[ovs.git] / datapath / vport-netdev.c

/*
 * Copyright (c) 2007-2012 Nicira, Inc.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of version 2 of the GNU General Public
 * License as published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful, but
 * WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
 * General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
 * 02110-1301, USA
 */

#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt

#include <linux/if_arp.h>
#include <linux/if_bridge.h>
#include <linux/if_vlan.h>
#include <linux/kernel.h>
#include <linux/llc.h>
#include <linux/rtnetlink.h>
#include <linux/skbuff.h>
#include <linux/openvswitch.h>

#include <net/llc.h>

#include "checksum.h"
#include "datapath.h"
#include "vlan.h"
#include "vport-internal_dev.h"
#include "vport-netdev.h"

#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,37) && \
	!defined(HAVE_VLAN_BUG_WORKAROUND)
#include <linux/module.h>

static int vlan_tso __read_mostly;
module_param(vlan_tso, int, 0644);
MODULE_PARM_DESC(vlan_tso, "Enable TSO for VLAN packets");
#else
#define vlan_tso true
#endif

static void netdev_port_receive(struct vport *vport, struct sk_buff *skb);

#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,39)
/* Called with rcu_read_lock and bottom-halves disabled. */
static rx_handler_result_t netdev_frame_hook(struct sk_buff **pskb)
{
	struct sk_buff *skb = *pskb;
	struct vport *vport;

	if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
		return RX_HANDLER_PASS;

	vport = ovs_netdev_get_vport(skb->dev);

	netdev_port_receive(vport, skb);

	return RX_HANDLER_CONSUMED;
}
#elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36) || \
      defined HAVE_RHEL_OVS_HOOK
/* Called with rcu_read_lock and bottom-halves disabled. */
static struct sk_buff *netdev_frame_hook(struct sk_buff *skb)
{
	struct vport *vport;

	if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
		return skb;

	vport = ovs_netdev_get_vport(skb->dev);

	netdev_port_receive(vport, skb);

	return NULL;
}
#elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
/*
 * Used as br_handle_frame_hook.  (Cannot run bridge at the same time, even on
 * different set of devices!)
 */
/* Called with rcu_read_lock and bottom-halves disabled. */
static struct sk_buff *netdev_frame_hook(struct net_bridge_port *p,
					 struct sk_buff *skb)
{
	netdev_port_receive((struct vport *)p, skb);
	return NULL;
}
#elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0)
/*
 * Used as br_handle_frame_hook.  (Cannot run bridge at the same time, even on
 * different set of devices!)
 */
/* Called with rcu_read_lock and bottom-halves disabled. */
static int netdev_frame_hook(struct net_bridge_port *p, struct sk_buff **pskb)
{
	netdev_port_receive((struct vport *)p, *pskb);
	return 1;
}
#else
#error
#endif

#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36) || \
    defined HAVE_RHEL_OVS_HOOK
static int netdev_init(void) { return 0; }
static void netdev_exit(void) { }
#else
static int port_count;

static void netdev_init(void)
{
	port_count++;
	if (port_count > 1)
		return;

	/* Hook into callback used by the bridge to intercept packets.
	 * Parasites we are. */
	br_handle_frame_hook = netdev_frame_hook;

	return;
}

static void netdev_exit(void)
{
	port_count--;
	if (port_count > 0)
		return;

	br_handle_frame_hook = NULL;
}
#endif

static struct net_device *get_dpdev(struct datapath *dp)
{
	struct vport *local;

	local = ovs_vport_ovsl(dp, OVSP_LOCAL);
	BUG_ON(!local);
	return netdev_vport_priv(local)->dev;
}

static struct vport *netdev_create(const struct vport_parms *parms)
{
	struct vport *vport;
	struct netdev_vport *netdev_vport;
	int err;

	vport = ovs_vport_alloc(sizeof(struct netdev_vport),
				&ovs_netdev_vport_ops, parms);
	if (IS_ERR(vport)) {
		err = PTR_ERR(vport);
		goto error;
	}

	netdev_vport = netdev_vport_priv(vport);

	netdev_vport->dev = dev_get_by_name(ovs_dp_get_net(vport->dp), parms->name);
	if (!netdev_vport->dev) {
		err = -ENODEV;
		goto error_free_vport;
	}

	if (netdev_vport->dev->flags & IFF_LOOPBACK ||
	    netdev_vport->dev->type != ARPHRD_ETHER ||
	    ovs_is_internal_dev(netdev_vport->dev)) {
		err = -EINVAL;
		goto error_put;
	}

	rtnl_lock();
	err = netdev_master_upper_dev_link(netdev_vport->dev,
					   get_dpdev(vport->dp));
	if (err)
		goto error_unlock;

	err = netdev_rx_handler_register(netdev_vport->dev, netdev_frame_hook,
					 vport);
	if (err)
		goto error_master_upper_dev_unlink;

	dev_set_promiscuity(netdev_vport->dev, 1);
#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,24)
	dev_disable_lro(netdev_vport->dev);
#endif
	netdev_vport->dev->priv_flags |= IFF_OVS_DATAPATH;
	rtnl_unlock();

	netdev_init();
	return vport;

error_master_upper_dev_unlink:
	netdev_upper_dev_unlink(netdev_vport->dev, get_dpdev(vport->dp));
error_unlock:
	rtnl_unlock();
error_put:
	dev_put(netdev_vport->dev);
error_free_vport:
	ovs_vport_free(vport);
error:
	return ERR_PTR(err);
}

static void free_port_rcu(struct rcu_head *rcu)
{
	struct netdev_vport *netdev_vport = container_of(rcu,
					struct netdev_vport, rcu);

	dev_put(netdev_vport->dev);
	ovs_vport_free(vport_from_priv(netdev_vport));
}

static void netdev_destroy(struct vport *vport)
{
	struct netdev_vport *netdev_vport = netdev_vport_priv(vport);

	netdev_exit();
	rtnl_lock();
	netdev_vport->dev->priv_flags &= ~IFF_OVS_DATAPATH;
	netdev_rx_handler_unregister(netdev_vport->dev);
	netdev_upper_dev_unlink(netdev_vport->dev, get_dpdev(vport->dp));
	dev_set_promiscuity(netdev_vport->dev, -1);
	rtnl_unlock();

	call_rcu(&netdev_vport->rcu, free_port_rcu);
}

const char *ovs_netdev_get_name(const struct vport *vport)
{
	const struct netdev_vport *netdev_vport = netdev_vport_priv(vport);
	return netdev_vport->dev->name;
}

/* Must be called with rcu_read_lock. */
static void netdev_port_receive(struct vport *vport, struct sk_buff *skb)
{
	if (unlikely(!vport))
		goto error;

	if (unlikely(skb_warn_if_lro(skb)))
		goto error;

	/* Make our own copy of the packet.  Otherwise we will mangle the
	 * packet for anyone who came before us (e.g. tcpdump via AF_PACKET).
	 * (No one comes after us, since we tell handle_bridge() that we took
	 * the packet.) */
	skb = skb_share_check(skb, GFP_ATOMIC);
	if (unlikely(!skb))
		return;

	if (unlikely(compute_ip_summed(skb, false)))
		goto error;

	skb_push(skb, ETH_HLEN);
	ovs_skb_postpush_rcsum(skb, skb->data, ETH_HLEN);

	vlan_copy_skb_tci(skb);

	ovs_vport_receive(vport, skb, NULL);
	return;

error:
	kfree_skb(skb);
}

static unsigned int packet_length(const struct sk_buff *skb)
{
	unsigned int length = skb->len - ETH_HLEN;

	if (skb->protocol == htons(ETH_P_8021Q))
		length -= VLAN_HLEN;

	return length;
}

static bool dev_supports_vlan_tx(struct net_device *dev)
{
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,37)
	/* Software fallback means every device supports vlan_tci on TX. */
	return true;
#elif defined(HAVE_VLAN_BUG_WORKAROUND)
	return dev->features & NETIF_F_HW_VLAN_TX;
#else
	/* Assume that the driver is buggy. */
	return false;
#endif
}

static int netdev_send(struct vport *vport, struct sk_buff *skb)
{
	struct netdev_vport *netdev_vport = netdev_vport_priv(vport);
	int mtu = netdev_vport->dev->mtu;
	int len;

	if (unlikely(packet_length(skb) > mtu && !skb_is_gso(skb))) {
		net_warn_ratelimited("%s: dropped over-mtu packet: %d > %d\n",
				     netdev_vport->dev->name,
				     packet_length(skb), mtu);
		goto drop;
	}

	skb->dev = netdev_vport->dev;
	forward_ip_summed(skb, true);

	if (vlan_tx_tag_present(skb) && !dev_supports_vlan_tx(skb->dev)) {
		int features;

		features = netif_skb_features(skb);

		if (!vlan_tso)
			features &= ~(NETIF_F_TSO | NETIF_F_TSO6 |
				      NETIF_F_UFO | NETIF_F_FSO);

		if (netif_needs_gso(skb, features)) {
			struct sk_buff *nskb;

			nskb = skb_gso_segment(skb, features);
			if (!nskb) {
				if (unlikely(skb_cloned(skb) &&
				    pskb_expand_head(skb, 0, 0, GFP_ATOMIC)))
					goto drop;

				skb_shinfo(skb)->gso_type &= ~SKB_GSO_DODGY;
				goto tag;
			}

			if (IS_ERR(nskb))
				goto drop;
			consume_skb(skb);
			skb = nskb;

			len = 0;
			do {
				nskb = skb->next;
				skb->next = NULL;

				skb = __vlan_put_tag(skb, skb->vlan_proto, vlan_tx_tag_get(skb));
				if (likely(skb)) {
					len += skb->len;
					vlan_set_tci(skb, 0);
					dev_queue_xmit(skb);
				}

				skb = nskb;
			} while (skb);

			return len;
		}

tag:
		skb = __vlan_put_tag(skb, skb->vlan_proto, vlan_tx_tag_get(skb));
		if (unlikely(!skb))
			return 0;
		vlan_set_tci(skb, 0);
	}

	len = skb->len;
	dev_queue_xmit(skb);

	return len;

drop:
	kfree_skb(skb);
	return 0;
}

/* Returns null if this device is not attached to a datapath. */
struct vport *ovs_netdev_get_vport(struct net_device *dev)
{
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36) || \
    defined HAVE_RHEL_OVS_HOOK
#if IFF_OVS_DATAPATH != 0
	if (likely(dev->priv_flags & IFF_OVS_DATAPATH))
#else
	if (likely(rcu_access_pointer(dev->rx_handler) == netdev_frame_hook))
#endif
#ifdef HAVE_RHEL_OVS_HOOK
		return (struct vport *)rcu_dereference_rtnl(dev->ax25_ptr);
#else
		return (struct vport *)rcu_dereference_rtnl(dev->rx_handler_data);
#endif
	else
		return NULL;
#else
	return (struct vport *)rcu_dereference_rtnl(dev->br_port);
#endif
}

const struct vport_ops ovs_netdev_vport_ops = {
	.type		= OVS_VPORT_TYPE_NETDEV,
	.create		= netdev_create,
	.destroy	= netdev_destroy,
	.get_name	= ovs_netdev_get_name,
	.send		= netdev_send,
};

#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,36) && \
    !defined HAVE_RHEL_OVS_HOOK
/*
 * Enforces, mutual exclusion with the Linux bridge module, by declaring and
 * exporting br_should_route_hook.  Because the bridge module also exports the
 * same symbol, the module loader will refuse to load both modules at the same
 * time (e.g. "bridge: exports duplicate symbol br_should_route_hook (owned by
 * openvswitch)").
 *
 * Before Linux 2.6.36, Open vSwitch cannot safely coexist with the Linux
 * bridge module, so openvswitch uses this macro in those versions.  In
 * Linux 2.6.36 and later, Open vSwitch can coexist with the bridge module.
 *
 * The use of "typeof" here avoids the need to track changes in the type of
 * br_should_route_hook over various kernel versions.
 */
typeof(br_should_route_hook) br_should_route_hook;
EXPORT_SYMBOL(br_should_route_hook);
#endif
Commit	Line	Data
f2459fe7	1	/*
e0edde6f	2	* Copyright (c) 2007-2012 Nicira, Inc.
f2459fe7	3	*
a9a29d22 JG	4	* This program is free software; you can redistribute it and/or
	5	* modify it under the terms of version 2 of the GNU General Public
	6	* License as published by the Free Software Foundation.
	7	*
	8	* This program is distributed in the hope that it will be useful, but
	9	* WITHOUT ANY WARRANTY; without even the implied warranty of
	10	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	11	* General Public License for more details.
	12	*
	13	* You should have received a copy of the GNU General Public License
	14	* along with this program; if not, write to the Free Software
	15	* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
	16	* 02110-1301, USA
f2459fe7 JG	17	*/
f2459fe7 JG	18
24b019f8 JP	19	#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
24b019f8 JP	20
f2459fe7 JG	21	#include <linux/if_arp.h>
	22	#include <linux/if_bridge.h>
	23	#include <linux/if_vlan.h>
	24	#include <linux/kernel.h>
	25	#include <linux/llc.h>
	26	#include <linux/rtnetlink.h>
	27	#include <linux/skbuff.h>
2b51596f	28	#include <linux/openvswitch.h>
f2459fe7 JG	29
	30	#include <net/llc.h>
	31
dd8d6b8c	32	#include "checksum.h"
f2459fe7	33	#include "datapath.h"
6e0ce48e	34	#include "vlan.h"
f2459fe7 JG	35	#include "vport-internal_dev.h"
	36	#include "vport-netdev.h"
	37
431488e6	38	#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,37) && \
6455100f	39	!defined(HAVE_VLAN_BUG_WORKAROUND)
3f6256af JG	40	#include <linux/module.h>
3f6256af JG	41
6455100f	42	static int vlan_tso __read_mostly;
3f6256af JG	43	module_param(vlan_tso, int, 0644);
3f6256af JG	44	MODULE_PARM_DESC(vlan_tso, "Enable TSO for VLAN packets");
431488e6 BP	45	#else
431488e6 BP	46	#define vlan_tso true
3f6256af JG	47	#endif
3f6256af JG	48
43158509	49	static void netdev_port_receive(struct vport vport, struct sk_buff skb);
f2459fe7	50
898ba403 AE	51	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,39)
	52	/* Called with rcu_read_lock and bottom-halves disabled. */
	53	static rx_handler_result_t netdev_frame_hook(struct sk_buff **pskb)
	54	{
	55	struct sk_buff skb = pskb;
	56	struct vport *vport;
	57
	58	if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
	59	return RX_HANDLER_PASS;
	60
850b6b3b	61	vport = ovs_netdev_get_vport(skb->dev);
898ba403 AE	62
	63	netdev_port_receive(vport, skb);
	64
	65	return RX_HANDLER_CONSUMED;
	66	}
f285d3e7 TG	67	#elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36) \|\| \
f285d3e7 TG	68	defined HAVE_RHEL_OVS_HOOK
43158509 SH	69	/* Called with rcu_read_lock and bottom-halves disabled. */
	70	static struct sk_buff netdev_frame_hook(struct sk_buff skb)
	71	{
	72	struct vport *vport;
	73
	74	if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
	75	return skb;
	76
0145d7ed	77	vport = ovs_netdev_get_vport(skb->dev);
43158509 SH	78
	79	netdev_port_receive(vport, skb);
	80
	81	return NULL;
	82	}
	83	#elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
f2459fe7 JG	84	/*
	85	* Used as br_handle_frame_hook. (Cannot run bridge at the same time, even on
	86	* different set of devices!)
	87	*/
f2459fe7	88	/* Called with rcu_read_lock and bottom-halves disabled. */
fceb2a5b JG	89	static struct sk_buff netdev_frame_hook(struct net_bridge_port p,
fceb2a5b JG	90	struct sk_buff *skb)
f2459fe7	91	{
43158509	92	netdev_port_receive((struct vport *)p, skb);
f2459fe7 JG	93	return NULL;
	94	}
	95	#elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0)
43158509 SH	96	/*
	97	* Used as br_handle_frame_hook. (Cannot run bridge at the same time, even on
	98	* different set of devices!)
	99	*/
f2459fe7	100	/* Called with rcu_read_lock and bottom-halves disabled. */
fceb2a5b	101	static int netdev_frame_hook(struct net_bridge_port p, struct sk_buff *pskb)
f2459fe7	102	{
43158509	103	netdev_port_receive((struct vport )p, pskb);
f2459fe7 JG	104	return 1;
	105	}
	106	#else
	107	#error
	108	#endif
	109
f285d3e7 TG	110	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36) \|\| \
f285d3e7 TG	111	defined HAVE_RHEL_OVS_HOOK
43158509 SH	112	static int netdev_init(void) { return 0; }
	113	static void netdev_exit(void) { }
	114	#else
806b46ef PS	115	static int port_count;
	116
	117	static void netdev_init(void)
f2459fe7	118	{
806b46ef PS	119	port_count++;
	120	if (port_count > 1)
	121	return;
	122
f2459fe7 JG	123	/* Hook into callback used by the bridge to intercept packets.
	124	* Parasites we are. */
	125	br_handle_frame_hook = netdev_frame_hook;
	126
806b46ef	127	return;
f2459fe7 JG	128	}
f2459fe7 JG	129
fbb29a54	130	static void netdev_exit(void)
f2459fe7	131	{
806b46ef PS	132	port_count--;
	133	if (port_count > 0)
	134	return;
	135
f2459fe7	136	br_handle_frame_hook = NULL;
f2459fe7	137	}
43158509	138	#endif
f2459fe7	139
2b51596f JP	140	static struct net_device get_dpdev(struct datapath dp)
	141	{
	142	struct vport *local;
	143
	144	local = ovs_vport_ovsl(dp, OVSP_LOCAL);
	145	BUG_ON(!local);
	146	return netdev_vport_priv(local)->dev;
	147	}
	148
94903c98	149	static struct vport netdev_create(const struct vport_parms parms)
f2459fe7 JG	150	{
	151	struct vport *vport;
	152	struct netdev_vport *netdev_vport;
	153	int err;
	154
850b6b3b JG	155	vport = ovs_vport_alloc(sizeof(struct netdev_vport),
850b6b3b JG	156	&ovs_netdev_vport_ops, parms);
f2459fe7 JG	157	if (IS_ERR(vport)) {
	158	err = PTR_ERR(vport);
	159	goto error;
	160	}
	161
	162	netdev_vport = netdev_vport_priv(vport);
	163
2a4999f3	164	netdev_vport->dev = dev_get_by_name(ovs_dp_get_net(vport->dp), parms->name);
f2459fe7 JG	165	if (!netdev_vport->dev) {
	166	err = -ENODEV;
	167	goto error_free_vport;
	168	}
	169
	170	if (netdev_vport->dev->flags & IFF_LOOPBACK \|\|
	171	netdev_vport->dev->type != ARPHRD_ETHER \|\|
850b6b3b	172	ovs_is_internal_dev(netdev_vport->dev)) {
f2459fe7 JG	173	err = -EINVAL;
	174	goto error_put;
	175	}
	176
cd2a59e9	177	rtnl_lock();
2b51596f JP	178	err = netdev_master_upper_dev_link(netdev_vport->dev,
	179	get_dpdev(vport->dp));
	180	if (err)
	181	goto error_unlock;
	182
7237e4f4 BP	183	err = netdev_rx_handler_register(netdev_vport->dev, netdev_frame_hook,
	184	vport);
	185	if (err)
2b51596f	186	goto error_master_upper_dev_unlink;
7237e4f4 BP	187
7237e4f4 BP	188	dev_set_promiscuity(netdev_vport->dev, 1);
24b019f8	189	#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,24)
7237e4f4	190	dev_disable_lro(netdev_vport->dev);
24b019f8	191	#endif
7237e4f4	192	netdev_vport->dev->priv_flags \|= IFF_OVS_DATAPATH;
cd2a59e9	193	rtnl_unlock();
7237e4f4	194
806b46ef	195	netdev_init();
f2459fe7 JG	196	return vport;
f2459fe7 JG	197
2b51596f JP	198	error_master_upper_dev_unlink:
2b51596f JP	199	netdev_upper_dev_unlink(netdev_vport->dev, get_dpdev(vport->dp));
cd2a59e9 PS	200	error_unlock:
cd2a59e9 PS	201	rtnl_unlock();
f2459fe7 JG	202	error_put:
	203	dev_put(netdev_vport->dev);
	204	error_free_vport:
850b6b3b	205	ovs_vport_free(vport);
f2459fe7 JG	206	error:
	207	return ERR_PTR(err);
	208	}
	209
e65361d6 JG	210	static void free_port_rcu(struct rcu_head *rcu)
	211	{
	212	struct netdev_vport *netdev_vport = container_of(rcu,
	213	struct netdev_vport, rcu);
	214
	215	dev_put(netdev_vport->dev);
	216	ovs_vport_free(vport_from_priv(netdev_vport));
	217	}
	218
3544358a	219	static void netdev_destroy(struct vport *vport)
f2459fe7 JG	220	{
	221	struct netdev_vport *netdev_vport = netdev_vport_priv(vport);
	222
806b46ef	223	netdev_exit();
cd2a59e9	224	rtnl_lock();
37f055c7	225	netdev_vport->dev->priv_flags &= ~IFF_OVS_DATAPATH;
43158509	226	netdev_rx_handler_unregister(netdev_vport->dev);
2b51596f	227	netdev_upper_dev_unlink(netdev_vport->dev, get_dpdev(vport->dp));
f2459fe7	228	dev_set_promiscuity(netdev_vport->dev, -1);
cd2a59e9	229	rtnl_unlock();
f2459fe7	230
e65361d6	231	call_rcu(&netdev_vport->rcu, free_port_rcu);
f2459fe7 JG	232	}
f2459fe7 JG	233
850b6b3b	234	const char ovs_netdev_get_name(const struct vport vport)
f2459fe7 JG	235	{
	236	const struct netdev_vport *netdev_vport = netdev_vport_priv(vport);
	237	return netdev_vport->dev->name;
	238	}
	239
f2459fe7	240	/* Must be called with rcu_read_lock. */
43158509	241	static void netdev_port_receive(struct vport vport, struct sk_buff skb)
f2459fe7	242	{
33e031e9 JG	243	if (unlikely(!vport))
	244	goto error;
	245
	246	if (unlikely(skb_warn_if_lro(skb)))
	247	goto error;
d825e2a5	248
f2459fe7 JG	249	/* Make our own copy of the packet. Otherwise we will mangle the
	250	* packet for anyone who came before us (e.g. tcpdump via AF_PACKET).
	251	* (No one comes after us, since we tell handle_bridge() that we took
	252	* the packet.) */
	253	skb = skb_share_check(skb, GFP_ATOMIC);
67c74f75	254	if (unlikely(!skb))
f2459fe7 JG	255	return;
f2459fe7 JG	256
33e031e9 JG	257	if (unlikely(compute_ip_summed(skb, false)))
	258	goto error;
	259
3cfede14 PS	260	skb_push(skb, ETH_HLEN);
	261	ovs_skb_postpush_rcsum(skb, skb->data, ETH_HLEN);
	262
6e0ce48e	263	vlan_copy_skb_tci(skb);
f2459fe7	264
db0c3134	265	ovs_vport_receive(vport, skb, NULL);
33e031e9 JG	266	return;
	267
	268	error:
	269	kfree_skb(skb);
f2459fe7 JG	270	}
f2459fe7 JG	271
f1b8a922	272	static unsigned int packet_length(const struct sk_buff *skb)
24b019f8	273	{
f1b8a922	274	unsigned int length = skb->len - ETH_HLEN;
24b019f8 JP	275
	276	if (skb->protocol == htons(ETH_P_8021Q))
	277	length -= VLAN_HLEN;
	278
	279	return length;
	280	}
	281
431488e6 BP	282	static bool dev_supports_vlan_tx(struct net_device *dev)
	283	{
	284	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,37)
	285	/* Software fallback means every device supports vlan_tci on TX. */
	286	return true;
	287	#elif defined(HAVE_VLAN_BUG_WORKAROUND)
	288	return dev->features & NETIF_F_HW_VLAN_TX;
	289	#else
	290	/* Assume that the driver is buggy. */
	291	return false;
	292	#endif
	293	}
	294
fceb2a5b	295	static int netdev_send(struct vport vport, struct sk_buff skb)
f2459fe7 JG	296	{
f2459fe7 JG	297	struct netdev_vport *netdev_vport = netdev_vport_priv(vport);
24b019f8	298	int mtu = netdev_vport->dev->mtu;
6ce39213	299	int len;
f2459fe7	300
24b019f8	301	if (unlikely(packet_length(skb) > mtu && !skb_is_gso(skb))) {
51885213	302	net_warn_ratelimited("%s: dropped over-mtu packet: %d > %d\n",
40a5a25d	303	netdev_vport->dev->name,
51885213	304	packet_length(skb), mtu);
be7cd27e	305	goto drop;
24b019f8 JP	306	}
24b019f8 JP	307
f2459fe7	308	skb->dev = netdev_vport->dev;
c3729ee4	309	forward_ip_summed(skb, true);
6ce39213	310
431488e6	311	if (vlan_tx_tag_present(skb) && !dev_supports_vlan_tx(skb->dev)) {
d9065a90	312	int features;
2d8916e5	313
d9065a90	314	features = netif_skb_features(skb);
6ce39213	315
3f6256af JG	316	if (!vlan_tso)
	317	features &= ~(NETIF_F_TSO \| NETIF_F_TSO6 \|
	318	NETIF_F_UFO \| NETIF_F_FSO);
	319
d9065a90	320	if (netif_needs_gso(skb, features)) {
6ce39213 JG	321	struct sk_buff *nskb;
6ce39213 JG	322
3f6256af JG	323	nskb = skb_gso_segment(skb, features);
	324	if (!nskb) {
	325	if (unlikely(skb_cloned(skb) &&
be7cd27e PS	326	pskb_expand_head(skb, 0, 0, GFP_ATOMIC)))
be7cd27e PS	327	goto drop;
3f6256af JG	328
	329	skb_shinfo(skb)->gso_type &= ~SKB_GSO_DODGY;
	330	goto tag;
	331	}
	332
be7cd27e PS	333	if (IS_ERR(nskb))
be7cd27e PS	334	goto drop;
5b95ab0e JG	335	consume_skb(skb);
5b95ab0e JG	336	skb = nskb;
6ce39213 JG	337
	338	len = 0;
	339	do {
	340	nskb = skb->next;
	341	skb->next = NULL;
	342
9b764edf	343	skb = __vlan_put_tag(skb, skb->vlan_proto, vlan_tx_tag_get(skb));
6ce39213 JG	344	if (likely(skb)) {
	345	len += skb->len;
	346	vlan_set_tci(skb, 0);
	347	dev_queue_xmit(skb);
	348	}
	349
	350	skb = nskb;
	351	} while (skb);
	352
	353	return len;
6ce39213	354	}
3f6256af JG	355
3f6256af JG	356	tag:
9b764edf	357	skb = __vlan_put_tag(skb, skb->vlan_proto, vlan_tx_tag_get(skb));
3f6256af JG	358	if (unlikely(!skb))
	359	return 0;
	360	vlan_set_tci(skb, 0);
6ce39213	361	}
6ce39213 JG	362
6ce39213 JG	363	len = skb->len;
f2459fe7 JG	364	dev_queue_xmit(skb);
	365
	366	return len;
24b019f8	367
be7cd27e	368	drop:
24b019f8	369	kfree_skb(skb);
24b019f8	370	return 0;
f2459fe7 JG	371	}
	372
	373	/* Returns null if this device is not attached to a datapath. */
850b6b3b	374	struct vport ovs_netdev_get_vport(struct net_device dev)
f2459fe7	375	{
f285d3e7 TG	376	#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36) \|\| \
f285d3e7 TG	377	defined HAVE_RHEL_OVS_HOOK
e25c55d2	378	#if IFF_OVS_DATAPATH != 0
055dae4f JG	379	if (likely(dev->priv_flags & IFF_OVS_DATAPATH))
055dae4f JG	380	#else
6455100f	381	if (likely(rcu_access_pointer(dev->rx_handler) == netdev_frame_hook))
055dae4f	382	#endif
f285d3e7 TG	383	#ifdef HAVE_RHEL_OVS_HOOK
	384	return (struct vport *)rcu_dereference_rtnl(dev->ax25_ptr);
	385	#else
e33adfd0	386	return (struct vport *)rcu_dereference_rtnl(dev->rx_handler_data);
f285d3e7	387	#endif
055dae4f	388	else
37f055c7	389	return NULL;
37f055c7	390	#else
e33adfd0	391	return (struct vport *)rcu_dereference_rtnl(dev->br_port);
37f055c7	392	#endif
f2459fe7 JG	393	}
f2459fe7 JG	394
850b6b3b	395	const struct vport_ops ovs_netdev_vport_ops = {
df2c07f4	396	.type = OVS_VPORT_TYPE_NETDEV,
f2459fe7 JG	397	.create = netdev_create,
f2459fe7 JG	398	.destroy = netdev_destroy,
850b6b3b	399	.get_name = ovs_netdev_get_name,
f2459fe7 JG	400	.send = netdev_send,
f2459fe7 JG	401	};
106eab5d	402
f285d3e7 TG	403	#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,36) && \
f285d3e7 TG	404	!defined HAVE_RHEL_OVS_HOOK
106eab5d	405	/*
5ca1ba48 PS	406	* Enforces, mutual exclusion with the Linux bridge module, by declaring and
	407	* exporting br_should_route_hook. Because the bridge module also exports the
	408	* same symbol, the module loader will refuse to load both modules at the same
	409	* time (e.g. "bridge: exports duplicate symbol br_should_route_hook (owned by
	410	* openvswitch)").
	411	*
	412	* Before Linux 2.6.36, Open vSwitch cannot safely coexist with the Linux
	413	* bridge module, so openvswitch uses this macro in those versions. In
	414	* Linux 2.6.36 and later, Open vSwitch can coexist with the bridge module.
	415	*
	416	* The use of "typeof" here avoids the need to track changes in the type of
	417	* br_should_route_hook over various kernel versions.
106eab5d	418	*/
5ca1ba48 PS	419	typeof(br_should_route_hook) br_should_route_hook;
5ca1ba48 PS	420	EXPORT_SYMBOL(br_should_route_hook);
37f055c7	421	#endif