v4bel@theori.io> Date: Wed, 25 Jan 2023 01:17:10 -0800 > On Tue, Jan 24, 2023 at 05:43:47PM -0800, Kuniyuki Iwashima wrote: > > Hi, > > > > I think the diff looks good but changelog is not correct. > > > > From: Hyunwoo Kim <v4bel@theori.io> > > Date: Mon, 23 Jan 2023 09:22:33 -0800 > > > If listen() and accept() are called on an AF_NETROM socket that > > > has already been connect()ed, accept() succeeds in connecting. > > > This is because nr_accept() dequeues the skb queued in > > > `sk->sk_receive_queue` in nr_connect(). > > > > This sentence is misleading. The skb sent by nr_connect() is queued > > up for the peer's sk_receive_queue (socket_2 below), and also the reply > > is not queued for connect()er (socket_1) because it is NR_CONNACK and > > just consumed by nr_state1_machine() to transition to NR_STATE_3 and > > TCP_ESTABLISHED. > > > > > > > > > > This causes nr_accept() to allocate and return a sock with the > > > sk of the parent AF_NETROM socket. > > > > This happens because the peer (socket_2) sends some data by nr_sendmsg(), > > whose type is NR_INFO and the data is queued up for the connect()er's > > (socket_1) sk_receive_queue by > > > > - nr_state3_machine > > - nr_queue_rx_frame > > - sock_queue_rcv_skb > > you're right. > I misanalyzed nr_rx_frame(). sorry. > > > > > > > > And here's where use-after-free > > > can happen through complex race conditions: > > > ``` > > > cpu0 cpu1 > > > 1. socket_2 = socket(AF_NETROM) > > > listen(socket_2) > > > accepted_socket = accept(socket_2) // loopback connection with socket_1 > > > 2. socket_1 = socket(AF_NETROM) > > > nr_create() // sk refcount : 1 > > > connect(socket_1) // loopback connection with socket_2 > > > nr_connect() > > > nr_establish_data_link() > > > nr_write_internal() > > > > This is NR_CONNREQ for socket_2. > > > > > > > nr_transmit_buffer() > > > nr_route_frame() > > > nr_loopback_queue() > > > nr_loopback_timer() > > > nr_rx_frame() > > > nr_process_rx_frame() > > > > So, nr_process_rx_frame() is not called from nr_rx_frame() because the > > kernel finds socket_2 by nr_find_listener(), and creates accepted_socket, > > and then responds with NR_CONNACK to socket_1. > > > > And accepted_socket sends some data with NR_INFO, then, nr_state3_machine() > > is called for socket_1 and the skb with sk pointing socket_1 is queued for > > socket_1's sk_receive_queue. > > > > I think this part is lacking in this UAF scenario. > > > > What do you think ? > > > > > > > nr_state3_machine() > > > nr_queue_rx_frame() > > > sock_queue_rcv_skb() > > > sock_queue_rcv_skb_reason() > > > __sock_queue_rcv_skb() > > > __skb_queue_tail(list, skb); // list : sk->sk_receive_queue > > > > > > 3. listen(socket_1) > > > nr_listen() > > > uaf_socket = accept(socket_1) > > > nr_accept() > > > skb_dequeue(&sk->sk_receive_queue); > > > 4. close(accepted_socket) > > > > And I guess this close() is not so important in this case. > > Even if we don't close the socket cleanly, the UAF will happen, right ? > > (it might be caused in a different place like timer though) > > IMHO, I think this `close(accepted_socket)` is a necessary procedure. > This is because `case NR_STATE_3:` in nr_release() is executed here, > and nr_write_internal(sk, NR_DISCREQ) ~ nr_disconnect() sets socket_1's sk to NR_STATE_0. > If sk of socket_1 is not set to NR_STATE_0, `case NR_STATE_3:` is executed instead of > `case NR_STATE_0:` in `close(socket_1)` and sk is not released. No, it's released by the heartbeat timer. The heartbeat timer finally calls nr_destroy_socket() as SOCK_DESTROY is flagged to the sk. > This means you will not be able to trigger UAF on close(uaf_socket). I didn't say close(uaf_socket) is not needed, I mean we can trigger UAF without close(__accepted_socket__). > > What do you think? > > So the modified flow chart looks like this: > ``` > cpu0 cpu1 > 1. socket_2 = socket(AF_NETROM) > . > . > listen(socket_2) > accepted_socket = accept(socket_2) > 2. socket_1 = socket(AF_NETROM) > nr_create() // sk refcount : 1 > connect(socket_1) > nr_connect() > nr_establish_data_link() > nr_write_internal() > nr_transmit_buffer() > nr_route_frame() > nr_loopback_queue() > nr_loopback_timer() > nr_rx_frame() > nr_process_rx_frame() > nr_state1_machine() > nr->state = NR_STATE_3; If you write the connect()'s stack trace, you should note the state change at the socket_1 side by NR_CONNACK. The NR_STATE_3 here is for socket_2 by NR_CONNREQ. We don't need the details of socket_2. Otherwise, you should just write connect(socket_1) only like the step 1 (socket(), listen() etc). > 3. write(accepted_socket) > nr_sendmsg() > nr_output() > nr_kick() > nr_send_iframe() > nr_transmit_buffer() > nr_route_frame() > nr_loopback_queue() > nr_loopback_timer() > nr_process_rx_frame(sk, skb); // sk : socket_1's sk > nr_state3_machine() > nr_queue_rx_frame() > sock_queue_rcv_skb() > sock_queue_rcv_skb_reason() > __sock_queue_rcv_skb() > __skb_queue_tail(list, skb); // list : socket_1's sk->sk_receive_queue > 4. listen(socket_1) > nr_listen() > uaf_socket = accept(socket_1) > nr_accept() > skb_dequeue(&sk->sk_receive_queue); > 5. close(accepted_socket) > nr_release() > nr_write_internal(sk, NR_DISCREQ) > nr_transmit_buffer() // NR_DISCREQ > nr_route_frame() > nr_loopback_queue() > nr_loopback_timer() > nr_rx_frame() // sk : socket_1's sk > nr_process_rx_frame() // NR_STATE_3 > nr_state3_machine() // NR_DISCREQ > nr_disconnect() > nr_sk(sk)->state = NR_STATE_0; > 6. close(socket_1) // sk refcount : 3 > nr_release() // NR_STATE_0 > sock_put(sk); // sk refcount : 0 > sk_free(sk); > close(uaf_socket) > nr_release() > sock_hold(sk); // UAF > ``` > > Anything wrong? > > > > > > > nr_release() > > > nr_write_internal(sk, NR_DISCREQ) > > > nr_transmit_buffer() // NR_DISCREQ > > > nr_route_frame() > > > nr_loopback_queue() > > > nr_loopback_timer() > > > nr_rx_frame() // sk : socket_1's sk > > > nr_process_rx_frame() // NR_STATE_3 > > > nr_state3_machine() // NR_DISCREQ > > > nr_disconnect() > > > nr_sk(sk)->state = NR_STATE_0; > > > 5. close(socket_1) // sk refcount : 3 > > > nr_release() // NR_STATE_0 > > > sock_put(sk); // sk refcount : 0 > > > sk_free(sk); > > > close(uaf_socket) > > > nr_release() > > > sock_hold(sk); // UAF > > > ``` > > > > > > KASAN report by syzbot: > > > ``` > > > BUG: KASAN: use-after-free in nr_release+0x66/0x460 net/netrom/af_netrom.c:520 > > > Write of size 4 at addr ffff8880235d8080 by task syz-executor564/5128 > > > > > > Call Trace: > > > <TASK> > > > __dump_stack lib/dump_stack.c:88 [inline] > > > dump_stack_lvl+0xd1/0x138 lib/dump_stack.c:106 > > > print_address_description mm/kasan/report.c:306 [inline] > > > print_report+0x15e/0x461 mm/kasan/report.c:417 > > > kasan_report+0xbf/0x1f0 mm/kasan/report.c:517 > > > check_region_inline mm/kasan/generic.c:183 [inline] > > > kasan_check_range+0x141/0x190 mm/kasan/generic.c:189 > > > instrument_atomic_read_write include/linux/instrumented.h:102 [inline] > > > atomic_fetch_add_relaxed include/linux/atomic/atomic-instrumented.h:116 [inline] > > > __refcount_add include/linux/refcount.h:193 [inline] > > > __refcount_inc include/linux/refcount.h:250 [inline] > > > refcount_inc include/linux/refcount.h:267 [inline] > > > sock_hold include/net/sock.h:775 [inline] > > > nr_release+0x66/0x460 net/netrom/af_netrom.c:520 > > > __sock_release+0xcd/0x280 net/socket.c:650 > > > sock_close+0x1c/0x20 net/socket.c:1365 > > > __fput+0x27c/0xa90 fs/file_table.c:320 > > > task_work_run+0x16f/0x270 kernel/task_work.c:179 > > > exit_task_work include/linux/task_work.h:38 [inline] > > > do_exit+0xaa8/0x2950 kernel/exit.c:867 > > > do_group_exit+0xd4/0x2a0 kernel/exit.c:1012 > > > get_signal+0x21c3/0x2450 kernel/signal.c:2859 > > > arch_do_signal_or_restart+0x79/0x5c0 arch/x86/kernel/signal.c:306 > > > exit_to_user_mode_loop kernel/entry/common.c:168 [inline] > > > exit_to_user_mode_prepare+0x15f/0x250 kernel/entry/common.c:203 > > > __syscall_exit_to_user_mode_work kernel/entry/common.c:285 [inline] > > > syscall_exit_to_user_mode+0x1d/0x50 kernel/entry/common.c:296 > > > do_syscall_64+0x46/0xb0 arch/x86/entry/common.c:86 > > > entry_SYSCALL_64_after_hwframe+0x63/0xcd > > > RIP: 0033:0x7f6c19e3c9b9 > > > Code: Unable to access opcode bytes at 0x7f6c19e3c98f. > > > RSP: 002b:00007fffd4ba2ce8 EFLAGS: 00000246 ORIG_RAX: 0000000000000133 > > > RAX: 0000000000000116 RBX: 0000000000000003 RCX: 00007f6c19e3c9b9 > > > RDX: 0000000000000318 RSI: 00000000200bd000 RDI: 0000000000000006 > > > RBP: 0000000000000003 R08: 000000000000000d R09: 000000000000000d > > > R10: 0000000000000000 R11: 0000000000000246 R12: 000055555566a2c0 > > > R13: 0000000000000011 R14: 0000000000000000 R15: 0000000000000000 > > > </TASK> > > > > > > Allocated by task 5128: > > > kasan_save_stack+0x22/0x40 mm/kasan/common.c:45 > > > kasan_set_track+0x25/0x30 mm/kasan/common.c:52 > > > ____kasan_kmalloc mm/kasan/common.c:371 [inline] > > > ____kasan_kmalloc mm/kasan/common.c:330 [inline] > > > __kasan_kmalloc+0xa3/0xb0 mm/kasan/common.c:380 > > > kasan_kmalloc include/linux/kasan.h:211 [inline] > > > __do_kmalloc_node mm/slab_common.c:968 [inline] > > > __kmalloc+0x5a/0xd0 mm/slab_common.c:981 > > > kmalloc include/linux/slab.h:584 [inline] > > > sk_prot_alloc+0x140/0x290 net/core/sock.c:2038 > > > sk_alloc+0x3a/0x7a0 net/core/sock.c:2091 > > > nr_create+0xb6/0x5f0 net/netrom/af_netrom.c:433 > > > __sock_create+0x359/0x790 net/socket.c:1515 > > > sock_create net/socket.c:1566 [inline] > > > __sys_socket_create net/socket.c:1603 [inline] > > > __sys_socket_create net/socket.c:1588 [inline] > > > __sys_socket+0x133/0x250 net/socket.c:1636 > > > __do_sys_socket net/socket.c:1649 [inline] > > > __se_sys_socket net/socket.c:1647 [inline] > > > __x64_sys_socket+0x73/0xb0 net/socket.c:1647 > > > do_syscall_x64 arch/x86/entry/common.c:50 [inline] > > > do_syscall_64+0x39/0xb0 arch/x86/entry/common.c:80 > > > entry_SYSCALL_64_after_hwframe+0x63/0xcd > > > > > > Freed by task 5128: > > > kasan_save_stack+0x22/0x40 mm/kasan/common.c:45 > > > kasan_set_track+0x25/0x30 mm/kasan/common.c:52 > > > kasan_save_free_info+0x2b/0x40 mm/kasan/generic.c:518 > > > ____kasan_slab_free mm/kasan/common.c:236 [inline] > > > ____kasan_slab_free+0x13b/0x1a0 mm/kasan/common.c:200 > > > kasan_slab_free include/linux/kasan.h:177 [inline] > > > __cache_free mm/slab.c:3394 [inline] > > > __do_kmem_cache_free mm/slab.c:3580 [inline] > > > __kmem_cache_free+0xcd/0x3b0 mm/slab.c:3587 > > > sk_prot_free net/core/sock.c:2074 [inline] > > > __sk_destruct+0x5df/0x750 net/core/sock.c:2166 > > > sk_destruct net/core/sock.c:2181 [inline] > > > __sk_free+0x175/0x460 net/core/sock.c:2192 > > > sk_free+0x7c/0xa0 net/core/sock.c:2203 > > > sock_put include/net/sock.h:1991 [inline] > > > nr_release+0x39e/0x460 net/netrom/af_netrom.c:554 > > > __sock_release+0xcd/0x280 net/socket.c:650 > > > sock_close+0x1c/0x20 net/socket.c:1365 > > > __fput+0x27c/0xa90 fs/file_table.c:320 > > > task_work_run+0x16f/0x270 kernel/task_work.c:179 > > > exit_task_work include/linux/task_work.h:38 [inline] > > > do_exit+0xaa8/0x2950 kernel/exit.c:867 > > > do_group_exit+0xd4/0x2a0 kernel/exit.c:1012 > > > get_signal+0x21c3/0x2450 kernel/signal.c:2859 > > > arch_do_signal_or_restart+0x79/0x5c0 arch/x86/kernel/signal.c:306 > > > exit_to_user_mode_loop kernel/entry/common.c:168 [inline] > > > exit_to_user_mode_prepare+0x15f/0x250 kernel/entry/common.c:203 > > > __syscall_exit_to_user_mode_work kernel/entry/common.c:285 [inline] > > > syscall_exit_to_user_mode+0x1d/0x50 kernel/entry/common.c:296 > > > do_syscall_64+0x46/0xb0 arch/x86/entry/common.c:86 > > > entry_SYSCALL_64_after_hwframe+0x63/0xcd > > > ``` > > > > > > To fix this problem, nr_listen() returns -EINVAL for sockets that successfully nr_connect(). > > > > > > Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2") > > > Reported-by: syzbot+caa188bdfc1eeafeb418@syzkaller.appspotmail.com > > > Signed-off-by: Hyunwoo Kim <v4bel@theori.io> > > > --- > > > net/netrom/af_netrom.c | 5 +++++ > > > 1 file changed, 5 insertions(+) > > > > > > diff --git a/net/netrom/af_netrom.c b/net/netrom/af_netrom.c > > > index 6f7f4392cffb..5a4cb796150f 100644 > > > --- a/net/netrom/af_netrom.c > > > +++ b/net/netrom/af_netrom.c > > > @@ -400,6 +400,11 @@ static int nr_listen(struct socket *sock, int backlog) > > > struct sock *sk = sock->sk; > > > > > > lock_sock(sk); > > > + if (sock->state != SS_UNCONNECTED) { > > > + release_sock(sk); > > > + return -EINVAL; > > > + } > > > + > > > if (sk->sk_state != TCP_LISTEN) { > > > memset(&nr_sk(sk)->user_addr, 0, AX25_ADDR_LEN); > > > sk->sk_max_ack_backlog = backlog; > > > -- > > > 2.25.1
On Wed, Jan 25, 2023 at 08:40:35AM -0800, Kuniyuki Iwashima wrote: > From: Hyunwoo Kim <v4bel@theori.io> > Date: Wed, 25 Jan 2023 01:17:10 -0800 > > On Tue, Jan 24, 2023 at 05:43:47PM -0800, Kuniyuki Iwashima wrote: > > > Hi, > > > > > > I think the diff looks good but changelog is not correct. > > > > > > From: Hyunwoo Kim <v4bel@theori.io> > > > Date: Mon, 23 Jan 2023 09:22:33 -0800 > > > > If listen() and accept() are called on an AF_NETROM socket that > > > > has already been connect()ed, accept() succeeds in connecting. > > > > This is because nr_accept() dequeues the skb queued in > > > > `sk->sk_receive_queue` in nr_connect(). > > > > > > This sentence is misleading. The skb sent by nr_connect() is queued > > > up for the peer's sk_receive_queue (socket_2 below), and also the reply > > > is not queued for connect()er (socket_1) because it is NR_CONNACK and > > > just consumed by nr_state1_machine() to transition to NR_STATE_3 and > > > TCP_ESTABLISHED. > > > > > > > > > > > > > > This causes nr_accept() to allocate and return a sock with the > > > > sk of the parent AF_NETROM socket. > > > > > > This happens because the peer (socket_2) sends some data by nr_sendmsg(), > > > whose type is NR_INFO and the data is queued up for the connect()er's > > > (socket_1) sk_receive_queue by > > > > > > - nr_state3_machine > > > - nr_queue_rx_frame > > > - sock_queue_rcv_skb > > > > you're right. > > I misanalyzed nr_rx_frame(). sorry. > > > > > > > > > > > > And here's where use-after-free > > > > can happen through complex race conditions: > > > > ``` > > > > cpu0 cpu1 > > > > 1. socket_2 = socket(AF_NETROM) > > > > listen(socket_2) > > > > accepted_socket = accept(socket_2) // loopback connection with socket_1 > > > > 2. socket_1 = socket(AF_NETROM) > > > > nr_create() // sk refcount : 1 > > > > connect(socket_1) // loopback connection with socket_2 > > > > nr_connect() > > > > nr_establish_data_link() > > > > nr_write_internal() > > > > > > This is NR_CONNREQ for socket_2. > > > > > > > > > > nr_transmit_buffer() > > > > nr_route_frame() > > > > nr_loopback_queue() > > > > nr_loopback_timer() > > > > nr_rx_frame() > > > > nr_process_rx_frame() > > > > > > So, nr_process_rx_frame() is not called from nr_rx_frame() because the > > > kernel finds socket_2 by nr_find_listener(), and creates accepted_socket, > > > and then responds with NR_CONNACK to socket_1. > > > > > > And accepted_socket sends some data with NR_INFO, then, nr_state3_machine() > > > is called for socket_1 and the skb with sk pointing socket_1 is queued for > > > socket_1's sk_receive_queue. > > > > > > I think this part is lacking in this UAF scenario. > > > > > > What do you think ? > > > > > > > > > > nr_state3_machine() > > > > nr_queue_rx_frame() > > > > sock_queue_rcv_skb() > > > > sock_queue_rcv_skb_reason() > > > > __sock_queue_rcv_skb() > > > > __skb_queue_tail(list, skb); // list : sk->sk_receive_queue > > > > > > > > 3. listen(socket_1) > > > > nr_listen() > > > > uaf_socket = accept(socket_1) > > > > nr_accept() > > > > skb_dequeue(&sk->sk_receive_queue); > > > > 4. close(accepted_socket) > > > > > > And I guess this close() is not so important in this case. > > > Even if we don't close the socket cleanly, the UAF will happen, right ? > > > (it might be caused in a different place like timer though) > > > > IMHO, I think this `close(accepted_socket)` is a necessary procedure. > > This is because `case NR_STATE_3:` in nr_release() is executed here, > > and nr_write_internal(sk, NR_DISCREQ) ~ nr_disconnect() sets socket_1's sk to NR_STATE_0. > > If sk of socket_1 is not set to NR_STATE_0, `case NR_STATE_3:` is executed instead of > > `case NR_STATE_0:` in `close(socket_1)` and sk is not released. > > No, it's released by the heartbeat timer. The heartbeat timer finally > calls nr_destroy_socket() as SOCK_DESTROY is flagged to the sk. > > > > This means you will not be able to trigger UAF on close(uaf_socket). > > I didn't say close(uaf_socket) is not needed, I mean we can > trigger UAF without close(__accepted_socket__). ok. understand. The reason why I put close(accepted_socket) is because Freed by task in the KASAN log I attached to the patch description is as follows: ``` Freed by task 5128: kasan_save_stack+0x22/0x40 mm/kasan/common.c:45 kasan_set_track+0x25/0x30 mm/kasan/common.c:52 kasan_save_free_info+0x2b/0x40 mm/kasan/generic.c:518 ____kasan_slab_free mm/kasan/common.c:236 [inline] ____kasan_slab_free+0x13b/0x1a0 mm/kasan/common.c:200 kasan_slab_free include/linux/kasan.h:177 [inline] __cache_free mm/slab.c:3394 [inline] __do_kmem_cache_free mm/slab.c:3580 [inline] __kmem_cache_free+0xcd/0x3b0 mm/slab.c:3587 sk_prot_free net/core/sock.c:2074 [inline] __sk_destruct+0x5df/0x750 net/core/sock.c:2166 sk_destruct net/core/sock.c:2181 [inline] __sk_free+0x175/0x460 net/core/sock.c:2192 sk_free+0x7c/0xa0 net/core/sock.c:2203 sock_put include/net/sock.h:1991 [inline] nr_release+0x39e/0x460 net/netrom/af_netrom.c:554 __sock_release+0xcd/0x280 net/socket.c:650 sock_close+0x1c/0x20 net/socket.c:1365 __fput+0x27c/0xa90 fs/file_table.c:320 task_work_run+0x16f/0x270 kernel/task_work.c:179 exit_task_work include/linux/task_work.h:38 [inline] do_exit+0xaa8/0x2950 kernel/exit.c:867 do_group_exit+0xd4/0x2a0 kernel/exit.c:1012 get_signal+0x21c3/0x2450 kernel/signal.c:2859 arch_do_signal_or_restart+0x79/0x5c0 arch/x86/kernel/signal.c:306 exit_to_user_mode_loop kernel/entry/common.c:168 [inline] exit_to_user_mode_prepare+0x15f/0x250 kernel/entry/common.c:203 __syscall_exit_to_user_mode_work kernel/entry/common.c:285 [inline] syscall_exit_to_user_mode+0x1d/0x50 kernel/entry/common.c:296 do_syscall_64+0x46/0xb0 arch/x86/entry/common.c:86 entry_SYSCALL_64_after_hwframe+0x63/0xcd ``` If, as you say, heartbeat is running and UAF is triggered, the KASAN Freed log will look like this: ``` Freed by task 14094: kasan_save_stack+0x22/0x40 mm/kasan/common.c:45 kasan_set_track+0x25/0x30 mm/kasan/common.c:52 kasan_save_free_info+0x2e/0x40 mm/kasan/generic.c:518 ____kasan_slab_free mm/kasan/common.c:236 [inline] ____kasan_slab_free+0x160/0x1c0 mm/kasan/common.c:200 kasan_slab_free include/linux/kasan.h:177 [inline] slab_free_hook mm/slub.c:1781 [inline] slab_free_freelist_hook+0x8b/0x1c0 mm/slub.c:1807 slab_free mm/slub.c:3787 [inline] __kmem_cache_free+0xaf/0x3b0 mm/slub.c:3800 sk_prot_free net/core/sock.c:2074 [inline] __sk_destruct+0x5df/0x750 net/core/sock.c:2166 sk_destruct net/core/sock.c:2181 [inline] __sk_free+0x175/0x460 net/core/sock.c:2192 sk_free+0x7c/0xa0 net/core/sock.c:2203 sock_put include/net/sock.h:1991 [inline] nr_heartbeat_expiry+0x1d7/0x460 net/netrom/nr_timer.c:148 call_timer_fn+0x1da/0x7c0 kernel/time/timer.c:1700 expire_timers+0x2c6/0x5c0 kernel/time/timer.c:1751 __run_timers kernel/time/timer.c:2022 [inline] __run_timers kernel/time/timer.c:1995 [inline] run_timer_softirq+0x326/0x910 kernel/time/timer.c:2035 __do_softirq+0x1fb/0xadc kernel/softirq.c:571 ``` Anyway, I'm trying to keep close(accepted_socket) in the flow chart because the KASAN log I added to the first patch description is the first one. > > > > > > What do you think? > > > > So the modified flow chart looks like this: > > ``` > > cpu0 cpu1 > > 1. socket_2 = socket(AF_NETROM) > > . > > . > > listen(socket_2) > > accepted_socket = accept(socket_2) > > 2. socket_1 = socket(AF_NETROM) > > nr_create() // sk refcount : 1 > > connect(socket_1) > > nr_connect() > > nr_establish_data_link() > > nr_write_internal() > > nr_transmit_buffer() > > nr_route_frame() > > nr_loopback_queue() > > nr_loopback_timer() > > nr_rx_frame() > > nr_process_rx_frame() > > nr_state1_machine() > > nr->state = NR_STATE_3; > > If you write the connect()'s stack trace, you should note the state > change at the socket_1 side by NR_CONNACK. The NR_STATE_3 here is > for socket_2 by NR_CONNREQ. We don't need the details of socket_2. > > Otherwise, you should just write connect(socket_1) only like the > step 1 (socket(), listen() etc). So, the modified flow chart is: ``` cpu0 cpu1 1. socket_2 = socket(AF_NETROM) . . listen(socket_2) accepted_socket = accept(socket_2) 2. socket_1 = socket(AF_NETROM) nr_create() // sk refcount : 1 connect(socket_1) 3. write(accepted_socket) nr_sendmsg() nr_output() nr_kick() nr_send_iframe() nr_transmit_buffer() nr_route_frame() nr_loopback_queue() nr_loopback_timer() nr_rx_frame() nr_process_rx_frame(sk, skb); // sk : socket_1's sk nr_state3_machine() nr_queue_rx_frame() sock_queue_rcv_skb() sock_queue_rcv_skb_reason() __sock_queue_rcv_skb() __skb_queue_tail(list, skb); // list : socket_1's sk->sk_receive_queue 4. listen(socket_1) nr_listen() uaf_socket = accept(socket_1) nr_accept() skb_dequeue(&sk->sk_receive_queue); 5. close(accepted_socket) nr_release() nr_write_internal(sk, NR_DISCREQ) nr_transmit_buffer() // NR_DISCREQ nr_route_frame() nr_loopback_queue() nr_loopback_timer() nr_rx_frame() // sk : socket_1's sk nr_process_rx_frame() // NR_STATE_3 nr_state3_machine() // NR_DISCREQ nr_disconnect() nr_sk(sk)->state = NR_STATE_0; 6. close(socket_1) // sk refcount : 3 nr_release() // NR_STATE_0 sock_put(sk); // sk refcount : 0 sk_free(sk); close(uaf_socket) nr_release() sock_hold(sk); // UAF ``` And sorry for sending the v3 netrom patch first. I received an email that the v2 x25 patch with the wrong patch description was applied to the tree, so I made a v3 x25 patch in a hurry and sent the netrom as well. > > > > 3. write(accepted_socket) > > nr_sendmsg() > > nr_output() > > nr_kick() > > nr_send_iframe() > > nr_transmit_buffer() > > nr_route_frame() > > nr_loopback_queue() > > nr_loopback_timer() > > nr_process_rx_frame(sk, skb); // sk : socket_1's sk > > nr_state3_machine() > > nr_queue_rx_frame() > > sock_queue_rcv_skb() > > sock_queue_rcv_skb_reason() > > __sock_queue_rcv_skb() > > __skb_queue_tail(list, skb); // list : socket_1's sk->sk_receive_queue > > 4. listen(socket_1) > > nr_listen() > > uaf_socket = accept(socket_1) > > nr_accept() > > skb_dequeue(&sk->sk_receive_queue); > > 5. close(accepted_socket) > > nr_release() > > nr_write_internal(sk, NR_DISCREQ) > > nr_transmit_buffer() // NR_DISCREQ > > nr_route_frame() > > nr_loopback_queue() > > nr_loopback_timer() > > nr_rx_frame() // sk : socket_1's sk > > nr_process_rx_frame() // NR_STATE_3 > > nr_state3_machine() // NR_DISCREQ > > nr_disconnect() > > nr_sk(sk)->state = NR_STATE_0; > > 6. close(socket_1) // sk refcount : 3 > > nr_release() // NR_STATE_0 > > sock_put(sk); // sk refcount : 0 > > sk_free(sk); > > close(uaf_socket) > > nr_release() > > sock_hold(sk); // UAF > > ``` > > > > Anything wrong? > > > > > > > > > > > nr_release() > > > > nr_write_internal(sk, NR_DISCREQ) > > > > nr_transmit_buffer() // NR_DISCREQ > > > > nr_route_frame() > > > > nr_loopback_queue() > > > > nr_loopback_timer() > > > > nr_rx_frame() // sk : socket_1's sk > > > > nr_process_rx_frame() // NR_STATE_3 > > > > nr_state3_machine() // NR_DISCREQ > > > > nr_disconnect() > > > > nr_sk(sk)->state = NR_STATE_0; > > > > 5. close(socket_1) // sk refcount : 3 > > > > nr_release() // NR_STATE_0 > > > > sock_put(sk); // sk refcount : 0 > > > > sk_free(sk); > > > > close(uaf_socket) > > > > nr_release() > > > > sock_hold(sk); // UAF > > > > ``` > > > > > > > > KASAN report by syzbot: > > > > ``` > > > > BUG: KASAN: use-after-free in nr_release+0x66/0x460 net/netrom/af_netrom.c:520 > > > > Write of size 4 at addr ffff8880235d8080 by task syz-executor564/5128 > > > > > > > > Call Trace: > > > > <TASK> > > > > __dump_stack lib/dump_stack.c:88 [inline] > > > > dump_stack_lvl+0xd1/0x138 lib/dump_stack.c:106 > > > > print_address_description mm/kasan/report.c:306 [inline] > > > > print_report+0x15e/0x461 mm/kasan/report.c:417 > > > > kasan_report+0xbf/0x1f0 mm/kasan/report.c:517 > > > > check_region_inline mm/kasan/generic.c:183 [inline] > > > > kasan_check_range+0x141/0x190 mm/kasan/generic.c:189 > > > > instrument_atomic_read_write include/linux/instrumented.h:102 [inline] > > > > atomic_fetch_add_relaxed include/linux/atomic/atomic-instrumented.h:116 [inline] > > > > __refcount_add include/linux/refcount.h:193 [inline] > > > > __refcount_inc include/linux/refcount.h:250 [inline] > > > > refcount_inc include/linux/refcount.h:267 [inline] > > > > sock_hold include/net/sock.h:775 [inline] > > > > nr_release+0x66/0x460 net/netrom/af_netrom.c:520 > > > > __sock_release+0xcd/0x280 net/socket.c:650 > > > > sock_close+0x1c/0x20 net/socket.c:1365 > > > > __fput+0x27c/0xa90 fs/file_table.c:320 > > > > task_work_run+0x16f/0x270 kernel/task_work.c:179 > > > > exit_task_work include/linux/task_work.h:38 [inline] > > > > do_exit+0xaa8/0x2950 kernel/exit.c:867 > > > > do_group_exit+0xd4/0x2a0 kernel/exit.c:1012 > > > > get_signal+0x21c3/0x2450 kernel/signal.c:2859 > > > > arch_do_signal_or_restart+0x79/0x5c0 arch/x86/kernel/signal.c:306 > > > > exit_to_user_mode_loop kernel/entry/common.c:168 [inline] > > > > exit_to_user_mode_prepare+0x15f/0x250 kernel/entry/common.c:203 > > > > __syscall_exit_to_user_mode_work kernel/entry/common.c:285 [inline] > > > > syscall_exit_to_user_mode+0x1d/0x50 kernel/entry/common.c:296 > > > > do_syscall_64+0x46/0xb0 arch/x86/entry/common.c:86 > > > > entry_SYSCALL_64_after_hwframe+0x63/0xcd > > > > RIP: 0033:0x7f6c19e3c9b9 > > > > Code: Unable to access opcode bytes at 0x7f6c19e3c98f. > > > > RSP: 002b:00007fffd4ba2ce8 EFLAGS: 00000246 ORIG_RAX: 0000000000000133 > > > > RAX: 0000000000000116 RBX: 0000000000000003 RCX: 00007f6c19e3c9b9 > > > > RDX: 0000000000000318 RSI: 00000000200bd000 RDI: 0000000000000006 > > > > RBP: 0000000000000003 R08: 000000000000000d R09: 000000000000000d > > > > R10: 0000000000000000 R11: 0000000000000246 R12: 000055555566a2c0 > > > > R13: 0000000000000011 R14: 0000000000000000 R15: 0000000000000000 > > > > </TASK> > > > > > > > > Allocated by task 5128: > > > > kasan_save_stack+0x22/0x40 mm/kasan/common.c:45 > > > > kasan_set_track+0x25/0x30 mm/kasan/common.c:52 > > > > ____kasan_kmalloc mm/kasan/common.c:371 [inline] > > > > ____kasan_kmalloc mm/kasan/common.c:330 [inline] > > > > __kasan_kmalloc+0xa3/0xb0 mm/kasan/common.c:380 > > > > kasan_kmalloc include/linux/kasan.h:211 [inline] > > > > __do_kmalloc_node mm/slab_common.c:968 [inline] > > > > __kmalloc+0x5a/0xd0 mm/slab_common.c:981 > > > > kmalloc include/linux/slab.h:584 [inline] > > > > sk_prot_alloc+0x140/0x290 net/core/sock.c:2038 > > > > sk_alloc+0x3a/0x7a0 net/core/sock.c:2091 > > > > nr_create+0xb6/0x5f0 net/netrom/af_netrom.c:433 > > > > __sock_create+0x359/0x790 net/socket.c:1515 > > > > sock_create net/socket.c:1566 [inline] > > > > __sys_socket_create net/socket.c:1603 [inline] > > > > __sys_socket_create net/socket.c:1588 [inline] > > > > __sys_socket+0x133/0x250 net/socket.c:1636 > > > > __do_sys_socket net/socket.c:1649 [inline] > > > > __se_sys_socket net/socket.c:1647 [inline] > > > > __x64_sys_socket+0x73/0xb0 net/socket.c:1647 > > > > do_syscall_x64 arch/x86/entry/common.c:50 [inline] > > > > do_syscall_64+0x39/0xb0 arch/x86/entry/common.c:80 > > > > entry_SYSCALL_64_after_hwframe+0x63/0xcd > > > > > > > > Freed by task 5128: > > > > kasan_save_stack+0x22/0x40 mm/kasan/common.c:45 > > > > kasan_set_track+0x25/0x30 mm/kasan/common.c:52 > > > > kasan_save_free_info+0x2b/0x40 mm/kasan/generic.c:518 > > > > ____kasan_slab_free mm/kasan/common.c:236 [inline] > > > > ____kasan_slab_free+0x13b/0x1a0 mm/kasan/common.c:200 > > > > kasan_slab_free include/linux/kasan.h:177 [inline] > > > > __cache_free mm/slab.c:3394 [inline] > > > > __do_kmem_cache_free mm/slab.c:3580 [inline] > > > > __kmem_cache_free+0xcd/0x3b0 mm/slab.c:3587 > > > > sk_prot_free net/core/sock.c:2074 [inline] > > > > __sk_destruct+0x5df/0x750 net/core/sock.c:2166 > > > > sk_destruct net/core/sock.c:2181 [inline] > > > > __sk_free+0x175/0x460 net/core/sock.c:2192 > > > > sk_free+0x7c/0xa0 net/core/sock.c:2203 > > > > sock_put include/net/sock.h:1991 [inline] > > > > nr_release+0x39e/0x460 net/netrom/af_netrom.c:554 > > > > __sock_release+0xcd/0x280 net/socket.c:650 > > > > sock_close+0x1c/0x20 net/socket.c:1365 > > > > __fput+0x27c/0xa90 fs/file_table.c:320 > > > > task_work_run+0x16f/0x270 kernel/task_work.c:179 > > > > exit_task_work include/linux/task_work.h:38 [inline] > > > > do_exit+0xaa8/0x2950 kernel/exit.c:867 > > > > do_group_exit+0xd4/0x2a0 kernel/exit.c:1012 > > > > get_signal+0x21c3/0x2450 kernel/signal.c:2859 > > > > arch_do_signal_or_restart+0x79/0x5c0 arch/x86/kernel/signal.c:306 > > > > exit_to_user_mode_loop kernel/entry/common.c:168 [inline] > > > > exit_to_user_mode_prepare+0x15f/0x250 kernel/entry/common.c:203 > > > > __syscall_exit_to_user_mode_work kernel/entry/common.c:285 [inline] > > > > syscall_exit_to_user_mode+0x1d/0x50 kernel/entry/common.c:296 > > > > do_syscall_64+0x46/0xb0 arch/x86/entry/common.c:86 > > > > entry_SYSCALL_64_after_hwframe+0x63/0xcd > > > > ``` > > > > > > > > To fix this problem, nr_listen() returns -EINVAL for sockets that successfully nr_connect(). > > > > > > > > Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2") > > > > Reported-by: syzbot+caa188bdfc1eeafeb418@syzkaller.appspotmail.com > > > > Signed-off-by: Hyunwoo Kim <v4bel@theori.io> > > > > --- > > > > net/netrom/af_netrom.c | 5 +++++ > > > > 1 file changed, 5 insertions(+) > > > > > > > > diff --git a/net/netrom/af_netrom.c b/net/netrom/af_netrom.c > > > > index 6f7f4392cffb..5a4cb796150f 100644 > > > > --- a/net/netrom/af_netrom.c > > > > +++ b/net/netrom/af_netrom.c > > > > @@ -400,6 +400,11 @@ static int nr_listen(struct socket *sock, int backlog) > > > > struct sock *sk = sock->sk; > > > > > > > > lock_sock(sk); > > > > + if (sock->state != SS_UNCONNECTED) { > > > > + release_sock(sk); > > > > + return -EINVAL; > > > > + } > > > > + > > > > if (sk->sk_state != TCP_LISTEN) { > > > > memset(&nr_sk(sk)->user_addr, 0, AX25_ADDR_LEN); > > > > sk->sk_max_ack_backlog = backlog; > > > > -- > > > > 2.25.1
From: Hyunwoo Kim <v4bel@theori.io> Date: Wed, 25 Jan 2023 20:16:36 -0800 > On Wed, Jan 25, 2023 at 08:40:35AM -0800, Kuniyuki Iwashima wrote: > > From: Hyunwoo Kim <v4bel@theori.io> > > Date: Wed, 25 Jan 2023 01:17:10 -0800 > > > On Tue, Jan 24, 2023 at 05:43:47PM -0800, Kuniyuki Iwashima wrote: > > > > Hi, > > > > > > > > I think the diff looks good but changelog is not correct. > > > > > > > > From: Hyunwoo Kim <v4bel@theori.io> > > > > Date: Mon, 23 Jan 2023 09:22:33 -0800 > > > > > If listen() and accept() are called on an AF_NETROM socket that > > > > > has already been connect()ed, accept() succeeds in connecting. > > > > > This is because nr_accept() dequeues the skb queued in > > > > > `sk->sk_receive_queue` in nr_connect(). > > > > > > > > This sentence is misleading. The skb sent by nr_connect() is queued > > > > up for the peer's sk_receive_queue (socket_2 below), and also the reply > > > > is not queued for connect()er (socket_1) because it is NR_CONNACK and > > > > just consumed by nr_state1_machine() to transition to NR_STATE_3 and > > > > TCP_ESTABLISHED. > > > > > > > > > > > > > > > > > > This causes nr_accept() to allocate and return a sock with the > > > > > sk of the parent AF_NETROM socket. > > > > > > > > This happens because the peer (socket_2) sends some data by nr_sendmsg(), > > > > whose type is NR_INFO and the data is queued up for the connect()er's > > > > (socket_1) sk_receive_queue by > > > > > > > > - nr_state3_machine > > > > - nr_queue_rx_frame > > > > - sock_queue_rcv_skb > > > > > > you're right. > > > I misanalyzed nr_rx_frame(). sorry. > > > > > > > > > > > > > > > > And here's where use-after-free > > > > > can happen through complex race conditions: > > > > > ``` > > > > > cpu0 cpu1 > > > > > 1. socket_2 = socket(AF_NETROM) > > > > > listen(socket_2) > > > > > accepted_socket = accept(socket_2) // loopback connection with socket_1 > > > > > 2. socket_1 = socket(AF_NETROM) > > > > > nr_create() // sk refcount : 1 > > > > > connect(socket_1) // loopback connection with socket_2 > > > > > nr_connect() > > > > > nr_establish_data_link() > > > > > nr_write_internal() > > > > > > > > This is NR_CONNREQ for socket_2. > > > > > > > > > > > > > nr_transmit_buffer() > > > > > nr_route_frame() > > > > > nr_loopback_queue() > > > > > nr_loopback_timer() > > > > > nr_rx_frame() > > > > > nr_process_rx_frame() > > > > > > > > So, nr_process_rx_frame() is not called from nr_rx_frame() because the > > > > kernel finds socket_2 by nr_find_listener(), and creates accepted_socket, > > > > and then responds with NR_CONNACK to socket_1. > > > > > > > > And accepted_socket sends some data with NR_INFO, then, nr_state3_machine() > > > > is called for socket_1 and the skb with sk pointing socket_1 is queued for > > > > socket_1's sk_receive_queue. > > > > > > > > I think this part is lacking in this UAF scenario. > > > > > > > > What do you think ? > > > > > > > > > > > > > nr_state3_machine() > > > > > nr_queue_rx_frame() > > > > > sock_queue_rcv_skb() > > > > > sock_queue_rcv_skb_reason() > > > > > __sock_queue_rcv_skb() > > > > > __skb_queue_tail(list, skb); // list : sk->sk_receive_queue > > > > > > > > > > 3. listen(socket_1) > > > > > nr_listen() > > > > > uaf_socket = accept(socket_1) > > > > > nr_accept() > > > > > skb_dequeue(&sk->sk_receive_queue); > > > > > 4. close(accepted_socket) > > > > > > > > And I guess this close() is not so important in this case. > > > > Even if we don't close the socket cleanly, the UAF will happen, right ? > > > > (it might be caused in a different place like timer though) > > > > > > IMHO, I think this `close(accepted_socket)` is a necessary procedure. > > > This is because `case NR_STATE_3:` in nr_release() is executed here, > > > and nr_write_internal(sk, NR_DISCREQ) ~ nr_disconnect() sets socket_1's sk to NR_STATE_0. > > > If sk of socket_1 is not set to NR_STATE_0, `case NR_STATE_3:` is executed instead of > > > `case NR_STATE_0:` in `close(socket_1)` and sk is not released. > > > > No, it's released by the heartbeat timer. The heartbeat timer finally > > calls nr_destroy_socket() as SOCK_DESTROY is flagged to the sk. > > > > > > > This means you will not be able to trigger UAF on close(uaf_socket). > > > > I didn't say close(uaf_socket) is not needed, I mean we can > > trigger UAF without close(__accepted_socket__). > > ok. understand. The reason why I put close(accepted_socket) > is because Freed by task in the KASAN log I attached to the > patch description is as follows: Ah, right. Sorry, I was confused with 3 close()s :/ > ``` > Freed by task 5128: > kasan_save_stack+0x22/0x40 mm/kasan/common.c:45 > kasan_set_track+0x25/0x30 mm/kasan/common.c:52 > kasan_save_free_info+0x2b/0x40 mm/kasan/generic.c:518 > ____kasan_slab_free mm/kasan/common.c:236 [inline] > ____kasan_slab_free+0x13b/0x1a0 mm/kasan/common.c:200 > kasan_slab_free include/linux/kasan.h:177 [inline] > __cache_free mm/slab.c:3394 [inline] > __do_kmem_cache_free mm/slab.c:3580 [inline] > __kmem_cache_free+0xcd/0x3b0 mm/slab.c:3587 > sk_prot_free net/core/sock.c:2074 [inline] > __sk_destruct+0x5df/0x750 net/core/sock.c:2166 > sk_destruct net/core/sock.c:2181 [inline] > __sk_free+0x175/0x460 net/core/sock.c:2192 > sk_free+0x7c/0xa0 net/core/sock.c:2203 > sock_put include/net/sock.h:1991 [inline] > nr_release+0x39e/0x460 net/netrom/af_netrom.c:554 > __sock_release+0xcd/0x280 net/socket.c:650 > sock_close+0x1c/0x20 net/socket.c:1365 > __fput+0x27c/0xa90 fs/file_table.c:320 > task_work_run+0x16f/0x270 kernel/task_work.c:179 > exit_task_work include/linux/task_work.h:38 [inline] > do_exit+0xaa8/0x2950 kernel/exit.c:867 > do_group_exit+0xd4/0x2a0 kernel/exit.c:1012 > get_signal+0x21c3/0x2450 kernel/signal.c:2859 > arch_do_signal_or_restart+0x79/0x5c0 arch/x86/kernel/signal.c:306 > exit_to_user_mode_loop kernel/entry/common.c:168 [inline] > exit_to_user_mode_prepare+0x15f/0x250 kernel/entry/common.c:203 > __syscall_exit_to_user_mode_work kernel/entry/common.c:285 [inline] > syscall_exit_to_user_mode+0x1d/0x50 kernel/entry/common.c:296 > do_syscall_64+0x46/0xb0 arch/x86/entry/common.c:86 > entry_SYSCALL_64_after_hwframe+0x63/0xcd > ``` > > If, as you say, heartbeat is running and UAF is triggered, > the KASAN Freed log will look like this: > ``` > Freed by task 14094: > kasan_save_stack+0x22/0x40 mm/kasan/common.c:45 > kasan_set_track+0x25/0x30 mm/kasan/common.c:52 > kasan_save_free_info+0x2e/0x40 mm/kasan/generic.c:518 > ____kasan_slab_free mm/kasan/common.c:236 [inline] > ____kasan_slab_free+0x160/0x1c0 mm/kasan/common.c:200 > kasan_slab_free include/linux/kasan.h:177 [inline] > slab_free_hook mm/slub.c:1781 [inline] > slab_free_freelist_hook+0x8b/0x1c0 mm/slub.c:1807 > slab_free mm/slub.c:3787 [inline] > __kmem_cache_free+0xaf/0x3b0 mm/slub.c:3800 > sk_prot_free net/core/sock.c:2074 [inline] > __sk_destruct+0x5df/0x750 net/core/sock.c:2166 > sk_destruct net/core/sock.c:2181 [inline] > __sk_free+0x175/0x460 net/core/sock.c:2192 > sk_free+0x7c/0xa0 net/core/sock.c:2203 > sock_put include/net/sock.h:1991 [inline] > nr_heartbeat_expiry+0x1d7/0x460 net/netrom/nr_timer.c:148 > call_timer_fn+0x1da/0x7c0 kernel/time/timer.c:1700 > expire_timers+0x2c6/0x5c0 kernel/time/timer.c:1751 > __run_timers kernel/time/timer.c:2022 [inline] > __run_timers kernel/time/timer.c:1995 [inline] > run_timer_softirq+0x326/0x910 kernel/time/timer.c:2035 > __do_softirq+0x1fb/0xadc kernel/softirq.c:571 > ``` > > Anyway, I'm trying to keep close(accepted_socket) in the flow chart > because the KASAN log I added to the first patch description is the first one. > > > > > > > > > > > What do you think? > > > > > > So the modified flow chart looks like this: > > > ``` > > > cpu0 cpu1 > > > 1. socket_2 = socket(AF_NETROM) > > > . > > > . > > > listen(socket_2) > > > accepted_socket = accept(socket_2) > > > 2. socket_1 = socket(AF_NETROM) > > > nr_create() // sk refcount : 1 > > > connect(socket_1) > > > nr_connect() > > > nr_establish_data_link() > > > nr_write_internal() > > > nr_transmit_buffer() > > > nr_route_frame() > > > nr_loopback_queue() > > > nr_loopback_timer() > > > nr_rx_frame() > > > nr_process_rx_frame() > > > nr_state1_machine() > > > nr->state = NR_STATE_3; > > > > If you write the connect()'s stack trace, you should note the state > > change at the socket_1 side by NR_CONNACK. The NR_STATE_3 here is > > for socket_2 by NR_CONNREQ. We don't need the details of socket_2. > > > > Otherwise, you should just write connect(socket_1) only like the > > step 1 (socket(), listen() etc). > > So, the modified flow chart is: > ``` > cpu0 cpu1 > 1. socket_2 = socket(AF_NETROM) > . > . > listen(socket_2) > accepted_socket = accept(socket_2) > 2. socket_1 = socket(AF_NETROM) > nr_create() // sk refcount : 1 > connect(socket_1) > 3. write(accepted_socket) > nr_sendmsg() > nr_output() > nr_kick() > nr_send_iframe() > nr_transmit_buffer() > nr_route_frame() > nr_loopback_queue() > nr_loopback_timer() > nr_rx_frame() > nr_process_rx_frame(sk, skb); // sk : socket_1's sk > nr_state3_machine() > nr_queue_rx_frame() > sock_queue_rcv_skb() > sock_queue_rcv_skb_reason() > __sock_queue_rcv_skb() > __skb_queue_tail(list, skb); // list : socket_1's sk->sk_receive_queue > 4. listen(socket_1) > nr_listen() > uaf_socket = accept(socket_1) > nr_accept() > skb_dequeue(&sk->sk_receive_queue); > 5. close(accepted_socket) > nr_release() > nr_write_internal(sk, NR_DISCREQ) > nr_transmit_buffer() // NR_DISCREQ > nr_route_frame() > nr_loopback_queue() > nr_loopback_timer() > nr_rx_frame() // sk : socket_1's sk > nr_process_rx_frame() // NR_STATE_3 > nr_state3_machine() // NR_DISCREQ > nr_disconnect() > nr_sk(sk)->state = NR_STATE_0; > 6. close(socket_1) // sk refcount : 3 > nr_release() // NR_STATE_0 > sock_put(sk); // sk refcount : 0 > sk_free(sk); > close(uaf_socket) > nr_release() > sock_hold(sk); // UAF > ``` Ok, this looks good. Please send v4 without my Reviewed-by tag. Thank you. > > And sorry for sending the v3 netrom patch first. > I received an email that the v2 x25 patch with the wrong patch > description was applied to the tree, > so I made a v3 x25 patch in a hurry and sent the netrom as well. > > > > > > > > 3. write(accepted_socket) > > > nr_sendmsg() > > > nr_output() > > > nr_kick() > > > nr_send_iframe() > > > nr_transmit_buffer() > > > nr_route_frame() > > > nr_loopback_queue() > > > nr_loopback_timer() > > > nr_process_rx_frame(sk, skb); // sk : socket_1's sk > > > nr_state3_machine() > > > nr_queue_rx_frame() > > > sock_queue_rcv_skb() > > > sock_queue_rcv_skb_reason() > > > __sock_queue_rcv_skb() > > > __skb_queue_tail(list, skb); // list : socket_1's sk->sk_receive_queue > > > 4. listen(socket_1) > > > nr_listen() > > > uaf_socket = accept(socket_1) > > > nr_accept() > > > skb_dequeue(&sk->sk_receive_queue); > > > 5. close(accepted_socket) > > > nr_release() > > > nr_write_internal(sk, NR_DISCREQ) > > > nr_transmit_buffer() // NR_DISCREQ > > > nr_route_frame() > > > nr_loopback_queue() > > > nr_loopback_timer() > > > nr_rx_frame() // sk : socket_1's sk > > > nr_process_rx_frame() // NR_STATE_3 > > > nr_state3_machine() // NR_DISCREQ > > > nr_disconnect() > > > nr_sk(sk)->state = NR_STATE_0; > > > 6. close(socket_1) // sk refcount : 3 > > > nr_release() // NR_STATE_0 > > > sock_put(sk); // sk refcount : 0 > > > sk_free(sk); > > > close(uaf_socket) > > > nr_release() > > > sock_hold(sk); // UAF > > > ``` > > > > > > Anything wrong? > > > > > > > > > > > > > > > nr_release() > > > > > nr_write_internal(sk, NR_DISCREQ) > > > > > nr_transmit_buffer() // NR_DISCREQ > > > > > nr_route_frame() > > > > > nr_loopback_queue() > > > > > nr_loopback_timer() > > > > > nr_rx_frame() // sk : socket_1's sk > > > > > nr_process_rx_frame() // NR_STATE_3 > > > > > nr_state3_machine() // NR_DISCREQ > > > > > nr_disconnect() > > > > > nr_sk(sk)->state = NR_STATE_0; > > > > > 5. close(socket_1) // sk refcount : 3 > > > > > nr_release() // NR_STATE_0 > > > > > sock_put(sk); // sk refcount : 0 > > > > > sk_free(sk); > > > > > close(uaf_socket) > > > > > nr_release() > > > > > sock_hold(sk); // UAF > > > > > ``` > > > > > > > > > > KASAN report by syzbot: > > > > > ``` > > > > > BUG: KASAN: use-after-free in nr_release+0x66/0x460 net/netrom/af_netrom.c:520 > > > > > Write of size 4 at addr ffff8880235d8080 by task syz-executor564/5128 > > > > > > > > > > Call Trace: > > > > > <TASK> > > > > > __dump_stack lib/dump_stack.c:88 [inline] > > > > > dump_stack_lvl+0xd1/0x138 lib/dump_stack.c:106 > > > > > print_address_description mm/kasan/report.c:306 [inline] > > > > > print_report+0x15e/0x461 mm/kasan/report.c:417 > > > > > kasan_report+0xbf/0x1f0 mm/kasan/report.c:517 > > > > > check_region_inline mm/kasan/generic.c:183 [inline] > > > > > kasan_check_range+0x141/0x190 mm/kasan/generic.c:189 > > > > > instrument_atomic_read_write include/linux/instrumented.h:102 [inline] > > > > > atomic_fetch_add_relaxed include/linux/atomic/atomic-instrumented.h:116 [inline] > > > > > __refcount_add include/linux/refcount.h:193 [inline] > > > > > __refcount_inc include/linux/refcount.h:250 [inline] > > > > > refcount_inc include/linux/refcount.h:267 [inline] > > > > > sock_hold include/net/sock.h:775 [inline] > > > > > nr_release+0x66/0x460 net/netrom/af_netrom.c:520 > > > > > __sock_release+0xcd/0x280 net/socket.c:650 > > > > > sock_close+0x1c/0x20 net/socket.c:1365 > > > > > __fput+0x27c/0xa90 fs/file_table.c:320 > > > > > task_work_run+0x16f/0x270 kernel/task_work.c:179 > > > > > exit_task_work include/linux/task_work.h:38 [inline] > > > > > do_exit+0xaa8/0x2950 kernel/exit.c:867 > > > > > do_group_exit+0xd4/0x2a0 kernel/exit.c:1012 > > > > > get_signal+0x21c3/0x2450 kernel/signal.c:2859 > > > > > arch_do_signal_or_restart+0x79/0x5c0 arch/x86/kernel/signal.c:306 > > > > > exit_to_user_mode_loop kernel/entry/common.c:168 [inline] > > > > > exit_to_user_mode_prepare+0x15f/0x250 kernel/entry/common.c:203 > > > > > __syscall_exit_to_user_mode_work kernel/entry/common.c:285 [inline] > > > > > syscall_exit_to_user_mode+0x1d/0x50 kernel/entry/common.c:296 > > > > > do_syscall_64+0x46/0xb0 arch/x86/entry/common.c:86 > > > > > entry_SYSCALL_64_after_hwframe+0x63/0xcd > > > > > RIP: 0033:0x7f6c19e3c9b9 > > > > > Code: Unable to access opcode bytes at 0x7f6c19e3c98f. > > > > > RSP: 002b:00007fffd4ba2ce8 EFLAGS: 00000246 ORIG_RAX: 0000000000000133 > > > > > RAX: 0000000000000116 RBX: 0000000000000003 RCX: 00007f6c19e3c9b9 > > > > > RDX: 0000000000000318 RSI: 00000000200bd000 RDI: 0000000000000006 > > > > > RBP: 0000000000000003 R08: 000000000000000d R09: 000000000000000d > > > > > R10: 0000000000000000 R11: 0000000000000246 R12: 000055555566a2c0 > > > > > R13: 0000000000000011 R14: 0000000000000000 R15: 0000000000000000 > > > > > </TASK> > > > > > > > > > > Allocated by task 5128: > > > > > kasan_save_stack+0x22/0x40 mm/kasan/common.c:45 > > > > > kasan_set_track+0x25/0x30 mm/kasan/common.c:52 > > > > > ____kasan_kmalloc mm/kasan/common.c:371 [inline] > > > > > ____kasan_kmalloc mm/kasan/common.c:330 [inline] > > > > > __kasan_kmalloc+0xa3/0xb0 mm/kasan/common.c:380 > > > > > kasan_kmalloc include/linux/kasan.h:211 [inline] > > > > > __do_kmalloc_node mm/slab_common.c:968 [inline] > > > > > __kmalloc+0x5a/0xd0 mm/slab_common.c:981 > > > > > kmalloc include/linux/slab.h:584 [inline] > > > > > sk_prot_alloc+0x140/0x290 net/core/sock.c:2038 > > > > > sk_alloc+0x3a/0x7a0 net/core/sock.c:2091 > > > > > nr_create+0xb6/0x5f0 net/netrom/af_netrom.c:433 > > > > > __sock_create+0x359/0x790 net/socket.c:1515 > > > > > sock_create net/socket.c:1566 [inline] > > > > > __sys_socket_create net/socket.c:1603 [inline] > > > > > __sys_socket_create net/socket.c:1588 [inline] > > > > > __sys_socket+0x133/0x250 net/socket.c:1636 > > > > > __do_sys_socket net/socket.c:1649 [inline] > > > > > __se_sys_socket net/socket.c:1647 [inline] > > > > > __x64_sys_socket+0x73/0xb0 net/socket.c:1647 > > > > > do_syscall_x64 arch/x86/entry/common.c:50 [inline] > > > > > do_syscall_64+0x39/0xb0 arch/x86/entry/common.c:80 > > > > > entry_SYSCALL_64_after_hwframe+0x63/0xcd > > > > > > > > > > Freed by task 5128: > > > > > kasan_save_stack+0x22/0x40 mm/kasan/common.c:45 > > > > > kasan_set_track+0x25/0x30 mm/kasan/common.c:52 > > > > > kasan_save_free_info+0x2b/0x40 mm/kasan/generic.c:518 > > > > > ____kasan_slab_free mm/kasan/common.c:236 [inline] > > > > > ____kasan_slab_free+0x13b/0x1a0 mm/kasan/common.c:200 > > > > > kasan_slab_free include/linux/kasan.h:177 [inline] > > > > > __cache_free mm/slab.c:3394 [inline] > > > > > __do_kmem_cache_free mm/slab.c:3580 [inline] > > > > > __kmem_cache_free+0xcd/0x3b0 mm/slab.c:3587 > > > > > sk_prot_free net/core/sock.c:2074 [inline] > > > > > __sk_destruct+0x5df/0x750 net/core/sock.c:2166 > > > > > sk_destruct net/core/sock.c:2181 [inline] > > > > > __sk_free+0x175/0x460 net/core/sock.c:2192 > > > > > sk_free+0x7c/0xa0 net/core/sock.c:2203 > > > > > sock_put include/net/sock.h:1991 [inline] > > > > > nr_release+0x39e/0x460 net/netrom/af_netrom.c:554 > > > > > __sock_release+0xcd/0x280 net/socket.c:650 > > > > > sock_close+0x1c/0x20 net/socket.c:1365 > > > > > __fput+0x27c/0xa90 fs/file_table.c:320 > > > > > task_work_run+0x16f/0x270 kernel/task_work.c:179 > > > > > exit_task_work include/linux/task_work.h:38 [inline] > > > > > do_exit+0xaa8/0x2950 kernel/exit.c:867 > > > > > do_group_exit+0xd4/0x2a0 kernel/exit.c:1012 > > > > > get_signal+0x21c3/0x2450 kernel/signal.c:2859 > > > > > arch_do_signal_or_restart+0x79/0x5c0 arch/x86/kernel/signal.c:306 > > > > > exit_to_user_mode_loop kernel/entry/common.c:168 [inline] > > > > > exit_to_user_mode_prepare+0x15f/0x250 kernel/entry/common.c:203 > > > > > __syscall_exit_to_user_mode_work kernel/entry/common.c:285 [inline] > > > > > syscall_exit_to_user_mode+0x1d/0x50 kernel/entry/common.c:296 > > > > > do_syscall_64+0x46/0xb0 arch/x86/entry/common.c:86 > > > > > entry_SYSCALL_64_after_hwframe+0x63/0xcd > > > > > ``` > > > > > > > > > > To fix this problem, nr_listen() returns -EINVAL for sockets that successfully nr_connect(). > > > > > > > > > > Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2") > > > > > Reported-by: syzbot+caa188bdfc1eeafeb418@syzkaller.appspotmail.com > > > > > Signed-off-by: Hyunwoo Kim <v4bel@theori.io> > > > > > --- > > > > > net/netrom/af_netrom.c | 5 +++++ > > > > > 1 file changed, 5 insertions(+) > > > > > > > > > > diff --git a/net/netrom/af_netrom.c b/net/netrom/af_netrom.c > > > > > index 6f7f4392cffb..5a4cb796150f 100644 > > > > > --- a/net/netrom/af_netrom.c > > > > > +++ b/net/netrom/af_netrom.c > > > > > @@ -400,6 +400,11 @@ static int nr_listen(struct socket *sock, int backlog) > > > > > struct sock *sk = sock->sk; > > > > > > > > > > lock_sock(sk); > > > > > + if (sock->state != SS_UNCONNECTED) { > > > > > + release_sock(sk); > > > > > + return -EINVAL; > > > > > + } > > > > > + > > > > > if (sk->sk_state != TCP_LISTEN) { > > > > > memset(&nr_sk(sk)->user_addr, 0, AX25_ADDR_LEN); > > > > > sk->sk_max_ack_backlog = backlog; > > > > > -- > > > > > 2.25.1
I noticed this on two systems except I did not get the kernel dump readout as X was running. I thought it may have had to do with VirtualBox, but then I am getting similar behavior on a real install. I was lucky that I did not have X running and have nice output that I OCR'd from a photo of the screen. How to reproduce this: Setup 3 pseudo tty's for loopback to FBB BBS kissnetd -p 3 & kissattach /dev/pts/1 kissparms -c 1 -p lb0 kissattach /dev/pts/2 kissparms -c 1 -p lb1 fbb configured with a real radio on /dev/ttyS0 for port 1 and the lb0 AX.25 kernel interface for port 2. I leave the third pseudo tty for connecting JNOS to FBB over this "loopback" net. If I try to pass bulls (or if I remember correctly even connect from JNOS) it locks cold. Last night (where this screen shot kernel panic came from) was an actual radio port where I was pushing bulls to a newly installed instance of FBB on Arch Linux with the latest kernel 6.2.5. I have not had issues in slackware (and I was also running the latest stable kernel as I am too lazy to keep patching the LTS kernel, but I have not played as much there. I think I may install slack on this box too and see if I can get it to dump like this. If you want the photo, just email me directly as it looks like the OCR is not that good, but I do not know how photos will go over here. Here is the screen dump: 11 fm KOolIP: 1 to NoTJ-1 ctI DISC: 1100216.0098271 BUG: kernel MULL [106216.0109741 #PF supervisor read access in kernel mode 1106216.0121001 APE: error codetox0000) not present page [106216.0132301 PCD 0 P4D 0 [106216.0143551 Dops: 0000 [#11 PREEMPT SMP PTI [106216.0151781 CPU: 0 PID: 39178 Comm: xibbd Tainted: G DE 6.2.5-arch1-1 #1 Fel?0e9497e04500 [106216.0166331 Hardware name: Dell Inc. Optifiex 790/ONKUGV, BIOS A17 03/14/2013 [106216.0177931 RIP: 0010 :ax25_addr_ax25dev+0x44/0xb0 [ax251 [106216.0189621 Code: c1 53 ed 9 b3 61 P9 48 86 14 40 16 01 00 48 85 db 74 41 41 bc 01 00 00 00 eb 08 48 Bb 16 0 03 00 00 eB do fb M Tr 85 co 75 e1 48 Ba bb 90 00 00° [106216.0214491 RSP: 0018: Frffa52b0107bdeB EFLAGS: 00010286 [106216.022714] RAX: 0000000000000000 RBX: FIff9457c84126c0 RCX: 0000000000000000 [106216.023976] RDX: 0000000000000001 RSI: ffffffffc1541100 RDI: ffffa5260107be68 [106216.025229] RBP: ffFfa52b0107be68 ROB: 0000000000000009 R09: 0000000000000000 [106216.0264951 R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000001 [106216.0277661 R13: 0000000000000000 R14: 0000000000000000 R15: 0000000000000000 [106216.029015] FS: 00007/7421a26740 (0000) GS:ffff9458a9c00000 (0000) knIGS: 0000000000000000 [106216.0302591 CS: 0010 DS: 0000 ES: 0000 CRO: 0000000080050033 [106216.031589] CR2: 0000000000000340 CR3: 000000008c804005 CR4: 0000000000060650 [106216.0330441 Call Trace: [106216.0337581 <TASK> [106216.0344601 ax25_bind+0x1e2/0x210 [ax25 0149579197c9004716ce47844d0cb0c56b9a4c841 [106216.0351811 _sys_bind+Oxe8/OxfO [106216.0358931 _x64_sus_bind+0x18/0×20 [106216.0365951 [106216.037296] do_suscal1_64+0x5f/0×90 ? syscall_exit_to_user_mode+0x1b/0×40 [106216.038000] ? do_syscal1_64+0x6b/0x90 [106216.0386931 entry_SYSCALL_64_after_huframe+0x72/0xdc [106216.039390] RIP: 0033:0x717421b3791b [106216.0400931 Code: c3 66 0f 1 44 00 00 48 86 15 51 e4 0c 00 f7 d8 64 89 02 b8 ff fr if ff eb be of 11 44 00 00. 13 of 1e fa b8 31 00 00 00 of 05 <48> 34 01 1 0 ff ff 73 01 c3 48 86 0d 25 e4 0c 00 f7 d8 64 89 01 48 [106216.041578] RSP: 0026:00007ffd15dab688 EFLAGS: 00000206 ORIG RAX: 0000000000000031 [106216.0423321 RAX: fffffffffffffrda RBX: 00007ffd15dabac2 RCX: 000077421b37916 [106216.043092] RDX: 0000000000000048 RSI: 00007ffd15dab700 RDI: 0000000000000009 [106216.043853] RBP: 0000563ef2403900 ROB: 0000000000000004 R09: 00000000ffffffff [106216.0446231 R10: 00007rd15dab6c0 R11: 0000000000000206 R12: 0000000000000009 [106216.045395] R13: 0000000000000048 R14: 0000563ef2403c00 R15: 0000000000000010 [106216.0461671 </TASK> [106216.0469201 Modules linked in: mkiss ax25 crc16 cp210x tun rperdma rdma_cm iw_cm ib_cm ib_core ufat fat intel_rap _msr intel_rapl common ×86_pkg_temp_therma I intel powerc lamp coretemp kum_ intel kum (rqbupass cret10dif pcimul cre32 pcimul polyual cimulni poluual generic grizimul ghash cimulni intel crypta sha512 53s e3 snd Ida _codec_hdmi snd hda_codec_-realtek snd _hda_codec generic rapl ledtrig _audio mousedev snd hda_intel intel _cstate at24 snd_intel_dspefg snd_intel_sdu acp i snd_hda_codec snd_hda_core snd_hudep snd_pem mei_pxp 12c_1801 intel_uncore mei_hdep snd_timer iTCO_wt snd dedbas intel_pc_bxt iTCO_uendor_support mei wat cf gBO211 rikill soundcore pespr 12c _smblis mei me mel lpe_ich mac hid e1000e nfsd auth epcyss nfs acl lockd grace sunrpe uboxnetfit (OF) Uboxmetadp(OF) uboxdru(OF» dm _mod loop fuse bpf_preload ip_tables x tables btrfs blakeZb _generic xor raid_pq liberc32c usbhid cre32c _generic 1915 drm_buddy intel_gtt crc32c_intel drm_di splay_helper sr_mod cdrom cec ttm video uni [106216.0526781 CRZ: 0000000000000340 [106216.053515] ---[ end trace 0000000000000000 1- [106216.0543461 RIP: 0010 :ax25_ addr_ax25deu+0x44/Oxb0 Lax25] [106216.055173] Code: c1 53 8 9 63 61 9 48 86 1d 40 16 01 00 48 85 ab 74 4f 41 bc 01 00 00 00 eb 0B 48 8b 16 48 85 b 74 31 48 86 43 08 48. 89 ef <48> 86 60 0 03 00 00 e8 do fb fr ff 85 cO 75 e1 48 8d bb 90 00 00 [106216.056921] RSP: 0018:ffffa52b0107bded EFLAGS: 00010286 1106216.0578041 RAX: 0000000000000000 RBX: fFff9457c84126c0 RCX: 0000000000000000 [106216.058688] RDX: 0000000000000001 RSI: ffffffffc1541100 RDI: ffffa520107be68 [106216.059577] RBP: ffffa52b0107be68 ROB: 0000000000000009 R09: 0000000000000000 [106216.060473] R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000001 [106216.0613741 R13: 0000000000000000 R14: 0000000000000000 R15: 0000000000000000 [106216.0622811 FS* 0000757421a26740 (0000) GS:ffFf9458a9c00000(0000) knIGS:0000000000000000 [106216.063200] C8: 0010 DS: 0000 ES: 0000 CRO: 0000000080050033 [106216.0641291 CR2: 0000000000000340 CR3: 000000008c804005 CR4: 000000000006060 [106216.0650721 Kernel panic - not syncing: Fatal exception in interrupt [106216.0662201 Kernel Offset: 0x38e00000 from Oxffffffff81000000 (relocation range: Oxfffffff80000000-oxfrrrrrrrrrrrrr) [106216.0672191 ---[ end Kernel panic - not syncing: Fatal exception in interrupt ]-. Regards, Chris KQ6UP
Hi, From: Chris Maness <christopher.maness@gmail.com> Date: Wed, 15 Mar 2023 09:54:03 -0700 > I noticed this on two systems except I did not get the kernel dump > readout as X was running. I thought it may have had to do with > VirtualBox, but then I am getting similar behavior on a real install. > I was lucky that I did not have X running and have nice output that I > OCR'd from a photo of the screen. > > How to reproduce this: > > Setup 3 pseudo tty's for loopback to FBB BBS > kissnetd -p 3 & > kissattach /dev/pts/1 > kissparms -c 1 -p lb0 > kissattach /dev/pts/2 > kissparms -c 1 -p lb1 > fbb configured with a real radio on /dev/ttyS0 for port 1 > and the lb0 AX.25 kernel interface for port 2. > > I leave the third pseudo tty for connecting JNOS to FBB over this > "loopback" net. > > If I try to pass bulls (or if I remember correctly even connect from > JNOS) it locks cold. > > Last night (where this screen shot kernel panic came from) was an > actual radio port where I was pushing bulls to a newly installed > instance of FBB on Arch Linux with the latest kernel 6.2.5. I have > not had issues in slackware (and I was also running the latest stable > kernel as I am too lazy to keep patching the LTS kernel, but I have > not played as much there. I think I may install slack on this box too > and see if I can get it to dump like this. If you want the photo, > just email me directly as it looks like the OCR is not that good, but > I do not know how photos will go over here. > > Here is the screen dump: > > 11 fm KOolIP: 1 to NoTJ-1 ctI DISC: 1100216.0098271 BUG: kernel MULL > > [106216.0109741 #PF supervisor read access in kernel mode > > 1106216.0121001 APE: error codetox0000) > > not present page > > [106216.0132301 PCD 0 P4D 0 > > [106216.0143551 Dops: 0000 [#11 PREEMPT SMP PTI > > [106216.0151781 CPU: 0 PID: 39178 Comm: xibbd Tainted: G > > DE > > 6.2.5-arch1-1 #1 Fel?0e9497e04500 > > [106216.0166331 Hardware name: Dell Inc. Optifiex 790/ONKUGV, BIOS A17 > 03/14/2013 > > [106216.0177931 RIP: 0010 :ax25_addr_ax25dev+0x44/0xb0 [ax251 > > [106216.0189621 Code: c1 53 ed 9 b3 61 P9 48 86 14 40 16 01 00 48 85 > db 74 41 41 bc 01 00 00 00 eb 08 48 Bb 16 > > 0 03 00 00 eB do fb M Tr 85 co 75 e1 48 Ba bb 90 00 00=C2=B0 > > [106216.0214491 RSP: 0018: Frffa52b0107bdeB EFLAGS: 00010286 > > [106216.022714] RAX: 0000000000000000 RBX: FIff9457c84126c0 RCX: > 0000000000000000 > > [106216.023976] RDX: 0000000000000001 RSI: ffffffffc1541100 RDI: > ffffa5260107be68 > > [106216.025229] RBP: ffFfa52b0107be68 ROB: 0000000000000009 R09: > 0000000000000000 > > [106216.0264951 R10: 0000000000000000 R11: 0000000000000000 R12: > 0000000000000001 > > [106216.0277661 R13: 0000000000000000 R14: 0000000000000000 R15: > 0000000000000000 > > [106216.029015] FS: > > 00007/7421a26740 (0000) GS:ffff9458a9c00000 (0000) knIGS: 0000000000000000 > > [106216.0302591 CS: > > 0010 DS: 0000 ES: 0000 CRO: 0000000080050033 > > [106216.031589] CR2: 0000000000000340 CR3: 000000008c804005 CR4: > 0000000000060650 > > [106216.0330441 Call Trace: > > [106216.0337581 <TASK> > > [106216.0344601 ax25_bind+0x1e2/0x210 [ax25 > 0149579197c9004716ce47844d0cb0c56b9a4c841 It seems to be a new null-deref ? I think at least this patch is not related to the issue and same for other patches from Hyunwoom. All of these commits did not touch ax25. 611792920925 ("netrom: Fix use-after-free caused by accept on already connected socket") f2b0b5210f67 ("net/x25: Fix to not accept on connected socket") 14caefcf9837 ("net/rose: Fix to not accept on connected socket") Thanks, Kuniyuki > > [106216.0351811 > > _sys_bind+Oxe8/OxfO > > [106216.0358931 > > _x64_sus_bind+0x18/0=C3=9720 > > [106216.0365951 > > [106216.037296] > > do_suscal1_64+0x5f/0=C3=9790 > > ? syscall_exit_to_user_mode+0x1b/0=C3=9740 > > [106216.038000] > > ? do_syscal1_64+0x6b/0x90 [106216.0386931 > entry_SYSCALL_64_after_huframe+0x72/0xdc > > [106216.039390] RIP: 0033:0x717421b3791b > > [106216.0400931 Code: c3 66 0f 1 44 00 00 48 86 15 51 e4 0c 00 f7 d8 > 64 89 02 b8 ff fr if ff eb be of 11 44 00 00. 13 of 1e fa b8 31 00 00 > 00 of 05 <48> 34 01 1 > > 0 ff ff 73 01 c3 48 86 0d 25 e4 0c 00 f7 d8 64 89 01 48 > > [106216.041578] RSP: 0026:00007ffd15dab688 EFLAGS: 00000206 ORIG RAX: > 0000000000000031 > > [106216.0423321 RAX: fffffffffffffrda RBX: 00007ffd15dabac2 RCX: 000077421b= > 37916 > > [106216.043092] RDX: 0000000000000048 RSI: 00007ffd15dab700 RDI: > 0000000000000009 > > [106216.043853] RBP: 0000563ef2403900 ROB: 0000000000000004 R09: > 00000000ffffffff > > [106216.0446231 R10: 00007rd15dab6c0 R11: 0000000000000206 R12: 00000000000= > 00009 > > [106216.045395] R13: 0000000000000048 R14: 0000563ef2403c00 R15: > 0000000000000010 > > [106216.0461671 > > </TASK> > > [106216.0469201 Modules linked in: mkiss ax25 crc16 cp210x tun rperdma > rdma_cm iw_cm ib_cm ib_core ufat fat intel_rap _msr intel_rapl common > =C3=9786_pkg_temp_therma I intel powerc lamp coretemp kum_ intel kum > (rqbupass cret10dif pcimul cre32 pcimul polyual cimulni poluual > generic grizimul ghash cimulni intel crypta sha512 53s e3 snd Ida > _codec_hdmi snd hda_codec_-realtek snd _hda_codec generic rapl ledtrig > _audio mousedev snd hda_intel intel _cstate at24 snd_intel_dspefg > snd_intel_sdu acp i snd_hda_codec snd_hda_core snd_hudep snd_pem > mei_pxp 12c_1801 intel_uncore mei_hdep snd_timer iTCO_wt snd dedbas > intel_pc_bxt iTCO_uendor_support mei wat cf gBO211 rikill soundcore > pespr 12c _smblis mei me mel lpe_ich mac hid e1000e nfsd auth epcyss > nfs acl lockd grace sunrpe uboxnetfit (OF) Uboxmetadp(OF) uboxdru(OF=C2=BB > dm _mod loop fuse bpf_preload ip_tables x tables btrfs blakeZb > _generic xor raid_pq liberc32c usbhid cre32c _generic 1915 drm_buddy > intel_gtt crc32c_intel drm_di splay_helper sr_mod cdrom cec ttm video > uni > > [106216.0526781 CRZ: 0000000000000340 > > [106216.053515] ---[ end trace 0000000000000000 1- > > [106216.0543461 RIP: 0010 :ax25_ addr_ax25deu+0x44/Oxb0 Lax25] > > [106216.055173] Code: c1 53 8 9 63 61 9 48 86 1d 40 16 01 00 48 85 ab > 74 4f 41 bc 01 00 00 00 eb 0B 48 8b 16 48 85 b 74 31 48 86 43 08 48. > 89 ef <48> 86 60 > > 0 03 00 00 e8 do fb fr ff 85 cO 75 e1 48 8d bb 90 00 00 > > [106216.056921] RSP: 0018:ffffa52b0107bded EFLAGS: 00010286 > > 1106216.0578041 RAX: 0000000000000000 RBX: fFff9457c84126c0 RCX: > 0000000000000000 > > [106216.058688] RDX: 0000000000000001 RSI: ffffffffc1541100 RDI: ffffa52010= > 7be68 > > [106216.059577] RBP: ffffa52b0107be68 ROB: 0000000000000009 R09: > 0000000000000000 > > [106216.060473] R10: 0000000000000000 R11: 0000000000000000 R12: > 0000000000000001 > > [106216.0613741 R13: 0000000000000000 R14: 0000000000000000 R15: > 0000000000000000 > > [106216.0622811 FS* > > 0000757421a26740 (0000) GS:ffFf9458a9c00000(0000) knIGS:0000000000000000 > > [106216.063200] C8: > > 0010 DS: 0000 ES: 0000 CRO: 0000000080050033 > > [106216.0641291 CR2: 0000000000000340 CR3: 000000008c804005 CR4: 0000000000= > 06060 > > [106216.0650721 Kernel panic - not syncing: Fatal exception in interrupt > > [106216.0662201 Kernel Offset: 0x38e00000 from Oxffffffff81000000 > (relocation range: Oxfffffff80000000-oxfrrrrrrrrrrrrr) > > [106216.0672191 > > ---[ end Kernel panic > > - not syncing: Fatal exception in interrupt ]-. > > Regards, > Chris KQ6UP
diff --git a/net/netrom/af_netrom.c b/net/netrom/af_netrom.c index 6f7f4392cffb..5a4cb796150f 100644 --- a/net/netrom/af_netrom.c +++ b/net/netrom/af_netrom.c @@ -400,6 +400,11 @@ static int nr_listen(struct socket *sock, int backlog) struct sock *sk = sock->sk; lock_sock(sk); + if (sock->state != SS_UNCONNECTED) { + release_sock(sk); + return -EINVAL; + } + if (sk->sk_state != TCP_LISTEN) { memset(&nr_sk(sk)->user_addr, 0, AX25_ADDR_LEN); sk->sk_max_ack_backlog = backlog;
If listen() and accept() are called on an AF_NETROM socket that has already been connect()ed, accept() succeeds in connecting. This is because nr_accept() dequeues the skb queued in `sk->sk_receive_queue` in nr_connect(). This causes nr_accept() to allocate and return a sock with the sk of the parent AF_NETROM socket. And here's where use-after-free can happen through complex race conditions: ``` cpu0 cpu1 1. socket_2 = socket(AF_NETROM) listen(socket_2) accepted_socket = accept(socket_2) // loopback connection with socket_1 2. socket_1 = socket(AF_NETROM) nr_create() // sk refcount : 1 connect(socket_1) // loopback connection with socket_2 nr_connect() nr_establish_data_link() nr_write_internal() nr_transmit_buffer() nr_route_frame() nr_loopback_queue() nr_loopback_timer() nr_rx_frame() nr_process_rx_frame() nr_state3_machine() nr_queue_rx_frame() sock_queue_rcv_skb() sock_queue_rcv_skb_reason() __sock_queue_rcv_skb() __skb_queue_tail(list, skb); // list : sk->sk_receive_queue 3. listen(socket_1) nr_listen() uaf_socket = accept(socket_1) nr_accept() skb_dequeue(&sk->sk_receive_queue); 4. close(accepted_socket) nr_release() nr_write_internal(sk, NR_DISCREQ) nr_transmit_buffer() // NR_DISCREQ nr_route_frame() nr_loopback_queue() nr_loopback_timer() nr_rx_frame() // sk : socket_1's sk nr_process_rx_frame() // NR_STATE_3 nr_state3_machine() // NR_DISCREQ nr_disconnect() nr_sk(sk)->state = NR_STATE_0; 5. close(socket_1) // sk refcount : 3 nr_release() // NR_STATE_0 sock_put(sk); // sk refcount : 0 sk_free(sk); close(uaf_socket) nr_release() sock_hold(sk); // UAF ``` KASAN report by syzbot: ``` BUG: KASAN: use-after-free in nr_release+0x66/0x460 net/netrom/af_netrom.c:520 Write of size 4 at addr ffff8880235d8080 by task syz-executor564/5128 Call Trace: <TASK> __dump_stack lib/dump_stack.c:88 [inline] dump_stack_lvl+0xd1/0x138 lib/dump_stack.c:106 print_address_description mm/kasan/report.c:306 [inline] print_report+0x15e/0x461 mm/kasan/report.c:417 kasan_report+0xbf/0x1f0 mm/kasan/report.c:517 check_region_inline mm/kasan/generic.c:183 [inline] kasan_check_range+0x141/0x190 mm/kasan/generic.c:189 instrument_atomic_read_write include/linux/instrumented.h:102 [inline] atomic_fetch_add_relaxed include/linux/atomic/atomic-instrumented.h:116 [inline] __refcount_add include/linux/refcount.h:193 [inline] __refcount_inc include/linux/refcount.h:250 [inline] refcount_inc include/linux/refcount.h:267 [inline] sock_hold include/net/sock.h:775 [inline] nr_release+0x66/0x460 net/netrom/af_netrom.c:520 __sock_release+0xcd/0x280 net/socket.c:650 sock_close+0x1c/0x20 net/socket.c:1365 __fput+0x27c/0xa90 fs/file_table.c:320 task_work_run+0x16f/0x270 kernel/task_work.c:179 exit_task_work include/linux/task_work.h:38 [inline] do_exit+0xaa8/0x2950 kernel/exit.c:867 do_group_exit+0xd4/0x2a0 kernel/exit.c:1012 get_signal+0x21c3/0x2450 kernel/signal.c:2859 arch_do_signal_or_restart+0x79/0x5c0 arch/x86/kernel/signal.c:306 exit_to_user_mode_loop kernel/entry/common.c:168 [inline] exit_to_user_mode_prepare+0x15f/0x250 kernel/entry/common.c:203 __syscall_exit_to_user_mode_work kernel/entry/common.c:285 [inline] syscall_exit_to_user_mode+0x1d/0x50 kernel/entry/common.c:296 do_syscall_64+0x46/0xb0 arch/x86/entry/common.c:86 entry_SYSCALL_64_after_hwframe+0x63/0xcd RIP: 0033:0x7f6c19e3c9b9 Code: Unable to access opcode bytes at 0x7f6c19e3c98f. RSP: 002b:00007fffd4ba2ce8 EFLAGS: 00000246 ORIG_RAX: 0000000000000133 RAX: 0000000000000116 RBX: 0000000000000003 RCX: 00007f6c19e3c9b9 RDX: 0000000000000318 RSI: 00000000200bd000 RDI: 0000000000000006 RBP: 0000000000000003 R08: 000000000000000d R09: 000000000000000d R10: 0000000000000000 R11: 0000000000000246 R12: 000055555566a2c0 R13: 0000000000000011 R14: 0000000000000000 R15: 0000000000000000 </TASK> Allocated by task 5128: kasan_save_stack+0x22/0x40 mm/kasan/common.c:45 kasan_set_track+0x25/0x30 mm/kasan/common.c:52 ____kasan_kmalloc mm/kasan/common.c:371 [inline] ____kasan_kmalloc mm/kasan/common.c:330 [inline] __kasan_kmalloc+0xa3/0xb0 mm/kasan/common.c:380 kasan_kmalloc include/linux/kasan.h:211 [inline] __do_kmalloc_node mm/slab_common.c:968 [inline] __kmalloc+0x5a/0xd0 mm/slab_common.c:981 kmalloc include/linux/slab.h:584 [inline] sk_prot_alloc+0x140/0x290 net/core/sock.c:2038 sk_alloc+0x3a/0x7a0 net/core/sock.c:2091 nr_create+0xb6/0x5f0 net/netrom/af_netrom.c:433 __sock_create+0x359/0x790 net/socket.c:1515 sock_create net/socket.c:1566 [inline] __sys_socket_create net/socket.c:1603 [inline] __sys_socket_create net/socket.c:1588 [inline] __sys_socket+0x133/0x250 net/socket.c:1636 __do_sys_socket net/socket.c:1649 [inline] __se_sys_socket net/socket.c:1647 [inline] __x64_sys_socket+0x73/0xb0 net/socket.c:1647 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x39/0xb0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x63/0xcd Freed by task 5128: kasan_save_stack+0x22/0x40 mm/kasan/common.c:45 kasan_set_track+0x25/0x30 mm/kasan/common.c:52 kasan_save_free_info+0x2b/0x40 mm/kasan/generic.c:518 ____kasan_slab_free mm/kasan/common.c:236 [inline] ____kasan_slab_free+0x13b/0x1a0 mm/kasan/common.c:200 kasan_slab_free include/linux/kasan.h:177 [inline] __cache_free mm/slab.c:3394 [inline] __do_kmem_cache_free mm/slab.c:3580 [inline] __kmem_cache_free+0xcd/0x3b0 mm/slab.c:3587 sk_prot_free net/core/sock.c:2074 [inline] __sk_destruct+0x5df/0x750 net/core/sock.c:2166 sk_destruct net/core/sock.c:2181 [inline] __sk_free+0x175/0x460 net/core/sock.c:2192 sk_free+0x7c/0xa0 net/core/sock.c:2203 sock_put include/net/sock.h:1991 [inline] nr_release+0x39e/0x460 net/netrom/af_netrom.c:554 __sock_release+0xcd/0x280 net/socket.c:650 sock_close+0x1c/0x20 net/socket.c:1365 __fput+0x27c/0xa90 fs/file_table.c:320 task_work_run+0x16f/0x270 kernel/task_work.c:179 exit_task_work include/linux/task_work.h:38 [inline] do_exit+0xaa8/0x2950 kernel/exit.c:867 do_group_exit+0xd4/0x2a0 kernel/exit.c:1012 get_signal+0x21c3/0x2450 kernel/signal.c:2859 arch_do_signal_or_restart+0x79/0x5c0 arch/x86/kernel/signal.c:306 exit_to_user_mode_loop kernel/entry/common.c:168 [inline] exit_to_user_mode_prepare+0x15f/0x250 kernel/entry/common.c:203 __syscall_exit_to_user_mode_work kernel/entry/common.c:285 [inline] syscall_exit_to_user_mode+0x1d/0x50 kernel/entry/common.c:296 do_syscall_64+0x46/0xb0 arch/x86/entry/common.c:86 entry_SYSCALL_64_after_hwframe+0x63/0xcd ``` To fix this problem, nr_listen() returns -EINVAL for sockets that successfully nr_connect(). Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2") Reported-by: syzbot+caa188bdfc1eeafeb418@syzkaller.appspotmail.com Signed-off-by: Hyunwoo Kim <v4bel@theori.io> --- net/netrom/af_netrom.c | 5 +++++ 1 file changed, 5 insertions(+)