* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
- *
- * $Id: ipoib_main.c 1377 2004-12-23 19:57:12Z roland $
*/
#include "ipoib.h"
#include <linux/init.h>
#include <linux/slab.h>
#include <linux/kernel.h>
+#include <linux/vmalloc.h>
#include <linux/if_arp.h> /* For ARPHRD_xxx */
module_param_named(recv_queue_size, ipoib_recvq_size, int, 0444);
MODULE_PARM_DESC(recv_queue_size, "Number of descriptors in receive queue");
+static int lro;
+module_param(lro, bool, 0444);
+MODULE_PARM_DESC(lro, "Enable LRO (Large Receive Offload)");
+
+static int lro_max_aggr = IPOIB_LRO_MAX_AGGR;
+module_param(lro_max_aggr, int, 0644);
+MODULE_PARM_DESC(lro_max_aggr, "LRO: Max packets to be aggregated "
+ "(default = 64)");
+
#ifdef CONFIG_INFINIBAND_IPOIB_DEBUG
int ipoib_debug_level;
struct ipoib_dev_priv *priv = netdev_priv(dev);
/* dev->mtu > 2K ==> connected mode */
- if (ipoib_cm_admin_enabled(dev) && new_mtu <= IPOIB_CM_MTU) {
+ if (ipoib_cm_admin_enabled(dev)) {
+ if (new_mtu > ipoib_cm_max_mtu(dev))
+ return -EINVAL;
+
if (new_mtu > priv->mcast_mtu)
ipoib_warn(priv, "mtu > %d will cause multicast packet drops.\n",
priv->mcast_mtu);
+
dev->mtu = new_mtu;
return 0;
}
- if (new_mtu > IPOIB_PACKET_SIZE - IPOIB_ENCAP_LEN) {
+ if (new_mtu > IPOIB_UD_MTU(priv->max_ib_mtu))
return -EINVAL;
- }
priv->admin_mtu = new_mtu;
#endif /* CONFIG_INFINIBAND_IPOIB_DEBUG */
+void ipoib_mark_paths_invalid(struct net_device *dev)
+{
+ struct ipoib_dev_priv *priv = netdev_priv(dev);
+ struct ipoib_path *path, *tp;
+
+ spin_lock_irq(&priv->lock);
+
+ list_for_each_entry_safe(path, tp, &priv->path_list, list) {
+ ipoib_dbg(priv, "mark path LID 0x%04x GID " IPOIB_GID_FMT " invalid\n",
+ be16_to_cpu(path->pathrec.dlid),
+ IPOIB_GID_ARG(path->pathrec.dgid));
+ path->valid = 0;
+ }
+
+ spin_unlock_irq(&priv->lock);
+}
+
void ipoib_flush_paths(struct net_device *dev)
{
struct ipoib_dev_priv *priv = netdev_priv(dev);
spin_lock_irq(&priv->tx_lock);
spin_lock(&priv->lock);
- list_splice(&priv->path_list, &remove_list);
- INIT_LIST_HEAD(&priv->path_list);
+ list_splice_init(&priv->path_list, &remove_list);
list_for_each_entry(path, &remove_list, list)
rb_erase(&path->rb_node, &priv->path_tree);
struct net_device *dev = path->dev;
struct ipoib_dev_priv *priv = netdev_priv(dev);
struct ipoib_ah *ah = NULL;
+ struct ipoib_ah *old_ah;
struct ipoib_neigh *neigh, *tn;
struct sk_buff_head skqueue;
struct sk_buff *skb;
spin_lock_irqsave(&priv->lock, flags);
+ old_ah = path->ah;
path->ah = ah;
if (ah) {
__skb_queue_tail(&skqueue, skb);
list_for_each_entry_safe(neigh, tn, &path->neigh_list, list) {
+ if (neigh->ah) {
+ WARN_ON(neigh->ah != old_ah);
+ /*
+ * Dropping the ah reference inside
+ * priv->lock is safe here, because we
+ * will hold one more reference from
+ * the original value of path->ah (ie
+ * old_ah).
+ */
+ ipoib_put_ah(neigh->ah);
+ }
kref_get(&path->ah->ref);
neigh->ah = path->ah;
memcpy(&neigh->dgid.raw, &path->pathrec.dgid.raw,
while ((skb = __skb_dequeue(&neigh->queue)))
__skb_queue_tail(&skqueue, skb);
}
+ path->valid = 1;
}
path->query = NULL;
spin_unlock_irqrestore(&priv->lock, flags);
+ if (old_ah)
+ ipoib_put_ah(old_ah);
+
while ((skb = __skb_dequeue(&skqueue))) {
skb->dev = dev;
if (dev_queue_xmit(skb))
INIT_LIST_HEAD(&path->neigh_list);
memcpy(path->pathrec.dgid.raw, gid, sizeof (union ib_gid));
- path->pathrec.sgid = priv->local_gid;
- path->pathrec.pkey = cpu_to_be16(priv->pkey);
+ path->pathrec.sgid = priv->local_gid;
+ path->pathrec.pkey = cpu_to_be16(priv->pkey);
path->pathrec.numb_path = 1;
path->pathrec.traffic_class = priv->broadcast->mcmember.traffic_class;
spin_lock(&priv->lock);
path = __path_find(dev, phdr->hwaddr + 4);
- if (!path) {
- path = path_rec_create(dev, phdr->hwaddr + 4);
+ if (!path || !path->valid) {
+ if (!path)
+ path = path_rec_create(dev, phdr->hwaddr + 4);
if (path) {
/* put pseudoheader back on for next time */
skb_push(skb, sizeof *phdr);
if (unlikely(!spin_trylock_irqsave(&priv->tx_lock, flags)))
return NETDEV_TX_LOCKED;
- /*
- * Check if our queue is stopped. Since we have the LLTX bit
- * set, we can't rely on netif_stop_queue() preventing our
- * xmit function from being called with a full queue.
- */
- if (unlikely(netif_queue_stopped(dev))) {
- spin_unlock_irqrestore(&priv->tx_lock, flags);
- return NETDEV_TX_BUSY;
- }
-
if (likely(skb->dst && skb->dst->neighbour)) {
if (unlikely(!*to_ipoib_neigh(skb->dst->neighbour))) {
ipoib_path_lookup(skb, dev);
neigh = *to_ipoib_neigh(skb->dst->neighbour);
- if (ipoib_cm_get(neigh)) {
- if (ipoib_cm_up(neigh)) {
- ipoib_cm_send(dev, skb, ipoib_cm_get(neigh));
- goto out;
- }
- } else if (neigh->ah) {
+ if (neigh->ah)
if (unlikely((memcmp(&neigh->dgid.raw,
skb->dst->neighbour->ha + 4,
sizeof(union ib_gid))) ||
goto out;
}
+ if (ipoib_cm_get(neigh)) {
+ if (ipoib_cm_up(neigh)) {
+ ipoib_cm_send(dev, skb, ipoib_cm_get(neigh));
+ goto out;
+ }
+ } else if (neigh->ah) {
ipoib_send(dev, skb, neigh->ah, IPOIB_QPN(skb->dst->neighbour->ha));
goto out;
}
struct ipoib_ah *ah = NULL;
neigh = *to_ipoib_neigh(n);
- if (neigh) {
+ if (neigh)
priv = netdev_priv(neigh->dev);
- ipoib_dbg(priv, "neigh_destructor for bonding device: %s\n",
- n->dev->name);
- } else
+ else
return;
ipoib_dbg(priv,
"neigh_cleanup for %06x " IPOIB_GID_FMT "\n",
goto out;
}
- priv->tx_ring = kzalloc(ipoib_sendq_size * sizeof *priv->tx_ring,
- GFP_KERNEL);
+ priv->tx_ring = vmalloc(ipoib_sendq_size * sizeof *priv->tx_ring);
if (!priv->tx_ring) {
printk(KERN_WARNING "%s: failed to allocate TX ring (%d entries)\n",
ca->name, ipoib_sendq_size);
goto out_rx_ring_cleanup;
}
+ memset(priv->tx_ring, 0, ipoib_sendq_size * sizeof *priv->tx_ring);
/* priv->tx_head, tx_tail & tx_outstanding are already 0 */
return 0;
out_tx_ring_cleanup:
- kfree(priv->tx_ring);
+ vfree(priv->tx_ring);
out_rx_ring_cleanup:
kfree(priv->rx_ring);
ipoib_ib_dev_cleanup(dev);
kfree(priv->rx_ring);
- kfree(priv->tx_ring);
+ vfree(priv->tx_ring);
priv->rx_ring = NULL;
priv->tx_ring = NULL;
.create = ipoib_hard_header,
};
+static int get_skb_hdr(struct sk_buff *skb, void **iphdr,
+ void **tcph, u64 *hdr_flags, void *priv)
+{
+ unsigned int ip_len;
+ struct iphdr *iph;
+
+ if (unlikely(skb->protocol != htons(ETH_P_IP)))
+ return -1;
+
+ /*
+ * In the future we may add an else clause that verifies the
+ * checksum and allows devices which do not calculate checksum
+ * to use LRO.
+ */
+ if (unlikely(skb->ip_summed != CHECKSUM_UNNECESSARY))
+ return -1;
+
+ /* Check for non-TCP packet */
+ skb_reset_network_header(skb);
+ iph = ip_hdr(skb);
+ if (iph->protocol != IPPROTO_TCP)
+ return -1;
+
+ ip_len = ip_hdrlen(skb);
+ skb_set_transport_header(skb, ip_len);
+ *tcph = tcp_hdr(skb);
+
+ /* check if IP header and TCP header are complete */
+ if (ntohs(iph->tot_len) < ip_len + tcp_hdrlen(skb))
+ return -1;
+
+ *hdr_flags = LRO_IPV4 | LRO_TCP;
+ *iphdr = iph;
+
+ return 0;
+}
+
+static void ipoib_lro_setup(struct ipoib_dev_priv *priv)
+{
+ priv->lro.lro_mgr.max_aggr = lro_max_aggr;
+ priv->lro.lro_mgr.max_desc = IPOIB_MAX_LRO_DESCRIPTORS;
+ priv->lro.lro_mgr.lro_arr = priv->lro.lro_desc;
+ priv->lro.lro_mgr.get_skb_header = get_skb_hdr;
+ priv->lro.lro_mgr.features = LRO_F_NAPI;
+ priv->lro.lro_mgr.dev = priv->dev;
+ priv->lro.lro_mgr.ip_summed_aggr = CHECKSUM_UNNECESSARY;
+}
+
static void ipoib_setup(struct net_device *dev)
{
struct ipoib_dev_priv *priv = netdev_priv(dev);
- dev->open = ipoib_open;
- dev->stop = ipoib_stop;
- dev->change_mtu = ipoib_change_mtu;
- dev->hard_start_xmit = ipoib_start_xmit;
- dev->tx_timeout = ipoib_timeout;
- dev->header_ops = &ipoib_header_ops;
- dev->set_multicast_list = ipoib_set_mcast_list;
- dev->neigh_setup = ipoib_neigh_setup_dev;
+ dev->open = ipoib_open;
+ dev->stop = ipoib_stop;
+ dev->change_mtu = ipoib_change_mtu;
+ dev->hard_start_xmit = ipoib_start_xmit;
+ dev->tx_timeout = ipoib_timeout;
+ dev->header_ops = &ipoib_header_ops;
+ dev->set_multicast_list = ipoib_set_mcast_list;
+ dev->neigh_setup = ipoib_neigh_setup_dev;
+
+ ipoib_set_ethtool_ops(dev);
netif_napi_add(dev, &priv->napi, ipoib_poll, 100);
- dev->watchdog_timeo = HZ;
+ dev->watchdog_timeo = HZ;
- dev->flags |= IFF_BROADCAST | IFF_MULTICAST;
+ dev->flags |= IFF_BROADCAST | IFF_MULTICAST;
/*
* We add in INFINIBAND_ALEN to allow for the destination
* address "pseudoheader" for skbs without neighbour struct.
*/
- dev->hard_header_len = IPOIB_ENCAP_LEN + INFINIBAND_ALEN;
- dev->addr_len = INFINIBAND_ALEN;
- dev->type = ARPHRD_INFINIBAND;
- dev->tx_queue_len = ipoib_sendq_size * 2;
- dev->features = NETIF_F_VLAN_CHALLENGED | NETIF_F_LLTX;
-
- /* MTU will be reset when mcast join happens */
- dev->mtu = IPOIB_PACKET_SIZE - IPOIB_ENCAP_LEN;
- priv->mcast_mtu = priv->admin_mtu = dev->mtu;
+ dev->hard_header_len = IPOIB_ENCAP_LEN + INFINIBAND_ALEN;
+ dev->addr_len = INFINIBAND_ALEN;
+ 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);
priv->dev = dev;
+ ipoib_lro_setup(priv);
+
spin_lock_init(&priv->lock);
spin_lock_init(&priv->tx_lock);
- mutex_init(&priv->mcast_mutex);
mutex_init(&priv->vlan_mutex);
INIT_LIST_HEAD(&priv->path_list);
INIT_LIST_HEAD(&priv->multicast_list);
INIT_DELAYED_WORK(&priv->pkey_poll_task, ipoib_pkey_poll);
- INIT_WORK(&priv->pkey_event_task, ipoib_pkey_event);
INIT_DELAYED_WORK(&priv->mcast_task, ipoib_mcast_join_task);
- INIT_WORK(&priv->flush_task, ipoib_ib_dev_flush);
+ INIT_WORK(&priv->flush_light, ipoib_ib_dev_flush_light);
+ INIT_WORK(&priv->flush_normal, ipoib_ib_dev_flush_normal);
+ INIT_WORK(&priv->flush_heavy, ipoib_ib_dev_flush_heavy);
INIT_WORK(&priv->restart_task, ipoib_mcast_restart_task);
INIT_DELAYED_WORK(&priv->ah_reap_task, ipoib_reap_ah);
}
struct ib_device *hca, u8 port)
{
struct ipoib_dev_priv *priv;
+ struct ib_device_attr *device_attr;
+ struct ib_port_attr attr;
int result = -ENOMEM;
priv = ipoib_intf_alloc(format);
SET_NETDEV_DEV(priv->dev, hca->dma_device);
+ if (!ib_query_port(hca, port, &attr))
+ priv->max_ib_mtu = ib_mtu_enum_to_int(attr.max_mtu);
+ else {
+ printk(KERN_WARNING "%s: ib_query_port %d failed\n",
+ hca->name, port);
+ goto device_init_failed;
+ }
+
+ /* MTU will be reset when mcast join happens */
+ priv->dev->mtu = IPOIB_UD_MTU(priv->max_ib_mtu);
+ priv->mcast_mtu = priv->admin_mtu = priv->dev->mtu;
+
result = ib_query_pkey(hca, port, 0, &priv->pkey);
if (result) {
printk(KERN_WARNING "%s: ib_query_pkey port %d failed (ret = %d)\n",
goto device_init_failed;
}
+ device_attr = kmalloc(sizeof *device_attr, GFP_KERNEL);
+ if (!device_attr) {
+ printk(KERN_WARNING "%s: allocation of %zu bytes failed\n",
+ hca->name, sizeof *device_attr);
+ goto device_init_failed;
+ }
+
+ result = ib_query_device(hca, device_attr);
+ if (result) {
+ printk(KERN_WARNING "%s: ib_query_device failed (ret = %d)\n",
+ hca->name, result);
+ kfree(device_attr);
+ goto device_init_failed;
+ }
+ priv->hca_caps = device_attr->device_cap_flags;
+
+ kfree(device_attr);
+
+ if (priv->hca_caps & IB_DEVICE_UD_IP_CSUM) {
+ set_bit(IPOIB_FLAG_CSUM, &priv->flags);
+ priv->dev->features |= NETIF_F_SG | NETIF_F_IP_CSUM;
+ }
+
+ if (lro)
+ priv->dev->features |= NETIF_F_LRO;
+
/*
* Set the full membership bit, so that we join the right
* broadcast group, etc.
} else
memcpy(priv->dev->dev_addr + 4, priv->local_gid.raw, sizeof (union ib_gid));
-
result = ipoib_dev_init(priv->dev, hca, port);
if (result < 0) {
printk(KERN_WARNING "%s: failed to initialize port %d (ret = %d)\n",
goto event_failed;
}
+ if (priv->dev->features & NETIF_F_SG && priv->hca_caps & IB_DEVICE_UD_TSO)
+ priv->dev->features |= NETIF_F_TSO;
+
result = register_netdev(priv->dev);
if (result) {
printk(KERN_WARNING "%s: couldn't register ipoib port %d; error %d\n",
ipoib_sendq_size = roundup_pow_of_two(ipoib_sendq_size);
ipoib_sendq_size = min(ipoib_sendq_size, IPOIB_MAX_QUEUE_SIZE);
- ipoib_sendq_size = max(ipoib_sendq_size, IPOIB_MIN_QUEUE_SIZE);
+ ipoib_sendq_size = max(ipoib_sendq_size, max(2 * MAX_SEND_CQE,
+ IPOIB_MIN_QUEUE_SIZE));
+#ifdef CONFIG_INFINIBAND_IPOIB_CM
+ ipoib_max_conn_qp = min(ipoib_max_conn_qp, IPOIB_CM_MAX_CONN_QP);
+#endif
+
+ /*
+ * When copying small received packets, we only copy from the
+ * linear data part of the SKB, so we rely on this condition.
+ */
+ BUILD_BUG_ON(IPOIB_CM_COPYBREAK > IPOIB_CM_HEAD_SIZE);
ret = ipoib_register_debugfs();
if (ret)