macvtap: add TUNSETQUEUE ioctl
This patch adds TUNSETQUEUE ioctl to let userspace can temporarily disable or
enable a queue of macvtap. This is used to be compatible at API layer of tuntap
to simplify the userspace to manage the queues. This is done through introducing
a linked list to track all taps while using vlan->taps array to only track
active taps.
Signed-off-by: Jason Wang <jasowang@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
diff --git a/drivers/net/macvtap.c b/drivers/net/macvtap.c
index 5ccba99b..d2d1d55 100644
--- a/drivers/net/macvtap.c
+++ b/drivers/net/macvtap.c
@@ -45,6 +45,8 @@
struct file *file;
unsigned int flags;
u16 queue_index;
+ bool enabled;
+ struct list_head next;
};
static struct proto macvtap_proto = {
@@ -85,14 +87,36 @@
*/
static DEFINE_SPINLOCK(macvtap_lock);
-static int macvtap_set_queue(struct net_device *dev, struct file *file,
+static int macvtap_enable_queue(struct net_device *dev, struct file *file,
struct macvtap_queue *q)
{
struct macvlan_dev *vlan = netdev_priv(dev);
+ int err = -EINVAL;
+
+ spin_lock(&macvtap_lock);
+
+ if (q->enabled)
+ goto out;
+
+ err = 0;
+ rcu_assign_pointer(vlan->taps[vlan->numvtaps], q);
+ q->queue_index = vlan->numvtaps;
+ q->enabled = true;
+
+ vlan->numvtaps++;
+out:
+ spin_unlock(&macvtap_lock);
+ return err;
+}
+
+static int macvtap_set_queue(struct net_device *dev, struct file *file,
+ struct macvtap_queue *q)
+{
+ struct macvlan_dev *vlan = netdev_priv(dev);
int err = -EBUSY;
spin_lock(&macvtap_lock);
- if (vlan->numvtaps == MAX_MACVTAP_QUEUES)
+ if (vlan->numqueues == MAX_MACVTAP_QUEUES)
goto out;
err = 0;
@@ -102,15 +126,57 @@
q->file = file;
q->queue_index = vlan->numvtaps;
+ q->enabled = true;
file->private_data = q;
+ list_add_tail(&q->next, &vlan->queue_list);
vlan->numvtaps++;
+ vlan->numqueues++;
out:
spin_unlock(&macvtap_lock);
return err;
}
+static int __macvtap_disable_queue(struct macvtap_queue *q)
+{
+ struct macvlan_dev *vlan;
+ struct macvtap_queue *nq;
+
+ vlan = rcu_dereference_protected(q->vlan,
+ lockdep_is_held(&macvtap_lock));
+
+ if (!q->enabled)
+ return -EINVAL;
+
+ if (vlan) {
+ int index = q->queue_index;
+ BUG_ON(index >= vlan->numvtaps);
+ nq = rcu_dereference_protected(vlan->taps[vlan->numvtaps - 1],
+ lockdep_is_held(&macvtap_lock));
+ nq->queue_index = index;
+
+ rcu_assign_pointer(vlan->taps[index], nq);
+ RCU_INIT_POINTER(vlan->taps[vlan->numvtaps - 1], NULL);
+ q->enabled = false;
+
+ vlan->numvtaps--;
+ }
+
+ return 0;
+}
+
+static int macvtap_disable_queue(struct macvtap_queue *q)
+{
+ int err;
+
+ spin_lock(&macvtap_lock);
+ err = __macvtap_disable_queue(q);
+ spin_unlock(&macvtap_lock);
+
+ return err;
+}
+
/*
* The file owning the queue got closed, give up both
* the reference that the files holds as well as the
@@ -121,25 +187,19 @@
*/
static void macvtap_put_queue(struct macvtap_queue *q)
{
- struct macvtap_queue *nq;
struct macvlan_dev *vlan;
spin_lock(&macvtap_lock);
vlan = rcu_dereference_protected(q->vlan,
lockdep_is_held(&macvtap_lock));
if (vlan) {
- int index = q->queue_index;
- BUG_ON(index >= vlan->numvtaps);
+ if (q->enabled)
+ BUG_ON(__macvtap_disable_queue(q));
- nq = rcu_dereference_protected(vlan->taps[vlan->numvtaps - 1],
- lockdep_is_held(&macvtap_lock));
- rcu_assign_pointer(vlan->taps[index], nq);
- nq->queue_index = index;
-
- RCU_INIT_POINTER(vlan->taps[vlan->numvtaps - 1], NULL);
+ vlan->numqueues--;
RCU_INIT_POINTER(q->vlan, NULL);
sock_put(&q->sk);
- --vlan->numvtaps;
+ list_del_init(&q->next);
}
spin_unlock(&macvtap_lock);
@@ -160,6 +220,11 @@
{
struct macvlan_dev *vlan = netdev_priv(dev);
struct macvtap_queue *tap = NULL;
+ /* Access to taps array is protected by rcu, but access to numvtaps
+ * isn't. Below we use it to lookup a queue, but treat it as a hint
+ * and validate that the result isn't NULL - in case we are
+ * racing against queue removal.
+ */
int numvtaps = ACCESS_ONCE(vlan->numvtaps);
__u32 rxq;
@@ -196,18 +261,22 @@
static void macvtap_del_queues(struct net_device *dev)
{
struct macvlan_dev *vlan = netdev_priv(dev);
- struct macvtap_queue *q, *qlist[MAX_MACVTAP_QUEUES];
+ struct macvtap_queue *q, *tmp, *qlist[MAX_MACVTAP_QUEUES];
int i, j = 0;
spin_lock(&macvtap_lock);
- for (i = 0; i < vlan->numvtaps; i++) {
- q = rcu_dereference_protected(vlan->taps[i],
- lockdep_is_held(&macvtap_lock));
- BUG_ON(q == NULL);
+ list_for_each_entry_safe(q, tmp, &vlan->queue_list, next) {
+ list_del_init(&q->next);
qlist[j++] = q;
- RCU_INIT_POINTER(vlan->taps[i], NULL);
RCU_INIT_POINTER(q->vlan, NULL);
+ if (q->enabled)
+ vlan->numvtaps--;
+ vlan->numqueues--;
}
+ for (i = 0; i < vlan->numvtaps; i++)
+ RCU_INIT_POINTER(vlan->taps[i], NULL);
+ BUG_ON(vlan->numvtaps);
+ BUG_ON(vlan->numqueues);
/* guarantee that any future macvtap_set_queue will fail */
vlan->numvtaps = MAX_MACVTAP_QUEUES;
spin_unlock(&macvtap_lock);
@@ -298,6 +367,9 @@
struct nlattr *tb[],
struct nlattr *data[])
{
+ struct macvlan_dev *vlan = netdev_priv(dev);
+ INIT_LIST_HEAD(&vlan->queue_list);
+
/* Don't put anything that may fail after macvlan_common_newlink
* because we can't undo what it does.
*/
@@ -887,6 +959,25 @@
dev_put(vlan->dev);
}
+static int macvtap_ioctl_set_queue(struct file *file, unsigned int flags)
+{
+ struct macvtap_queue *q = file->private_data;
+ struct macvlan_dev *vlan;
+ int ret;
+
+ vlan = macvtap_get_vlan(q);
+ if (!vlan)
+ return -EINVAL;
+
+ if (flags & IFF_ATTACH_QUEUE)
+ ret = macvtap_enable_queue(vlan->dev, file, q);
+ else if (flags & IFF_DETACH_QUEUE)
+ ret = macvtap_disable_queue(q);
+
+ macvtap_put_vlan(vlan);
+ return ret;
+}
+
/*
* provide compatibility with generic tun/tap interface
*/
@@ -910,7 +1001,8 @@
return -EFAULT;
ret = 0;
- if ((u & ~IFF_VNET_HDR) != (IFF_NO_PI | IFF_TAP))
+ if ((u & ~(IFF_VNET_HDR | IFF_MULTI_QUEUE)) !=
+ (IFF_NO_PI | IFF_TAP))
ret = -EINVAL;
else
q->flags = u;
@@ -929,6 +1021,11 @@
macvtap_put_vlan(vlan);
return ret;
+ case TUNSETQUEUE:
+ if (get_user(u, &ifr->ifr_flags))
+ return -EFAULT;
+ return macvtap_ioctl_set_queue(file, u);
+
case TUNGETFEATURES:
if (put_user(IFF_TAP | IFF_NO_PI | IFF_VNET_HDR, up))
return -EFAULT;