};
/*
- * Device private locking: tx_lock protects members used in TX fast
- * path (and we use LLTX so upper layers don't do extra locking).
- * lock protects everything else. lock nests inside of tx_lock (ie
- * tx_lock must be acquired first if needed).
+ * Device private locking: network stack tx_lock protects members used
+ * in TX fast path, lock protects everything else. lock nests inside
+ * of tx_lock (ie tx_lock must be acquired first if needed).
*/
struct ipoib_dev_priv {
spinlock_t lock;
struct ipoib_rx_buf *rx_ring;
- spinlock_t tx_lock;
struct ipoib_tx_buf *tx_ring;
unsigned tx_head;
unsigned tx_tail;
dev_kfree_skb_any(tx_req->skb);
- spin_lock_irqsave(&priv->tx_lock, flags);
+ netif_tx_lock(dev);
+
++tx->tx_tail;
if (unlikely(--priv->tx_outstanding == ipoib_sendq_size >> 1) &&
netif_queue_stopped(dev) &&
"(status=%d, wrid=%d vend_err %x)\n",
wc->status, wr_id, wc->vendor_err);
- spin_lock(&priv->lock);
+ spin_lock_irqsave(&priv->lock, flags);
neigh = tx->neigh;
if (neigh) {
clear_bit(IPOIB_FLAG_OPER_UP, &tx->flags);
- spin_unlock(&priv->lock);
+ spin_unlock_irqrestore(&priv->lock, flags);
}
- spin_unlock_irqrestore(&priv->tx_lock, flags);
+ netif_tx_unlock(dev);
}
int ipoib_cm_dev_open(struct net_device *dev)
{
struct ipoib_dev_priv *priv = netdev_priv(p->dev);
struct ipoib_cm_tx_buf *tx_req;
- unsigned long flags;
unsigned long begin;
ipoib_dbg(priv, "Destroy active connection 0x%x head 0x%x tail 0x%x\n",
DMA_TO_DEVICE);
dev_kfree_skb_any(tx_req->skb);
++p->tx_tail;
- spin_lock_irqsave(&priv->tx_lock, flags);
+ netif_tx_lock_bh(p->dev);
if (unlikely(--priv->tx_outstanding == ipoib_sendq_size >> 1) &&
netif_queue_stopped(p->dev) &&
test_bit(IPOIB_FLAG_ADMIN_UP, &priv->flags))
netif_wake_queue(p->dev);
- spin_unlock_irqrestore(&priv->tx_lock, flags);
+ netif_tx_unlock_bh(p->dev);
}
if (p->qp)
struct ipoib_dev_priv *priv = netdev_priv(tx->dev);
struct net_device *dev = priv->dev;
struct ipoib_neigh *neigh;
+ unsigned long flags;
int ret;
switch (event->event) {
case IB_CM_REJ_RECEIVED:
case IB_CM_TIMEWAIT_EXIT:
ipoib_dbg(priv, "CM error %d.\n", event->event);
- spin_lock_irq(&priv->tx_lock);
- spin_lock(&priv->lock);
+ netif_tx_lock_bh(dev);
+ spin_lock_irqsave(&priv->lock, flags);
neigh = tx->neigh;
if (neigh) {
queue_work(ipoib_workqueue, &priv->cm.reap_task);
}
- spin_unlock(&priv->lock);
- spin_unlock_irq(&priv->tx_lock);
+ spin_unlock_irqrestore(&priv->lock, flags);
+ netif_tx_unlock_bh(dev);
break;
default:
break;
struct ib_sa_path_rec pathrec;
u32 qpn;
- spin_lock_irqsave(&priv->tx_lock, flags);
- spin_lock(&priv->lock);
+ netif_tx_lock_bh(dev);
+ spin_lock_irqsave(&priv->lock, flags);
+
while (!list_empty(&priv->cm.start_list)) {
p = list_entry(priv->cm.start_list.next, typeof(*p), list);
list_del_init(&p->list);
neigh = p->neigh;
qpn = IPOIB_QPN(neigh->neighbour->ha);
memcpy(&pathrec, &p->path->pathrec, sizeof pathrec);
- spin_unlock(&priv->lock);
- spin_unlock_irqrestore(&priv->tx_lock, flags);
+
+ spin_unlock_irqrestore(&priv->lock, flags);
+ netif_tx_unlock_bh(dev);
+
ret = ipoib_cm_tx_init(p, qpn, &pathrec);
- spin_lock_irqsave(&priv->tx_lock, flags);
- spin_lock(&priv->lock);
+
+ netif_tx_lock_bh(dev);
+ spin_lock_irqsave(&priv->lock, flags);
+
if (ret) {
neigh = p->neigh;
if (neigh) {
kfree(p);
}
}
- spin_unlock(&priv->lock);
- spin_unlock_irqrestore(&priv->tx_lock, flags);
+
+ spin_unlock_irqrestore(&priv->lock, flags);
+ netif_tx_unlock_bh(dev);
}
static void ipoib_cm_tx_reap(struct work_struct *work)
{
struct ipoib_dev_priv *priv = container_of(work, struct ipoib_dev_priv,
cm.reap_task);
+ struct net_device *dev = priv->dev;
struct ipoib_cm_tx *p;
+ unsigned long flags;
+
+ netif_tx_lock_bh(dev);
+ spin_lock_irqsave(&priv->lock, flags);
- spin_lock_irq(&priv->tx_lock);
- spin_lock(&priv->lock);
while (!list_empty(&priv->cm.reap_list)) {
p = list_entry(priv->cm.reap_list.next, typeof(*p), list);
list_del(&p->list);
- spin_unlock(&priv->lock);
- spin_unlock_irq(&priv->tx_lock);
+ spin_unlock_irqrestore(&priv->lock, flags);
+ netif_tx_unlock_bh(dev);
ipoib_cm_tx_destroy(p);
- spin_lock_irq(&priv->tx_lock);
- spin_lock(&priv->lock);
+ netif_tx_lock_bh(dev);
+ spin_lock_irqsave(&priv->lock, flags);
}
- spin_unlock(&priv->lock);
- spin_unlock_irq(&priv->tx_lock);
+
+ spin_unlock_irqrestore(&priv->lock, flags);
+ netif_tx_unlock_bh(dev);
}
static void ipoib_cm_skb_reap(struct work_struct *work)
{
struct ipoib_dev_priv *priv = container_of(work, struct ipoib_dev_priv,
cm.skb_task);
+ struct net_device *dev = priv->dev;
struct sk_buff *skb;
-
+ unsigned long flags;
unsigned mtu = priv->mcast_mtu;
- spin_lock_irq(&priv->tx_lock);
- spin_lock(&priv->lock);
+ netif_tx_lock_bh(dev);
+ spin_lock_irqsave(&priv->lock, flags);
+
while ((skb = skb_dequeue(&priv->cm.skb_queue))) {
- spin_unlock(&priv->lock);
- spin_unlock_irq(&priv->tx_lock);
+ spin_unlock_irqrestore(&priv->lock, flags);
+ netif_tx_unlock_bh(dev);
+
if (skb->protocol == htons(ETH_P_IP))
icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, htonl(mtu));
#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, priv->dev);
#endif
dev_kfree_skb_any(skb);
- spin_lock_irq(&priv->tx_lock);
- spin_lock(&priv->lock);
+
+ netif_tx_lock_bh(dev);
+ spin_lock_irqsave(&priv->lock, flags);
}
- spin_unlock(&priv->lock);
- spin_unlock_irq(&priv->tx_lock);
+
+ spin_unlock_irqrestore(&priv->lock, flags);
+ netif_tx_unlock_bh(dev);
}
void ipoib_cm_skb_too_long(struct net_device *dev, struct sk_buff *skb,
static void drain_tx_cq(struct net_device *dev)
{
struct ipoib_dev_priv *priv = netdev_priv(dev);
- unsigned long flags;
- spin_lock_irqsave(&priv->tx_lock, flags);
+ netif_tx_lock(dev);
while (poll_tx(priv))
; /* nothing */
if (netif_queue_stopped(dev))
mod_timer(&priv->poll_timer, jiffies + 1);
- spin_unlock_irqrestore(&priv->tx_lock, flags);
+ netif_tx_unlock(dev);
}
void ipoib_send_comp_handler(struct ib_cq *cq, void *dev_ptr)
{
- drain_tx_cq((struct net_device *)dev_ptr);
+ struct ipoib_dev_priv *priv = netdev_priv(dev_ptr);
+
+ mod_timer(&priv->poll_timer, jiffies);
}
static inline int post_send(struct ipoib_dev_priv *priv,
struct ipoib_dev_priv *priv = netdev_priv(dev);
struct ipoib_ah *ah, *tah;
LIST_HEAD(remove_list);
+ unsigned long flags;
+
+ netif_tx_lock_bh(dev);
+ spin_lock_irqsave(&priv->lock, flags);
- spin_lock_irq(&priv->tx_lock);
- spin_lock(&priv->lock);
list_for_each_entry_safe(ah, tah, &priv->dead_ahs, list)
if ((int) priv->tx_tail - (int) ah->last_send >= 0) {
list_del(&ah->list);
ib_destroy_ah(ah->ah);
kfree(ah);
}
- spin_unlock(&priv->lock);
- spin_unlock_irq(&priv->tx_lock);
+
+ spin_unlock_irqrestore(&priv->lock, flags);
+ netif_tx_unlock_bh(dev);
}
void ipoib_reap_ah(struct work_struct *work)
{
struct ipoib_dev_priv *priv = netdev_priv(dev);
int i, n;
+
+ /*
+ * We call completion handling routines that expect to be
+ * called from the BH-disabled NAPI poll context, so disable
+ * BHs here too.
+ */
+ local_bh_disable();
+
do {
n = ib_poll_cq(priv->recv_cq, IPOIB_NUM_WC, priv->ibwc);
for (i = 0; i < n; ++i) {
while (poll_tx(priv))
; /* nothing */
+
+ local_bh_enable();
}
int ipoib_ib_dev_stop(struct net_device *dev, int flush)
struct ipoib_dev_priv *priv = netdev_priv(dev);
struct ipoib_path *path, *tp;
LIST_HEAD(remove_list);
+ unsigned long flags;
- spin_lock_irq(&priv->tx_lock);
- spin_lock(&priv->lock);
+ netif_tx_lock_bh(dev);
+ spin_lock_irqsave(&priv->lock, flags);
list_splice_init(&priv->path_list, &remove_list);
list_for_each_entry_safe(path, tp, &remove_list, list) {
if (path->query)
ib_sa_cancel_query(path->query_id, path->query);
- spin_unlock(&priv->lock);
- spin_unlock_irq(&priv->tx_lock);
+ spin_unlock_irqrestore(&priv->lock, flags);
+ netif_tx_unlock_bh(dev);
wait_for_completion(&path->done);
path_free(dev, path);
- spin_lock_irq(&priv->tx_lock);
- spin_lock(&priv->lock);
+ netif_tx_lock_bh(dev);
+ spin_lock_irqsave(&priv->lock, flags);
}
- spin_unlock(&priv->lock);
- spin_unlock_irq(&priv->tx_lock);
+
+ spin_unlock_irqrestore(&priv->lock, flags);
+ netif_tx_unlock_bh(dev);
}
static void path_rec_completion(int status,
struct ipoib_dev_priv *priv = netdev_priv(dev);
struct ipoib_path *path;
struct ipoib_neigh *neigh;
+ unsigned long flags;
neigh = ipoib_neigh_alloc(skb->dst->neighbour, skb->dev);
if (!neigh) {
return;
}
- /*
- * We can only be called from ipoib_start_xmit, so we're
- * inside tx_lock -- no need to save/restore flags.
- */
- spin_lock(&priv->lock);
+ spin_lock_irqsave(&priv->lock, flags);
path = __path_find(dev, skb->dst->neighbour->ha + 4);
if (!path) {
__skb_queue_tail(&neigh->queue, skb);
}
- spin_unlock(&priv->lock);
+ spin_unlock_irqrestore(&priv->lock, flags);
return;
err_list:
++dev->stats.tx_dropped;
dev_kfree_skb_any(skb);
- spin_unlock(&priv->lock);
+ spin_unlock_irqrestore(&priv->lock, flags);
}
static void ipoib_path_lookup(struct sk_buff *skb, struct net_device *dev)
{
struct ipoib_dev_priv *priv = netdev_priv(dev);
struct ipoib_path *path;
+ unsigned long flags;
- /*
- * We can only be called from ipoib_start_xmit, so we're
- * inside tx_lock -- no need to save/restore flags.
- */
- spin_lock(&priv->lock);
+ spin_lock_irqsave(&priv->lock, flags);
path = __path_find(dev, phdr->hwaddr + 4);
if (!path || !path->valid) {
__skb_queue_tail(&path->queue, skb);
if (path_rec_start(dev, path)) {
- spin_unlock(&priv->lock);
+ spin_unlock_irqrestore(&priv->lock, flags);
path_free(dev, path);
return;
} else
dev_kfree_skb_any(skb);
}
- spin_unlock(&priv->lock);
+ spin_unlock_irqrestore(&priv->lock, flags);
return;
}
dev_kfree_skb_any(skb);
}
- spin_unlock(&priv->lock);
+ spin_unlock_irqrestore(&priv->lock, flags);
}
static int ipoib_start_xmit(struct sk_buff *skb, struct net_device *dev)
struct ipoib_neigh *neigh;
unsigned long flags;
- if (unlikely(!spin_trylock_irqsave(&priv->tx_lock, flags)))
- return NETDEV_TX_LOCKED;
-
if (likely(skb->dst && skb->dst->neighbour)) {
if (unlikely(!*to_ipoib_neigh(skb->dst->neighbour))) {
ipoib_path_lookup(skb, dev);
- goto out;
+ return NETDEV_TX_OK;
}
neigh = *to_ipoib_neigh(skb->dst->neighbour);
skb->dst->neighbour->ha + 4,
sizeof(union ib_gid))) ||
(neigh->dev != dev))) {
- spin_lock(&priv->lock);
+ spin_lock_irqsave(&priv->lock, flags);
/*
* It's safe to call ipoib_put_ah() inside
* priv->lock here, because we know that
ipoib_put_ah(neigh->ah);
list_del(&neigh->list);
ipoib_neigh_free(dev, neigh);
- spin_unlock(&priv->lock);
+ spin_unlock_irqrestore(&priv->lock, flags);
ipoib_path_lookup(skb, dev);
- goto out;
+ return NETDEV_TX_OK;
}
if (ipoib_cm_get(neigh)) {
if (ipoib_cm_up(neigh)) {
ipoib_cm_send(dev, skb, ipoib_cm_get(neigh));
- goto out;
+ return NETDEV_TX_OK;
}
} else if (neigh->ah) {
ipoib_send(dev, skb, neigh->ah, IPOIB_QPN(skb->dst->neighbour->ha));
- goto out;
+ return NETDEV_TX_OK;
}
if (skb_queue_len(&neigh->queue) < IPOIB_MAX_PATH_REC_QUEUE) {
- spin_lock(&priv->lock);
+ spin_lock_irqsave(&priv->lock, flags);
__skb_queue_tail(&neigh->queue, skb);
- spin_unlock(&priv->lock);
+ spin_unlock_irqrestore(&priv->lock, flags);
} else {
++dev->stats.tx_dropped;
dev_kfree_skb_any(skb);
IPOIB_GID_RAW_ARG(phdr->hwaddr + 4));
dev_kfree_skb_any(skb);
++dev->stats.tx_dropped;
- goto out;
+ return NETDEV_TX_OK;
}
unicast_arp_send(skb, dev, phdr);
}
}
-out:
- spin_unlock_irqrestore(&priv->tx_lock, flags);
-
return NETDEV_TX_OK;
}
dev->type = ARPHRD_INFINIBAND;
dev->tx_queue_len = ipoib_sendq_size * 2;
dev->features = (NETIF_F_VLAN_CHALLENGED |
- NETIF_F_LLTX |
NETIF_F_HIGHDMA);
memcpy(dev->broadcast, ipv4_bcast_addr, INFINIBAND_ALEN);
ipoib_lro_setup(priv);
spin_lock_init(&priv->lock);
- spin_lock_init(&priv->tx_lock);
mutex_init(&priv->vlan_mutex);
struct net_device *dev = mcast->dev;
struct ipoib_dev_priv *priv = netdev_priv(dev);
struct ipoib_neigh *neigh, *tmp;
- unsigned long flags;
int tx_dropped = 0;
ipoib_dbg_mcast(netdev_priv(dev),
"deleting multicast group " IPOIB_GID_FMT "\n",
IPOIB_GID_ARG(mcast->mcmember.mgid));
- spin_lock_irqsave(&priv->lock, flags);
+ spin_lock_irq(&priv->lock);
list_for_each_entry_safe(neigh, tmp, &mcast->neigh_list, list) {
/*
ipoib_neigh_free(dev, neigh);
}
- spin_unlock_irqrestore(&priv->lock, flags);
+ spin_unlock_irq(&priv->lock);
if (mcast->ah)
ipoib_put_ah(mcast->ah);
dev_kfree_skb_any(skb_dequeue(&mcast->pkt_queue));
}
- spin_lock_irqsave(&priv->tx_lock, flags);
+ netif_tx_lock_bh(dev);
dev->stats.tx_dropped += tx_dropped;
- spin_unlock_irqrestore(&priv->tx_lock, flags);
+ netif_tx_unlock_bh(dev);
kfree(mcast);
}
}
/* actually send any queued packets */
- spin_lock_irq(&priv->tx_lock);
+ netif_tx_lock_bh(dev);
while (!skb_queue_empty(&mcast->pkt_queue)) {
struct sk_buff *skb = skb_dequeue(&mcast->pkt_queue);
- spin_unlock_irq(&priv->tx_lock);
+ netif_tx_unlock_bh(dev);
skb->dev = dev;
if (dev_queue_xmit(skb))
ipoib_warn(priv, "dev_queue_xmit failed to requeue packet\n");
- spin_lock_irq(&priv->tx_lock);
+ netif_tx_lock_bh(dev);
}
- spin_unlock_irq(&priv->tx_lock);
+ netif_tx_unlock_bh(dev);
return 0;
}
{
struct ipoib_mcast *mcast = multicast->context;
struct net_device *dev = mcast->dev;
- struct ipoib_dev_priv *priv = netdev_priv(dev);
/* We trap for port events ourselves. */
if (status == -ENETRESET)
IPOIB_GID_ARG(mcast->mcmember.mgid), status);
/* Flush out any queued packets */
- spin_lock_irq(&priv->tx_lock);
+ netif_tx_lock_bh(dev);
while (!skb_queue_empty(&mcast->pkt_queue)) {
++dev->stats.tx_dropped;
dev_kfree_skb_any(skb_dequeue(&mcast->pkt_queue));
}
- spin_unlock_irq(&priv->tx_lock);
+ netif_tx_unlock_bh(dev);
/* Clear the busy flag so we try again */
status = test_and_clear_bit(IPOIB_MCAST_FLAG_BUSY,
{
struct ipoib_dev_priv *priv = netdev_priv(dev);
struct ipoib_mcast *mcast;
+ unsigned long flags;
- /*
- * We can only be called from ipoib_start_xmit, so we're
- * inside tx_lock -- no need to save/restore flags.
- */
- spin_lock(&priv->lock);
+ spin_lock_irqsave(&priv->lock, flags);
if (!test_bit(IPOIB_FLAG_OPER_UP, &priv->flags) ||
!priv->broadcast ||
}
unlock:
- spin_unlock(&priv->lock);
+ spin_unlock_irqrestore(&priv->lock, flags);
}
void ipoib_mcast_dev_flush(struct net_device *dev)