When the link layer connection is broken, the rose->neighbour is
set to null. But rose->neighbour could be used by rose_connection()
and rose_release() later, because there is no synchronization among
them. As a result, the null-ptr-deref bugs will happen.

One of the null-ptr-deref bugs is shown below:

    (thread 1)                  |        (thread 2)
                                |  rose_connect
rose_kill_by_neigh              |    lock_sock(sk)
  spin_lock_bh(&rose_list_lock) |    if (!rose->neighbour)
  rose->neighbour = NULL;//(1)  |
                                |    rose->neighbour->use++;//(2)

The rose->neighbour is set to null in position (1) and dereferenced
in position (2).

The KASAN report triggered by POC is shown below:

KASAN: null-ptr-deref in range [0x0000000000000028-0x000000000000002f]
...
RIP: 0010:rose_connect+0x6c2/0xf30
RSP: 0018:ffff88800ab47d60 EFLAGS: 00000206
RAX: 0000000000000005 RBX: 000000000000002a RCX: 0000000000000000
RDX: ffff88800ab38000 RSI: ffff88800ab47e48 RDI: ffff88800ab38309
RBP: dffffc0000000000 R08: 0000000000000000 R09: ffffed1001567062
R10: dfffe91001567063 R11: 1ffff11001567061 R12: 1ffff11000d17cd0
R13: ffff8880068be680 R14: 0000000000000002 R15: 1ffff11000d17cd0
...
Call Trace:
  <TASK>
  ? __local_bh_enable_ip+0x54/0x80
  ? selinux_netlbl_socket_connect+0x26/0x30
  ? rose_bind+0x5b0/0x5b0
  __sys_connect+0x216/0x280
  __x64_sys_connect+0x71/0x80
  do_syscall_64+0x43/0x90
  entry_SYSCALL_64_after_hwframe+0x46/0xb0

This patch adds lock_sock() in rose_kill_by_neigh() in order to
synchronize with rose_connect() and rose_release(). Then, changing
type of 'neighbour->use' from unsigned short to atomic_t in order to
mitigate race conditions caused by holding different socket lock while
updating 'neighbour->use'.

Meanwhile, this patch adds sock_hold() protected by rose_list_lock
that could synchronize with rose_remove_socket() in order to mitigate
UAF bug caused by lock_sock() we add.

What's more, there is no need using rose_neigh_list_lock to protect
rose_kill_by_neigh(). Because we have already used rose_neigh_list_lock
to protect the state change of rose_neigh in rose_link_failed(), which
is well synchronized.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
---
Changes in v6:
  - Change sk_for_each() to sk_for_each_safe().
  - Change type of 'neighbour->use' from unsigned short to atomic_t.

 include/net/rose.h    |  2 +-
 net/rose/af_rose.c    | 19 +++++++++++++------
 net/rose/rose_in.c    | 12 ++++++------
 net/rose/rose_route.c | 24 ++++++++++++------------
 net/rose/rose_timer.c |  2 +-
 5 files changed, 33 insertions(+), 26 deletions(-)

diff --git a/include/net/rose.h b/include/net/rose.h
index 0f0a4ce0fee..d5ddebc556d 100644
--- a/include/net/rose.h
+++ b/include/net/rose.h
@@ -95,7 +95,7 @@ struct rose_neigh {
 	ax25_cb			*ax25;
 	struct net_device		*dev;
 	unsigned short		count;
-	unsigned short		use;
+	atomic_t		use;
 	unsigned int		number;
 	char			restarted;
 	char			dce_mode;
diff --git a/net/rose/af_rose.c b/net/rose/af_rose.c
index bf2d986a6bc..54e7b76c4f3 100644
--- a/net/rose/af_rose.c
+++ b/net/rose/af_rose.c
@@ -163,16 +163,23 @@ static void rose_remove_socket(struct sock *sk)
 void rose_kill_by_neigh(struct rose_neigh *neigh)
 {
 	struct sock *s;
+	struct hlist_node *tmp;
 
 	spin_lock_bh(&rose_list_lock);
-	sk_for_each(s, &rose_list) {
+	sk_for_each_safe(s, tmp, &rose_list) {
 		struct rose_sock *rose = rose_sk(s);
 
+		sock_hold(s);
+		spin_unlock_bh(&rose_list_lock);
+		lock_sock(s);
 		if (rose->neighbour == neigh) {
 			rose_disconnect(s, ENETUNREACH, ROSE_OUT_OF_ORDER, 0);
-			rose->neighbour->use--;
+			atomic_dec(&rose->neighbour->use);
 			rose->neighbour = NULL;
 		}
+		release_sock(s);
+		sock_put(s);
+		spin_lock_bh(&rose_list_lock);
 	}
 	spin_unlock_bh(&rose_list_lock);
 }
@@ -191,7 +198,7 @@ static void rose_kill_by_device(struct net_device *dev)
 		if (rose->device == dev) {
 			rose_disconnect(s, ENETUNREACH, ROSE_OUT_OF_ORDER, 0);
 			if (rose->neighbour)
-				rose->neighbour->use--;
+				atomic_dec(&rose->neighbour->use);
 			rose->device = NULL;
 		}
 	}
@@ -618,7 +625,7 @@ static int rose_release(struct socket *sock)
 		break;
 
 	case ROSE_STATE_2:
-		rose->neighbour->use--;
+		atomic_dec(&rose->neighbour->use);
 		release_sock(sk);
 		rose_disconnect(sk, 0, -1, -1);
 		lock_sock(sk);
@@ -819,7 +826,7 @@ static int rose_connect(struct socket *sock, struct sockaddr *uaddr, int addr_le
 
 	rose->state = ROSE_STATE_1;
 
-	rose->neighbour->use++;
+	atomic_inc(&rose->neighbour->use);
 
 	rose_write_internal(sk, ROSE_CALL_REQUEST);
 	rose_start_heartbeat(sk);
@@ -1019,7 +1026,7 @@ int rose_rx_call_request(struct sk_buff *skb, struct net_device *dev, struct ros
 	make_rose->device        = dev;
 	make_rose->facilities    = facilities;
 
-	make_rose->neighbour->use++;
+	atomic_inc(&make_rose->neighbour->use);
 
 	if (rose_sk(sk)->defer) {
 		make_rose->state = ROSE_STATE_5;
diff --git a/net/rose/rose_in.c b/net/rose/rose_in.c
index 4d67f36dce1..86168f29943 100644
--- a/net/rose/rose_in.c
+++ b/net/rose/rose_in.c
@@ -56,7 +56,7 @@ static int rose_state1_machine(struct sock *sk, struct sk_buff *skb, int framety
 	case ROSE_CLEAR_REQUEST:
 		rose_write_internal(sk, ROSE_CLEAR_CONFIRMATION);
 		rose_disconnect(sk, ECONNREFUSED, skb->data[3], skb->data[4]);
-		rose->neighbour->use--;
+		atomic_dec(&rose->neighbour->use);
 		break;
 
 	default:
@@ -79,12 +79,12 @@ static int rose_state2_machine(struct sock *sk, struct sk_buff *skb, int framety
 	case ROSE_CLEAR_REQUEST:
 		rose_write_internal(sk, ROSE_CLEAR_CONFIRMATION);
 		rose_disconnect(sk, 0, skb->data[3], skb->data[4]);
-		rose->neighbour->use--;
+		atomic_dec(&rose->neighbour->use);
 		break;
 
 	case ROSE_CLEAR_CONFIRMATION:
 		rose_disconnect(sk, 0, -1, -1);
-		rose->neighbour->use--;
+		atomic_dec(&rose->neighbour->use);
 		break;
 
 	default:
@@ -120,7 +120,7 @@ static int rose_state3_machine(struct sock *sk, struct sk_buff *skb, int framety
 	case ROSE_CLEAR_REQUEST:
 		rose_write_internal(sk, ROSE_CLEAR_CONFIRMATION);
 		rose_disconnect(sk, 0, skb->data[3], skb->data[4]);
-		rose->neighbour->use--;
+		atomic_dec(&rose->neighbour->use);
 		break;
 
 	case ROSE_RR:
@@ -233,7 +233,7 @@ static int rose_state4_machine(struct sock *sk, struct sk_buff *skb, int framety
 	case ROSE_CLEAR_REQUEST:
 		rose_write_internal(sk, ROSE_CLEAR_CONFIRMATION);
 		rose_disconnect(sk, 0, skb->data[3], skb->data[4]);
-		rose->neighbour->use--;
+		atomic_dec(&rose->neighbour->use);
 		break;
 
 	default:
@@ -253,7 +253,7 @@ static int rose_state5_machine(struct sock *sk, struct sk_buff *skb, int framety
 	if (frametype == ROSE_CLEAR_REQUEST) {
 		rose_write_internal(sk, ROSE_CLEAR_CONFIRMATION);
 		rose_disconnect(sk, 0, skb->data[3], skb->data[4]);
-		rose_sk(sk)->neighbour->use--;
+		atomic_dec(&rose_sk(sk)->neighbour->use);
 	}
 
 	return 0;
diff --git a/net/rose/rose_route.c b/net/rose/rose_route.c
index eb0b8197ac8..8be00a44540 100644
--- a/net/rose/rose_route.c
+++ b/net/rose/rose_route.c
@@ -93,7 +93,7 @@ static int __must_check rose_add_node(struct rose_route_struct *rose_route,
 		rose_neigh->ax25      = NULL;
 		rose_neigh->dev       = dev;
 		rose_neigh->count     = 0;
-		rose_neigh->use       = 0;
+		atomic_set(&rose_neigh->use, 0);
 		rose_neigh->dce_mode  = 0;
 		rose_neigh->loopback  = 0;
 		rose_neigh->number    = rose_neigh_no++;
@@ -263,10 +263,10 @@ static void rose_remove_route(struct rose_route *rose_route)
 	struct rose_route *s;
 
 	if (rose_route->neigh1 != NULL)
-		rose_route->neigh1->use--;
+		atomic_dec(&rose_route->neigh1->use);
 
 	if (rose_route->neigh2 != NULL)
-		rose_route->neigh2->use--;
+		atomic_dec(&rose_route->neigh2->use);
 
 	if ((s = rose_route_list) == rose_route) {
 		rose_route_list = rose_route->next;
@@ -331,7 +331,7 @@ static int rose_del_node(struct rose_route_struct *rose_route,
 		if (rose_node->neighbour[i] == rose_neigh) {
 			rose_neigh->count--;
 
-			if (rose_neigh->count == 0 && rose_neigh->use == 0)
+			if (rose_neigh->count == 0 && atomic_read(&rose_neigh->use) == 0)
 				rose_remove_neigh(rose_neigh);
 
 			rose_node->count--;
@@ -381,7 +381,7 @@ void rose_add_loopback_neigh(void)
 	sn->ax25      = NULL;
 	sn->dev       = NULL;
 	sn->count     = 0;
-	sn->use       = 0;
+	atomic_set(&sn->use, 0);
 	sn->dce_mode  = 1;
 	sn->loopback  = 1;
 	sn->number    = rose_neigh_no++;
@@ -573,7 +573,7 @@ static int rose_clear_routes(void)
 		s          = rose_neigh;
 		rose_neigh = rose_neigh->next;
 
-		if (s->use == 0 && !s->loopback) {
+		if (atomic_read(&s->use) == 0 && !s->loopback) {
 			s->count = 0;
 			rose_remove_neigh(s);
 		}
@@ -789,13 +789,13 @@ static void rose_del_route_by_neigh(struct rose_neigh *rose_neigh)
 		}
 
 		if (rose_route->neigh1 == rose_neigh) {
-			rose_route->neigh1->use--;
+			atomic_dec(&rose_route->neigh1->use);
 			rose_route->neigh1 = NULL;
 			rose_transmit_clear_request(rose_route->neigh2, rose_route->lci2, ROSE_OUT_OF_ORDER, 0);
 		}
 
 		if (rose_route->neigh2 == rose_neigh) {
-			rose_route->neigh2->use--;
+			atomic_dec(&rose_route->neigh2->use);
 			rose_route->neigh2 = NULL;
 			rose_transmit_clear_request(rose_route->neigh1, rose_route->lci1, ROSE_OUT_OF_ORDER, 0);
 		}
@@ -924,7 +924,7 @@ int rose_route_frame(struct sk_buff *skb, ax25_cb *ax25)
 			rose_clear_queues(sk);
 			rose->cause	 = ROSE_NETWORK_CONGESTION;
 			rose->diagnostic = 0;
-			rose->neighbour->use--;
+			atomic_dec(&rose->neighbour->use);
 			rose->neighbour	 = NULL;
 			rose->lci	 = 0;
 			rose->state	 = ROSE_STATE_0;
@@ -1067,8 +1067,8 @@ int rose_route_frame(struct sk_buff *skb, ax25_cb *ax25)
 	rose_route->lci2      = new_lci;
 	rose_route->neigh2    = new_neigh;
 
-	rose_route->neigh1->use++;
-	rose_route->neigh2->use++;
+	atomic_inc(&rose_route->neigh1->use);
+	atomic_inc(&rose_route->neigh2->use);
 
 	rose_route->next = rose_route_list;
 	rose_route_list  = rose_route;
@@ -1195,7 +1195,7 @@ static int rose_neigh_show(struct seq_file *seq, void *v)
 			   (rose_neigh->loopback) ? "RSLOOP-0" : ax2asc(buf, &rose_neigh->callsign),
 			   rose_neigh->dev ? rose_neigh->dev->name : "???",
 			   rose_neigh->count,
-			   rose_neigh->use,
+			   atomic_read(&rose_neigh->use),
 			   (rose_neigh->dce_mode) ? "DCE" : "DTE",
 			   (rose_neigh->restarted) ? "yes" : "no",
 			   ax25_display_timer(&rose_neigh->t0timer) / HZ,
diff --git a/net/rose/rose_timer.c b/net/rose/rose_timer.c
index f06ddbed3fe..9dfd4eae5d5 100644
--- a/net/rose/rose_timer.c
+++ b/net/rose/rose_timer.c
@@ -171,7 +171,7 @@ static void rose_timer_expiry(struct timer_list *t)
 		break;
 
 	case ROSE_STATE_2:	/* T3 */
-		rose->neighbour->use--;
+		atomic_dec(&rose->neighbour->use);
 		rose_disconnect(sk, ETIMEDOUT, -1, -1);
 		break;
 
-- 
2.17.1

On Mon, 2022-07-11 at 09:31 +0800, Duoming Zhou wrote:
> When the link layer connection is broken, the rose->neighbour is
> set to null. But rose->neighbour could be used by rose_connection()
> and rose_release() later, because there is no synchronization among
> them. As a result, the null-ptr-deref bugs will happen.
> 
> One of the null-ptr-deref bugs is shown below:
> 
>     (thread 1)                  |        (thread 2)
>                                 |  rose_connect
> rose_kill_by_neigh              |    lock_sock(sk)
>   spin_lock_bh(&rose_list_lock) |    if (!rose->neighbour)
>   rose->neighbour = NULL;//(1)  |
>                                 |    rose->neighbour->use++;//(2)
> 
> The rose->neighbour is set to null in position (1) and dereferenced
> in position (2).
> 
> The KASAN report triggered by POC is shown below:
> 
> KASAN: null-ptr-deref in range [0x0000000000000028-0x000000000000002f]
> ...
> RIP: 0010:rose_connect+0x6c2/0xf30
> RSP: 0018:ffff88800ab47d60 EFLAGS: 00000206
> RAX: 0000000000000005 RBX: 000000000000002a RCX: 0000000000000000
> RDX: ffff88800ab38000 RSI: ffff88800ab47e48 RDI: ffff88800ab38309
> RBP: dffffc0000000000 R08: 0000000000000000 R09: ffffed1001567062
> R10: dfffe91001567063 R11: 1ffff11001567061 R12: 1ffff11000d17cd0
> R13: ffff8880068be680 R14: 0000000000000002 R15: 1ffff11000d17cd0
> ...
> Call Trace:
>   <TASK>
>   ? __local_bh_enable_ip+0x54/0x80
>   ? selinux_netlbl_socket_connect+0x26/0x30
>   ? rose_bind+0x5b0/0x5b0
>   __sys_connect+0x216/0x280
>   __x64_sys_connect+0x71/0x80
>   do_syscall_64+0x43/0x90
>   entry_SYSCALL_64_after_hwframe+0x46/0xb0
> 
> This patch adds lock_sock() in rose_kill_by_neigh() in order to
> synchronize with rose_connect() and rose_release(). Then, changing
> type of 'neighbour->use' from unsigned short to atomic_t in order to
> mitigate race conditions caused by holding different socket lock while
> updating 'neighbour->use'.
> 
> Meanwhile, this patch adds sock_hold() protected by rose_list_lock
> that could synchronize with rose_remove_socket() in order to mitigate
> UAF bug caused by lock_sock() we add.
> 
> What's more, there is no need using rose_neigh_list_lock to protect
> rose_kill_by_neigh(). Because we have already used rose_neigh_list_lock
> to protect the state change of rose_neigh in rose_link_failed(), which
> is well synchronized.
> 
> Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
> Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
> ---
> Changes in v6:
>   - Change sk_for_each() to sk_for_each_safe().
>   - Change type of 'neighbour->use' from unsigned short to atomic_t.
> 
>  include/net/rose.h    |  2 +-
>  net/rose/af_rose.c    | 19 +++++++++++++------
>  net/rose/rose_in.c    | 12 ++++++------
>  net/rose/rose_route.c | 24 ++++++++++++------------
>  net/rose/rose_timer.c |  2 +-
>  5 files changed, 33 insertions(+), 26 deletions(-)
> 
> diff --git a/include/net/rose.h b/include/net/rose.h
> index 0f0a4ce0fee..d5ddebc556d 100644
> --- a/include/net/rose.h
> +++ b/include/net/rose.h
> @@ -95,7 +95,7 @@ struct rose_neigh {
>  	ax25_cb			*ax25;
>  	struct net_device		*dev;
>  	unsigned short		count;
> -	unsigned short		use;
> +	atomic_t		use;
>  	unsigned int		number;
>  	char			restarted;
>  	char			dce_mode;
> diff --git a/net/rose/af_rose.c b/net/rose/af_rose.c
> index bf2d986a6bc..54e7b76c4f3 100644
> --- a/net/rose/af_rose.c
> +++ b/net/rose/af_rose.c
> @@ -163,16 +163,23 @@ static void rose_remove_socket(struct sock *sk)
>  void rose_kill_by_neigh(struct rose_neigh *neigh)
>  {
>  	struct sock *s;
> +	struct hlist_node *tmp;
>  
>  	spin_lock_bh(&rose_list_lock);
> -	sk_for_each(s, &rose_list) {
> +	sk_for_each_safe(s, tmp, &rose_list) {
>  		struct rose_sock *rose = rose_sk(s);
>  
> +		sock_hold(s);
> +		spin_unlock_bh(&rose_list_lock);
> +		lock_sock(s);
>  		if (rose->neighbour == neigh) {
>  			rose_disconnect(s, ENETUNREACH, ROSE_OUT_OF_ORDER, 0);
> -			rose->neighbour->use--;
> +			atomic_dec(&rose->neighbour->use);
>  			rose->neighbour = NULL;
>  		}
> +		release_sock(s);
> +		sock_put(s);

I'm sorry, this does not work. At this point both 's' and 'tmp' sockets
can be freed and reused. Both iterators are not valid anymore when you
acquire the 'rose_list_lock' later.

I really think you should resort to something similar to the following
(completelly untested, just to give an idea). In any case it would be
better to split this change in 2 separate patches: the first patch
replaces 'int use;' with an antomic_t and the 2nd one addresses the
race you describe above.

---
diff --git a/net/rose/af_rose.c b/net/rose/af_rose.c
index bf2d986a6bc3..27b1027aaedf 100644
--- a/net/rose/af_rose.c
+++ b/net/rose/af_rose.c
@@ -156,25 +156,45 @@ static void rose_remove_socket(struct sock *sk)
 	spin_unlock_bh(&rose_list_lock);
 }
 
+static DEFINE_MUTEX(kill_lock);
+
 /*
  *	Kill all bound sockets on a broken link layer connection to a
  *	particular neighbour.
  */
 void rose_kill_by_neigh(struct rose_neigh *neigh)
 {
-	struct sock *s;
+	HLIST_HEAD(rose_list_copy);
+	struct sock *s, *tmp;
+
+	mutex_lock(&kill_lock);
 
 	spin_lock_bh(&rose_list_lock);
 	sk_for_each(s, &rose_list) {
+		sock_hold(s);
+		/* sk_bind_node is apparently unused by rose. Alternatively
+		 * you can add another hlist_node to rose_sock and use it here
+		 */
+		sk_add_bind_node(s, &rose_list_copy);
+	}
+	spin_unlock_bh(&rose_list_lock);
+
+	hlist_for_each_entry_safe(s, tmp, &rose_list_copy, sk_bind_node) {
 		struct rose_sock *rose = rose_sk(s);
 
+		__sk_del_bind_node(s);
+		lock_sock(s);
 		if (rose->neighbour == neigh) {
 			rose_disconnect(s, ENETUNREACH, ROSE_OUT_OF_ORDER, 0);
-			rose->neighbour->use--;
+			atomic_dec(&rose->neighbour->use);
 			rose->neighbour = NULL;
 		}
+		release_sock(s);
+
+		sock_put(s);
 	}
-	spin_unlock_bh(&rose_list_lock);
+
+	mutex_unlock(&kill_lock);
 }
 
 /*
---
/P

Hello,

On Tue, 12 Jul 2022 13:00:49 +0200 Paolo Abeni wrote:

> On Mon, 2022-07-11 at 09:31 +0800, Duoming Zhou wrote:
> > When the link layer connection is broken, the rose->neighbour is
> > set to null. But rose->neighbour could be used by rose_connection()
> > and rose_release() later, because there is no synchronization among
> > them. As a result, the null-ptr-deref bugs will happen.
> > 
> > One of the null-ptr-deref bugs is shown below:
> > 
> >     (thread 1)                  |        (thread 2)
> >                                 |  rose_connect
> > rose_kill_by_neigh              |    lock_sock(sk)
> >   spin_lock_bh(&rose_list_lock) |    if (!rose->neighbour)
> >   rose->neighbour = NULL;//(1)  |
> >                                 |    rose->neighbour->use++;//(2)
> > 
> > The rose->neighbour is set to null in position (1) and dereferenced
> > in position (2).
> > 
> > The KASAN report triggered by POC is shown below:
> > 
> > KASAN: null-ptr-deref in range [0x0000000000000028-0x000000000000002f]
> > ...
> > RIP: 0010:rose_connect+0x6c2/0xf30
> > RSP: 0018:ffff88800ab47d60 EFLAGS: 00000206
> > RAX: 0000000000000005 RBX: 000000000000002a RCX: 0000000000000000
> > RDX: ffff88800ab38000 RSI: ffff88800ab47e48 RDI: ffff88800ab38309
> > RBP: dffffc0000000000 R08: 0000000000000000 R09: ffffed1001567062
> > R10: dfffe91001567063 R11: 1ffff11001567061 R12: 1ffff11000d17cd0
> > R13: ffff8880068be680 R14: 0000000000000002 R15: 1ffff11000d17cd0
> > ...
> > Call Trace:
> >   <TASK>
> >   ? __local_bh_enable_ip+0x54/0x80
> >   ? selinux_netlbl_socket_connect+0x26/0x30
> >   ? rose_bind+0x5b0/0x5b0
> >   __sys_connect+0x216/0x280
> >   __x64_sys_connect+0x71/0x80
> >   do_syscall_64+0x43/0x90
> >   entry_SYSCALL_64_after_hwframe+0x46/0xb0
> > 
> > This patch adds lock_sock() in rose_kill_by_neigh() in order to
> > synchronize with rose_connect() and rose_release(). Then, changing
> > type of 'neighbour->use' from unsigned short to atomic_t in order to
> > mitigate race conditions caused by holding different socket lock while
> > updating 'neighbour->use'.
> > 
> > Meanwhile, this patch adds sock_hold() protected by rose_list_lock
> > that could synchronize with rose_remove_socket() in order to mitigate
> > UAF bug caused by lock_sock() we add.
> > 
> > What's more, there is no need using rose_neigh_list_lock to protect
> > rose_kill_by_neigh(). Because we have already used rose_neigh_list_lock
> > to protect the state change of rose_neigh in rose_link_failed(), which
> > is well synchronized.
> > 
> > Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
> > Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
> > ---
> > Changes in v6:
> >   - Change sk_for_each() to sk_for_each_safe().
> >   - Change type of 'neighbour->use' from unsigned short to atomic_t.
> > 
> >  include/net/rose.h    |  2 +-
> >  net/rose/af_rose.c    | 19 +++++++++++++------
> >  net/rose/rose_in.c    | 12 ++++++------
> >  net/rose/rose_route.c | 24 ++++++++++++------------
> >  net/rose/rose_timer.c |  2 +-
> >  5 files changed, 33 insertions(+), 26 deletions(-)
> > 
> > diff --git a/include/net/rose.h b/include/net/rose.h
> > index 0f0a4ce0fee..d5ddebc556d 100644
> > --- a/include/net/rose.h
> > +++ b/include/net/rose.h
> > @@ -95,7 +95,7 @@ struct rose_neigh {
> >  	ax25_cb			*ax25;
> >  	struct net_device		*dev;
> >  	unsigned short		count;
> > -	unsigned short		use;
> > +	atomic_t		use;
> >  	unsigned int		number;
> >  	char			restarted;
> >  	char			dce_mode;
> > diff --git a/net/rose/af_rose.c b/net/rose/af_rose.c
> > index bf2d986a6bc..54e7b76c4f3 100644
> > --- a/net/rose/af_rose.c
> > +++ b/net/rose/af_rose.c
> > @@ -163,16 +163,23 @@ static void rose_remove_socket(struct sock *sk)
> >  void rose_kill_by_neigh(struct rose_neigh *neigh)
> >  {
> >  	struct sock *s;
> > +	struct hlist_node *tmp;
> >  
> >  	spin_lock_bh(&rose_list_lock);
> > -	sk_for_each(s, &rose_list) {
> > +	sk_for_each_safe(s, tmp, &rose_list) {
> >  		struct rose_sock *rose = rose_sk(s);
> >  
> > +		sock_hold(s);
> > +		spin_unlock_bh(&rose_list_lock);
> > +		lock_sock(s);
> >  		if (rose->neighbour == neigh) {
> >  			rose_disconnect(s, ENETUNREACH, ROSE_OUT_OF_ORDER, 0);
> > -			rose->neighbour->use--;
> > +			atomic_dec(&rose->neighbour->use);
> >  			rose->neighbour = NULL;
> >  		}
> > +		release_sock(s);
> > +		sock_put(s);
> 
> I'm sorry, this does not work. At this point both 's' and 'tmp' sockets
> can be freed and reused. Both iterators are not valid anymore when you
> acquire the 'rose_list_lock' later.

Thank you for your time and reply! But I think both 's' and 'tmp' can not
be freed and reused in rose_kill_by_neigh(). Because rose_remove_socket()
calls sk_del_node_init() which is protected by rose_list_lock to delete the
socket node from the hlist and if sk->sk_refcnt equals to 1, the socket will
be deallocated.

static void rose_remove_socket(struct sock *sk)
{
	spin_lock_bh(&rose_list_lock);
	sk_del_node_init(sk);
	spin_unlock_bh(&rose_list_lock);
}

https://elixir.bootlin.com/linux/v5.19-rc6/source/net/rose/af_rose.c#L152

Both 's' and 'tmp' in rose_kill_by_neigh() is also protected by rose_list_lock.

If the socket is deleted from the hlist, sk_for_each_safe() could not find
the socket and the UAF bug could be prevented. 

If the socket could be found by sk_for_each_safe(), we use sock_hold(s)
to increase the refcount of the socket. As a result, the UAF bugs could
be prevented.

Best regards,
Duoming Zhou

On Wed, 2022-07-13 at 15:50 +0800, duoming@zju.edu.cn wrote:
> Hello,
> 
> On Tue, 12 Jul 2022 13:00:49 +0200 Paolo Abeni wrote:
> 
> > On Mon, 2022-07-11 at 09:31 +0800, Duoming Zhou wrote:
> > > When the link layer connection is broken, the rose->neighbour is
> > > set to null. But rose->neighbour could be used by rose_connection()
> > > and rose_release() later, because there is no synchronization among
> > > them. As a result, the null-ptr-deref bugs will happen.
> > > 
> > > One of the null-ptr-deref bugs is shown below:
> > > 
> > >     (thread 1)                  |        (thread 2)
> > >                                 |  rose_connect
> > > rose_kill_by_neigh              |    lock_sock(sk)
> > >   spin_lock_bh(&rose_list_lock) |    if (!rose->neighbour)
> > >   rose->neighbour = NULL;//(1)  |
> > >                                 |    rose->neighbour->use++;//(2)
> > > 
> > > The rose->neighbour is set to null in position (1) and dereferenced
> > > in position (2).
> > > 
> > > The KASAN report triggered by POC is shown below:
> > > 
> > > KASAN: null-ptr-deref in range [0x0000000000000028-0x000000000000002f]
> > > ...
> > > RIP: 0010:rose_connect+0x6c2/0xf30
> > > RSP: 0018:ffff88800ab47d60 EFLAGS: 00000206
> > > RAX: 0000000000000005 RBX: 000000000000002a RCX: 0000000000000000
> > > RDX: ffff88800ab38000 RSI: ffff88800ab47e48 RDI: ffff88800ab38309
> > > RBP: dffffc0000000000 R08: 0000000000000000 R09: ffffed1001567062
> > > R10: dfffe91001567063 R11: 1ffff11001567061 R12: 1ffff11000d17cd0
> > > R13: ffff8880068be680 R14: 0000000000000002 R15: 1ffff11000d17cd0
> > > ...
> > > Call Trace:
> > >   <TASK>
> > >   ? __local_bh_enable_ip+0x54/0x80
> > >   ? selinux_netlbl_socket_connect+0x26/0x30
> > >   ? rose_bind+0x5b0/0x5b0
> > >   __sys_connect+0x216/0x280
> > >   __x64_sys_connect+0x71/0x80
> > >   do_syscall_64+0x43/0x90
> > >   entry_SYSCALL_64_after_hwframe+0x46/0xb0
> > > 
> > > This patch adds lock_sock() in rose_kill_by_neigh() in order to
> > > synchronize with rose_connect() and rose_release(). Then, changing
> > > type of 'neighbour->use' from unsigned short to atomic_t in order to
> > > mitigate race conditions caused by holding different socket lock while
> > > updating 'neighbour->use'.
> > > 
> > > Meanwhile, this patch adds sock_hold() protected by rose_list_lock
> > > that could synchronize with rose_remove_socket() in order to mitigate
> > > UAF bug caused by lock_sock() we add.
> > > 
> > > What's more, there is no need using rose_neigh_list_lock to protect
> > > rose_kill_by_neigh(). Because we have already used rose_neigh_list_lock
> > > to protect the state change of rose_neigh in rose_link_failed(), which
> > > is well synchronized.
> > > 
> > > Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
> > > Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
> > > ---
> > > Changes in v6:
> > >   - Change sk_for_each() to sk_for_each_safe().
> > >   - Change type of 'neighbour->use' from unsigned short to atomic_t.
> > > 
> > >  include/net/rose.h    |  2 +-
> > >  net/rose/af_rose.c    | 19 +++++++++++++------
> > >  net/rose/rose_in.c    | 12 ++++++------
> > >  net/rose/rose_route.c | 24 ++++++++++++------------
> > >  net/rose/rose_timer.c |  2 +-
> > >  5 files changed, 33 insertions(+), 26 deletions(-)
> > > 
> > > diff --git a/include/net/rose.h b/include/net/rose.h
> > > index 0f0a4ce0fee..d5ddebc556d 100644
> > > --- a/include/net/rose.h
> > > +++ b/include/net/rose.h
> > > @@ -95,7 +95,7 @@ struct rose_neigh {
> > >  	ax25_cb			*ax25;
> > >  	struct net_device		*dev;
> > >  	unsigned short		count;
> > > -	unsigned short		use;
> > > +	atomic_t		use;
> > >  	unsigned int		number;
> > >  	char			restarted;
> > >  	char			dce_mode;
> > > diff --git a/net/rose/af_rose.c b/net/rose/af_rose.c
> > > index bf2d986a6bc..54e7b76c4f3 100644
> > > --- a/net/rose/af_rose.c
> > > +++ b/net/rose/af_rose.c
> > > @@ -163,16 +163,23 @@ static void rose_remove_socket(struct sock *sk)
> > >  void rose_kill_by_neigh(struct rose_neigh *neigh)
> > >  {
> > >  	struct sock *s;
> > > +	struct hlist_node *tmp;
> > >  
> > >  	spin_lock_bh(&rose_list_lock);
> > > -	sk_for_each(s, &rose_list) {
> > > +	sk_for_each_safe(s, tmp, &rose_list) {
> > >  		struct rose_sock *rose = rose_sk(s);
> > >  
> > > +		sock_hold(s);
> > > +		spin_unlock_bh(&rose_list_lock);
> > > +		lock_sock(s);
> > >  		if (rose->neighbour == neigh) {
> > >  			rose_disconnect(s, ENETUNREACH, ROSE_OUT_OF_ORDER, 0);
> > > -			rose->neighbour->use--;
> > > +			atomic_dec(&rose->neighbour->use);
> > >  			rose->neighbour = NULL;
> > >  		}
> > > +		release_sock(s);
> > > +		sock_put(s);
> > 
> > I'm sorry, this does not work. At this point both 's' and 'tmp' sockets
> > can be freed and reused. Both iterators are not valid anymore when you
> > acquire the 'rose_list_lock' later.
> 
> Thank you for your time and reply! But I think both 's' and 'tmp' can not
> be freed and reused in rose_kill_by_neigh(). Because rose_remove_socket()
> calls sk_del_node_init() which is protected by rose_list_lock to delete the
> socket node from the hlist and if sk->sk_refcnt equals to 1, the socket will
> be deallocated.
> 
> static void rose_remove_socket(struct sock *sk)
> {
> 	spin_lock_bh(&rose_list_lock);
> 	sk_del_node_init(sk);
> 	spin_unlock_bh(&rose_list_lock);
> }
> 
> https://elixir.bootlin.com/linux/v5.19-rc6/source/net/rose/af_rose.c#L152
> 
> Both 's' and 'tmp' in rose_kill_by_neigh() is also protected by rose_list_lock.

The above loop explicitly releases the rose_list_lock at each
iteration. Additionally, the reference count to 's' is released before
re-acquiring such lock. By the time rose_list_lock is re-acquired, some
other process could have removed from the list both 's' and 'tmp' and
even de-allocate them.

Moving the 'sock_put(s);' after re-acquiring the rose_list_lock could
protect from 's' being de-allocated, but can't protect from 'tmp' being
deallocated, neither from 's' and 'tmp' being removed from the list.

The above code is not safe.

/P

Hello,

On Wed, 13 Jul 2022 10:33:54 +0200 Paolo Abeni wrote:

> > > On Mon, 2022-07-11 at 09:31 +0800, Duoming Zhou wrote:
> > > > When the link layer connection is broken, the rose->neighbour is
> > > > set to null. But rose->neighbour could be used by rose_connection()
> > > > and rose_release() later, because there is no synchronization among
> > > > them. As a result, the null-ptr-deref bugs will happen.
> > > > 
> > > > One of the null-ptr-deref bugs is shown below:
> > > > 
> > > >     (thread 1)                  |        (thread 2)
> > > >                                 |  rose_connect
> > > > rose_kill_by_neigh              |    lock_sock(sk)
> > > >   spin_lock_bh(&rose_list_lock) |    if (!rose->neighbour)
> > > >   rose->neighbour = NULL;//(1)  |
> > > >                                 |    rose->neighbour->use++;//(2)
> > > > 
> > > > The rose->neighbour is set to null in position (1) and dereferenced
> > > > in position (2).
> > > > 
> > > > The KASAN report triggered by POC is shown below:
> > > > 
> > > > KASAN: null-ptr-deref in range [0x0000000000000028-0x000000000000002f]
> > > > ...
> > > > RIP: 0010:rose_connect+0x6c2/0xf30
> > > > RSP: 0018:ffff88800ab47d60 EFLAGS: 00000206
> > > > RAX: 0000000000000005 RBX: 000000000000002a RCX: 0000000000000000
> > > > RDX: ffff88800ab38000 RSI: ffff88800ab47e48 RDI: ffff88800ab38309
> > > > RBP: dffffc0000000000 R08: 0000000000000000 R09: ffffed1001567062
> > > > R10: dfffe91001567063 R11: 1ffff11001567061 R12: 1ffff11000d17cd0
> > > > R13: ffff8880068be680 R14: 0000000000000002 R15: 1ffff11000d17cd0
> > > > ...
> > > > Call Trace:
> > > >   <TASK>
> > > >   ? __local_bh_enable_ip+0x54/0x80
> > > >   ? selinux_netlbl_socket_connect+0x26/0x30
> > > >   ? rose_bind+0x5b0/0x5b0
> > > >   __sys_connect+0x216/0x280
> > > >   __x64_sys_connect+0x71/0x80
> > > >   do_syscall_64+0x43/0x90
> > > >   entry_SYSCALL_64_after_hwframe+0x46/0xb0
> > > > 
> > > > This patch adds lock_sock() in rose_kill_by_neigh() in order to
> > > > synchronize with rose_connect() and rose_release(). Then, changing
> > > > type of 'neighbour->use' from unsigned short to atomic_t in order to
> > > > mitigate race conditions caused by holding different socket lock while
> > > > updating 'neighbour->use'.
> > > > 
> > > > Meanwhile, this patch adds sock_hold() protected by rose_list_lock
> > > > that could synchronize with rose_remove_socket() in order to mitigate
> > > > UAF bug caused by lock_sock() we add.
> > > > 
> > > > What's more, there is no need using rose_neigh_list_lock to protect
> > > > rose_kill_by_neigh(). Because we have already used rose_neigh_list_lock
> > > > to protect the state change of rose_neigh in rose_link_failed(), which
> > > > is well synchronized.
> > > > 
> > > > Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
> > > > Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
> > > > ---
> > > > Changes in v6:
> > > >   - Change sk_for_each() to sk_for_each_safe().
> > > >   - Change type of 'neighbour->use' from unsigned short to atomic_t.
> > > > 
> > > >  include/net/rose.h    |  2 +-
> > > >  net/rose/af_rose.c    | 19 +++++++++++++------
> > > >  net/rose/rose_in.c    | 12 ++++++------
> > > >  net/rose/rose_route.c | 24 ++++++++++++------------
> > > >  net/rose/rose_timer.c |  2 +-
> > > >  5 files changed, 33 insertions(+), 26 deletions(-)
> > > > 
> > > > diff --git a/include/net/rose.h b/include/net/rose.h
> > > > index 0f0a4ce0fee..d5ddebc556d 100644
> > > > --- a/include/net/rose.h
> > > > +++ b/include/net/rose.h
> > > > @@ -95,7 +95,7 @@ struct rose_neigh {
> > > >  	ax25_cb			*ax25;
> > > >  	struct net_device		*dev;
> > > >  	unsigned short		count;
> > > > -	unsigned short		use;
> > > > +	atomic_t		use;
> > > >  	unsigned int		number;
> > > >  	char			restarted;
> > > >  	char			dce_mode;
> > > > diff --git a/net/rose/af_rose.c b/net/rose/af_rose.c
> > > > index bf2d986a6bc..54e7b76c4f3 100644
> > > > --- a/net/rose/af_rose.c
> > > > +++ b/net/rose/af_rose.c
> > > > @@ -163,16 +163,23 @@ static void rose_remove_socket(struct sock *sk)
> > > >  void rose_kill_by_neigh(struct rose_neigh *neigh)
> > > >  {
> > > >  	struct sock *s;
> > > > +	struct hlist_node *tmp;
> > > >  
> > > >  	spin_lock_bh(&rose_list_lock);
> > > > -	sk_for_each(s, &rose_list) {
> > > > +	sk_for_each_safe(s, tmp, &rose_list) {
> > > >  		struct rose_sock *rose = rose_sk(s);
> > > >  
> > > > +		sock_hold(s);
> > > > +		spin_unlock_bh(&rose_list_lock);
> > > > +		lock_sock(s);
> > > >  		if (rose->neighbour == neigh) {
> > > >  			rose_disconnect(s, ENETUNREACH, ROSE_OUT_OF_ORDER, 0);
> > > > -			rose->neighbour->use--;
> > > > +			atomic_dec(&rose->neighbour->use);
> > > >  			rose->neighbour = NULL;
> > > >  		}
> > > > +		release_sock(s);
> > > > +		sock_put(s);
> > > 
> > > I'm sorry, this does not work. At this point both 's' and 'tmp' sockets
> > > can be freed and reused. Both iterators are not valid anymore when you
> > > acquire the 'rose_list_lock' later.
> > 
> > Thank you for your time and reply! But I think both 's' and 'tmp' can not
> > be freed and reused in rose_kill_by_neigh(). Because rose_remove_socket()
> > calls sk_del_node_init() which is protected by rose_list_lock to delete the
> > socket node from the hlist and if sk->sk_refcnt equals to 1, the socket will
> > be deallocated.
> > 
> > static void rose_remove_socket(struct sock *sk)
> > {
> > 	spin_lock_bh(&rose_list_lock);
> > 	sk_del_node_init(sk);
> > 	spin_unlock_bh(&rose_list_lock);
> > }
> > 
> > https://elixir.bootlin.com/linux/v5.19-rc6/source/net/rose/af_rose.c#L152
> > 
> > Both 's' and 'tmp' in rose_kill_by_neigh() is also protected by rose_list_lock.
> 
> The above loop explicitly releases the rose_list_lock at each
> iteration. Additionally, the reference count to 's' is released before
> re-acquiring such lock. By the time rose_list_lock is re-acquired, some
> other process could have removed from the list both 's' and 'tmp' and
> even de-allocate them.
> 
> Moving the 'sock_put(s);' after re-acquiring the rose_list_lock could
> protect from 's' being de-allocated, but can't protect from 'tmp' being
> deallocated, neither from 's' and 'tmp' being removed from the list.
> 
> The above code is not safe.

I understand, I will improve the code , thank you!

Best regards,
Duoming Zhou

Hi,

I am an oldtimer FPAC / ROSE user and occasionnally debugger.

Let me take this opportunity to report a major issue present in rose 
module since kernel 5.4.83 (5.5.10).

The bug is an impossibility for a rose application to connect to rose 
socket.

Connect request was working until 5.4.81 kernel.

Here is an illustration using

Linux F6BVP-8 5.4.79-v7+ #1373 SMP Mon Nov 23 13:22:33 GMT 2020 armv7l 
GNU/Linux

and kernel downgraded to kernel 4.4.79 on a RaspbBerry Pi configured 
with ROSE / FPAC node f6bvp-8.

Connect request to co-located node on the same machine does not use 
Ethernet network.

pi@F6BVP-8:~ $ sudo rose_call rose0 f6bvp f6bvp-8 2080175520
F6BVP-8 (Commands = ?) : uilt May 15 2022) for LINUX (help = h)

Or success connecting a remote ROSE / FPAC node via Internet (AX25 over 
UDP frames) :

pi@F6BVP-8:/etc/ax25 $ sudo rose_call rose0 f6bvp f6kkr-8 2080178520
F6KKR-8 (Commands = ?) : uilt Nov 17 2019) for LINUX (help = h)
F6KKR-8 (Commands = ?) :

On listen AX25 tool screen dump (pid=1(X.25) means ROSE protocol

axudp: fm F6BVP-9 to F6KKR-9 ctl I11^ pid=1(X.25) len 60 15:25:04.162488
X.25: LCI 001 : CALL REQUEST - NbAlea: 7801
fm F6BVP-0   @2080,175520
to F6KKR-8   @2080,178520
axudp: fm F6KKR-9 to F6BVP-9 ctl I21^ pid=1(X.25) len 230 15:25:04.177346
X.25: LCI 001 : CALL ACCEPTED
axudp: fm F6KKR-9 to F6BVP-9 ctl I22+ pid=1(X.25) len 179 15:25:04.182222
X.25: LCI 001 : DATA R0 S0  len 176
0000  55 73 65 72 20 63 61 6C 6C 20 3A 20 46 36 42 56  | User call : F6BV
0010  50 2D 30 0D 57 65 6C 63 6F 6D 65 2F 42 69 65 6E  | P-0MWelcome/Bien
0020  76 65 6E 75 65 0D 46 36 4B 4B 52 20 52 61 6D 62  | venueMF6KKR Ramb
0030  6F 75 69 6C 6C 65 74 2C 20 37 38 20 2C 20 46 72  | ouillet, 78 , Fr
0040  61 6E 63 65 0D 35 30 6B 6D 20 53 57 20 6F 66 20  | anceM50km SW of
0050  50 61 72 69 73 0D 0D 46 50 41 43 2D 4E 6F 64 65  | ParisMMFPAC-Node
0060  20 76 20 34 2E 31 2E 31 2D 62 65 74 61 20 28 62  |  v 4.1.1-beta (b
0070  75 69 6C 74 20 4E 6F 76 20 31 37 20 32 30 31 39  | uilt Nov 17 2019
0080  29 20 66 6F 72 20 4C 49 4E 55 58 20 28 68 65 6C  | ) for LINUX (hel
0090  70 20 3D 20 68 29 0D 46 36 4B 4B 52 2D 38 20 28  | p = h)MF6KKR-8 (
00A0  43 6F 6D 6D 61 6E 64 73 20 3D 20 3F 29 20 3A 20  | Commands = ?) :
axudp: fm F6BVP-9 to F6KKR-9 ctl RR3- 15:25:04.184195


Using 5.18.11 kernel with up-to-date netdev ax25 and rose modules.

Linux ubuntu-f6bvp 5.18.11-F6BVP #1 SMP PREEMPT_DYNAMIC Tue Jul 12 
22:13:30 CEST 2022 x86_64 x86_64 x86_64 GNU/Linux

And performing the same connection sequences.

First connect request to co located node:

bernard@ubuntu-f6bvp:/etc/ax25$ sudo rose_call rose0 f6bvp f6bvp-4 
2080175524
Connecting to f6bvp-4 @ 2080175524 ...

infinite wait ...

And trying to connect a local network node does not show any packet 
going out when displaying ax25 activity with "listen" application :

bernard@ubuntu-f6bvp:/etc/ax25$ sudo rose_call rose0 f6bvp f6bvp-8 
2080175520
bernard@ubuntu-f6bvp:/etc/ax25$ 20 ...

No connection... and no outgoing frames on listen screen dump AX25 
application.

Again:

bernard@ubuntu-f6bvp:/etc/ax25$ sudo rose_call rose0 f6bvp f6kkr-8 
2080178520
bernard@ubuntu-f6bvp:/etc/ax25$ 20 ...

No connection.

The issue seems to be in rose socket connect ... I understand that some 
ROSE headers have been changed ... recently (???)

I would be pleased to check any patch to repair this nasty bug and be 
able to let 5.4.79 kernel away with its AX25 bugs ...

Bernard
Hemradio f6bvp / ai7bg
http://f6bvp.org

On Thu, Jul 14, 2022 at 04:11:44PM +0200, Bernard f6bvp wrote:
> Hi,
> 
> I am an oldtimer FPAC / ROSE user and occasionnally debugger.
> 
> Let me take this opportunity to report a major issue present in rose module
> since kernel 5.4.83 (5.5.10).

Do you think you going do a git bisect to find the buggy patch?

https://wiki.gentoo.org/wiki/Kernel_git-bisect

We should probably start a new email thread for that issue.

regards,
dan carpenter