can: remove CAN filters independent from namespace

When the devices namespace is changed the socket namespace and the device namespace might differ. The net_eq(dev_net(dev), sock_net(sk)) check in the CAN protocols netdev notifiers therefore led to skipping the required removal of the CAN filters from the (namespace changed) CAN devices.

This patch removes the namespace equality check in the netdev notifiers for BCM, ISOTP and RAW sockets. Since the struct net_device pointer is globally unique, the notifier should always process the unregister event and remove the CAN filters if it matches the original socket's bound device pointer.

In bcm.c netdevice comparisons were performed by checking the interface index (bo->ifindex and op->ifindex) which is not namespace-safe either. Introduce tracked netdevice pointers (bo->dev and op->tx_dev) for these referenced devices to enable namespace-save device comparisons. Additional put all bo->dev accesses in bcm_notify() under lock_sock().

In isotp.c the two missing can_rx_unregister() calling sites are converted to use dev_net(dev) instead of sock_net(sk) to get the correct namespace.

Fixes 8e8cda6d737d ("can: initial support for network namespaces")

Reported-by Norbert Szetei <[email protected]>

Link https://lore.kernel.org/linux-can/[email protected]/ open here

Cc [email protected]

Signed-off-by Oliver Hartkopp <[email protected]>

---
changes in v2:
- re-introduce the net_eq() check for ANYDEV case handling of op->ifdetected
- drop bo->dev reference tracking when procfs initialization fails
- add error handling for dev_get_by_index() of tx_dev in bcm_tx_setup()
- put all bo->dev accesses in bcm_notify() under lock_sock()
---

net/can/bcm.c   | 100 ++++++++++++++++++++++++++++++++++++------------
net/can/isotp.c |   7 +---
net/can/raw.c   |   3 --
3 files changed, 78 insertions(+), 32 deletions(-)

net/can/bcm.c +2 -0

index 3d637a1e0ac1..cd3522ec32c0 100644
@@ -128,18 +128,22 @@ struct bcm_op {
 struct canfd_frame sframe;
 struct canfd_frame last_sframe;
 struct sock *sk;
 struct net_device *rx_reg_dev;
 netdevice_tracker rx_reg_dev_tracker;
+ struct net_device *tx_dev;
+ netdevice_tracker tx_dev_tracker;

spinlock_t bcm_tx_lock; /* protect tx data and timer updates */ spinlock_t bcm_rx_update_lock; /* protect filter/timer data updates */ };

struct bcm_sock {
struct sock sk;
int bound;
int ifindex;
+ struct net_device *dev;
+ netdevice_tracker dev_tracker;
struct list_head notifier;
struct list_head rx_ops;
struct list_head tx_ops;
unsigned long dropped_usr_msgs;
struct proc_dir_entry *bcm_proc_read;

net/can/bcm.c +36 -2

@@ -933,10 +937,13 @@ static void bcm_free_op_work(struct work_struct *work)
 kfree(op->frames);
 
 if ((op->last_frames) && (op->last_frames != &op->last_sframe))
 kfree(op->last_frames);
 
+ if (op->tx_dev)
+ netdev_put(op->tx_dev, &op->tx_dev_tracker);
+
 /* the last possible access to op->timer/op->thrtimer has now
 * happened above via hrtimer_cancel() - op->sk is no longer
 * needed by any pending timer callback, so drop our reference
 */
 sock_put(op->sk);

@@ -1072,10 +1079,11 @@ static int bcm_tx_setup(struct bcm_msg_head *msg_head, struct msghdr *msg,
 int ifindex, struct sock *sk)
 {
 struct bcm_sock *bo = bcm_sk(sk);
 struct bcm_op *op;
 struct canfd_frame *cf;
+ struct net_device *tx_dev;
 bool add_op_to_list = false;
 unsigned int i;
 int err;
 
 /* we need a real device to send frames */

@@ -1103,10 +1111,26 @@ static int bcm_tx_setup(struct bcm_msg_head *msg_head, struct msghdr *msg,
 * therefore (complexity / locking) it is not supported.
 */
 if (msg_head->nframes > op->nframes)
 return -E2BIG;
 
+ /* Re-resolve and re-hold the target device if a concurrent
+ * NETDEV_UNREGISTER already cleared it (see bcm_notify()).
+ * op->ifindex and sock_net(sk) is unchanged.
+ */
+ if (!op->tx_dev) {
+ tx_dev = dev_get_by_index(sock_net(sk), ifindex);
+ if (tx_dev) {
+ op->tx_dev = tx_dev;
+ netdev_hold(tx_dev, &op->tx_dev_tracker,
+    GFP_KERNEL);
+ dev_put(tx_dev);
+ } else {
+ return -ENODEV;
+ }
+ }
+
 /* get new CAN frames content into a staging buffer before
 * locking: validate and normalize the frames there so that
 * bcm_can_tx() / bcm_tx_timeout_handler() never observe a
 * partially updated or unvalidated frame in op->frames
 */

@@ -1169,10 +1193,22 @@ static int bcm_tx_setup(struct bcm_msg_head *msg_head, struct msghdr *msg,
 
 op = kzalloc(OPSIZ, GFP_KERNEL);
 if (!op)
 return -ENOMEM;
 
+ tx_dev = dev_get_by_index(sock_net(sk), ifindex);
+ if (tx_dev) {
+ op->tx_dev = tx_dev;
+ netdev_hold(tx_dev, &op->tx_dev_tracker, GFP_KERNEL);
+ dev_put(tx_dev);
+ } else {
+ /* prepare op->frames for goto free_op */
+ op->frames = &op->sframe;
+ err = -ENODEV;
+ goto free_op;
+ }
+
 spin_lock_init(&op->bcm_tx_lock);
 op->can_id = msg_head->can_id;
 op->cfsiz = CFSIZ(msg_head->flags);
 op->flags = msg_head->flags;
 op->nframes = msg_head->nframes;

@@ -1184,12 +1220,14 @@ static int bcm_tx_setup(struct bcm_msg_head *msg_head, struct msghdr *msg,
 if (msg_head->nframes > 1) {
 op->frames = kmalloc_array(msg_head->nframes,
  op->cfsiz,
  GFP_KERNEL);
 if (!op->frames) {
- kfree(op);
- return -ENOMEM;
+ /* prepare op->frames for goto free_op */
+ op->frames = &op->sframe;
+ err = -ENOMEM;
+ goto free_op;
 }
 } else
 op->frames = &op->sframe;
 
 for (i = 0; i < msg_head->nframes; i++) {

@@ -1267,10 +1305,13 @@ static int bcm_tx_setup(struct bcm_msg_head *msg_head, struct msghdr *msg,
 bcm_tx_start_timer(op);
 
 return msg_head->nframes * op->cfsiz + MHSIZ;
 

free_op:
+ if (op->tx_dev)
+ netdev_put(op->tx_dev, &op->tx_dev_tracker);
+
if (op->frames != &op->sframe)
kfree(op->frames);
kfree(op);
return err;
}

net/can/bcm.c +2 -3

@@ -1790,46 +1831,47 @@ static int bcm_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
 static void bcm_notify(struct bcm_sock *bo, unsigned long msg,
      struct net_device *dev)
 {
 struct sock *sk = &bo->sk;
 struct bcm_op *op;
- int notify_enodev = 0;
+ int sk_err = 0;
 
- if (!net_eq(dev_net(dev), sock_net(sk)))
- return;
+ lock_sock(sk);
 
 switch (msg) {
 

case NETDEV_UNREGISTER:
- lock_sock(sk);

/* rx_ops: remove device specific receive entries */ list_for_each_entry(op, &bo->rx_ops, list) { if (op->rx_reg_dev == dev)
bcm_rx_unreg(dev, op);

/* release an ANYDEV op's claim (see bcm_rx_handler())
* on this now confirmed-gone interface.
*/
- if (!op->ifindex) {
+ if (!op->ifindex && net_eq(dev_net(dev), sock_net(sk))) { spin_lock_bh(&op->bcm_rx_update_lock);
if (op->if_detected == dev->ifindex) op->if_detected = 0; spin_unlock_bh(&op->bcm_rx_update_lock);
}
}

/* tx_ops: stop device specific cyclic transmissions on the
- * vanishing ifindex. Cancelling the timer is enough to stop
+ * vanishing device. Cancelling the timer is enough to stop
* cyclic bcm_can_tx() calls as there is no re-arming. */
list_for_each_entry(op, &bo->tx_ops, list)
- if (op->ifindex == dev->ifindex)
+ if (op->tx_dev == dev) { hrtimer_cancel(&op->timer);
+ netdev_put(op->tx_dev, &op->tx_dev_tracker);
+ op->tx_dev = NULL;
+ }

/* remove device reference, if this is our bound device */
- if (bo->bound && bo->ifindex == dev->ifindex) {
+ if (bo->bound && bo->dev == dev) {
#if IS_ENABLED(CONFIG_PROC_FS)
if (sock_net(sk)->can.bcmproc_dir && bo->bcm_proc_read) { remove_proc_entry(bo->procname, sock_net(sk)->can.bcmproc_dir); bo->bcm_proc_read = NULL;
}

net/can/bcm.c +3 -9

@@ -1839,28 +1881,27 @@ static void bcm_notify(struct bcm_sock *bo, unsigned long msg,
 * accessed under lock_sock() so it needs no
 * annotation.
 */
 WRITE_ONCE(bo->bound, 0);
 bo->ifindex = 0;
- notify_enodev = 1;
- }
-
- release_sock(sk);
-
- if (notify_enodev) {
- sk->sk_err = ENODEV;
- if (!sock_flag(sk, SOCK_DEAD))
- sk_error_report(sk);
+ netdev_put(bo->dev, &bo->dev_tracker);
+ bo->dev = NULL;
+ sk_err = ENODEV;
 }
 break;
 

case NETDEV_DOWN:
- if (bo->bound && bo->ifindex == dev->ifindex) {
- sk->sk_err = ENETDOWN;
- if (!sock_flag(sk, SOCK_DEAD))
- sk_error_report(sk);
- }
+ if (bo->bound && bo->dev == dev)
+ sk_err = ENETDOWN;
+ }

net/can/bcm.c +7 -0

+
+ release_sock(sk);
+
+ if (sk_err) {
+ sk->sk_err = sk_err;
+ if (!sock_flag(sk, SOCK_DEAD))
+ sk_error_report(sk);

}
}

static int bcm_notifier(struct notifier_block *nb, unsigned long msg, void *ptr)

net/can/bcm.c +11 -0

@@ -1982,10 +2023,14 @@ static int bcm_release(struct socket *sock)
 
 /* remove device reference */
 if (bo->bound) {
 WRITE_ONCE(bo->bound, 0);
 bo->ifindex = 0;
+ if (bo->dev) {
+ netdev_put(bo->dev, &bo->dev_tracker);
+ bo->dev = NULL;
+ }
 }
 
 sock_orphan(sk);
 sock->sk = NULL;
 

@@ -2029,25 +2074,32 @@ static int bcm_connect(struct socket *sock, struct sockaddr_unsized *uaddr, int
 ret = -ENODEV;
 goto fail;
 }
 
 bo->ifindex = dev->ifindex;
+ bo->dev = dev;
+ netdev_hold(dev, &bo->dev_tracker, GFP_KERNEL);
 dev_put(dev);
 
 } else {
 /* no interface reference for ifindex = 0 ('any' CAN device) */
 bo->ifindex = 0;
+ bo->dev = NULL;
 }
 
 #if IS_ENABLED(CONFIG_PROC_FS)
 if (net->can.bcmproc_dir) {
 /* unique socket address as filename */
 sprintf(bo->procname, "%llu", sock_i_ino(sk));
 bo->bcm_proc_read = proc_create_net_single(bo->procname, 0644,
    net->can.bcmproc_dir,
    bcm_proc_show, sk);
 if (!bo->bcm_proc_read) {
+ if (bo->dev) {
+ netdev_put(bo->dev, &bo->dev_tracker);
+ bo->dev = NULL;
+ }
 ret = -ENOMEM;
 goto fail;
 }
 }
 #endif /* CONFIG_PROC_FS */

net/can/isotp.c +2 -5

index 155530aedce2..0835a4758a72 100644
@@ -1490,15 +1490,15 @@ static int isotp_release(struct socket *sock)
 /* remove current filters & unregister
 * tracked reference so->dev is taken at bind() time with rtnl_lock
 */
 if (so->bound && so->dev) {
 if (isotp_register_rxid(so))
- can_rx_unregister(net, so->dev, so->rxid,
+ can_rx_unregister(dev_net(so->dev), so->dev, so->rxid,
 SINGLE_MASK(so->rxid),
 isotp_rcv, sk);
 
- can_rx_unregister(net, so->dev, so->txid,
+ can_rx_unregister(dev_net(so->dev), so->dev, so->txid,
 SINGLE_MASK(so->txid),
 isotp_rcv_echo, sk);
 netdev_put(so->dev, &so->dev_tracker);
 }
 

@@ -1846,13 +1846,10 @@ static int isotp_getsockopt(struct socket *sock, int level, int optname,
 static void isotp_notify(struct isotp_sock *so, unsigned long msg,
 struct net_device *dev)
 {
 struct sock *sk = &so->sk;
 
- if (!net_eq(dev_net(dev), sock_net(sk)))
- return;
-
 if (so->dev != dev)
 return;
 
 switch (msg) {

case NETDEV_UNREGISTER:

net/can/raw.c +0 -3

index 82d9c0499c95..c5596fc9aac5 100644
@@ -300,13 +300,10 @@ static int raw_enable_allfilters(struct net *net, struct net_device *dev,
 static void raw_notify(struct raw_sock *ro, unsigned long msg,
      struct net_device *dev)
 {
 struct sock *sk = &ro->sk;
 
- if (!net_eq(dev_net(dev), sock_net(sk)))
- return;
-
 if (ro->dev != dev)
 return;
 
 switch (msg) {

case NETDEV_UNREGISTER:

Signature · 1 line

2.53.0

Hi Oliver,

many thanks for v2. I ran it on 7.3-rc2 with vcan and my reproducer no longer triggers. The ifindex case is gone too, from another namespace I can no longer touch a bound BCM socket.

Tested-by Norbert Szetei <[email protected] <mailto:[email protected]>>

Thanks,
Norbert

Quoted 6 lines · Oliver Hartkopp

When the devices namespace is changed the socket namespace and the device namespace might differ. The net_eq(dev_net(dev), sock_net(sk)) check in the CAN protocols netdev notifiers therefore led to skipping the required removal of the CAN filters from the (namespace changed) CAN devices.

Sorry, my mail client turned the address into a link and mangled the tag. It should read:

Tested-by Norbert Szetei <[email protected]>

N.

Quoted 19 lines · Norbert Szetei

Hi Oliver,

many thanks for v2. I ran it on 7.3-rc2 with vcan and my reproducer no longer triggers. The ifindex case is gone too, from another namespace I can no longer touch a bound BCM socket.

Tested-by Norbert Szetei <[email protected] <mailto:[email protected]>>

Thanks,
Norbert

Quoted 6 lines · Oliver Hartkopp

When the devices namespace is changed the socket namespace and the device namespace might differ. The net_eq(dev_net(dev), sock_net(sk)) check in the CAN protocols netdev notifiers therefore led to skipping the required removal of the CAN filters from the (namespace changed) CAN devices.