[linux-block.git] / net / xfrm / xfrm_device.c

/*
 * xfrm_device.c - IPsec device offloading code.
 *
 * Copyright (c) 2015 secunet Security Networks AG
 *
 * Author:
 * Steffen Klassert <steffen.klassert@secunet.com>
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License
 * as published by the Free Software Foundation; either version
 * 2 of the License, or (at your option) any later version.
 */

#include <linux/errno.h>
#include <linux/module.h>
#include <linux/netdevice.h>
#include <linux/skbuff.h>
#include <linux/slab.h>
#include <linux/spinlock.h>
#include <net/dst.h>
#include <net/xfrm.h>
#include <linux/notifier.h>

#ifdef CONFIG_XFRM_OFFLOAD
struct sk_buff *validate_xmit_xfrm(struct sk_buff *skb, netdev_features_t features, bool *again)
{
	int err;
	unsigned long flags;
	struct xfrm_state *x;
	struct sk_buff *skb2;
	struct softnet_data *sd;
	netdev_features_t esp_features = features;
	struct xfrm_offload *xo = xfrm_offload(skb);

	if (!xo)
		return skb;

	if (!(features & NETIF_F_HW_ESP))
		esp_features = features & ~(NETIF_F_SG | NETIF_F_CSUM_MASK);

	x = skb->sp->xvec[skb->sp->len - 1];
	if (xo->flags & XFRM_GRO || x->xso.flags & XFRM_OFFLOAD_INBOUND)
		return skb;

	local_irq_save(flags);
	sd = this_cpu_ptr(&softnet_data);
	err = !skb_queue_empty(&sd->xfrm_backlog);
	local_irq_restore(flags);

	if (err) {
		*again = true;
		return skb;
	}

	if (skb_is_gso(skb)) {
		struct net_device *dev = skb->dev;

		if (unlikely(x->xso.dev != dev)) {
			struct sk_buff *segs;

			/* Packet got rerouted, fixup features and segment it. */
			esp_features = esp_features & ~(NETIF_F_HW_ESP
							| NETIF_F_GSO_ESP);

			segs = skb_gso_segment(skb, esp_features);
			if (IS_ERR(segs)) {
				kfree_skb(skb);
				atomic_long_inc(&dev->tx_dropped);
				return NULL;
			} else {
				consume_skb(skb);
				skb = segs;
			}
		}
	}

	if (!skb->next) {
		x->outer_mode->xmit(x, skb);

		xo->flags |= XFRM_DEV_RESUME;

		err = x->type_offload->xmit(x, skb, esp_features);
		if (err) {
			if (err == -EINPROGRESS)
				return NULL;

			XFRM_INC_STATS(xs_net(x), LINUX_MIB_XFRMOUTSTATEPROTOERROR);
			kfree_skb(skb);
			return NULL;
		}

		skb_push(skb, skb->data - skb_mac_header(skb));

		return skb;
	}

	skb2 = skb;

	do {
		struct sk_buff *nskb = skb2->next;
		skb2->next = NULL;

		xo = xfrm_offload(skb2);
		xo->flags |= XFRM_DEV_RESUME;

		x->outer_mode->xmit(x, skb2);

		err = x->type_offload->xmit(x, skb2, esp_features);
		if (!err) {
			skb2->next = nskb;
		} else if (err != -EINPROGRESS) {
			XFRM_INC_STATS(xs_net(x), LINUX_MIB_XFRMOUTSTATEPROTOERROR);
			skb2->next = nskb;
			kfree_skb_list(skb2);
			return NULL;
		} else {
			if (skb == skb2)
				skb = nskb;

			if (!skb)
				return NULL;

			goto skip_push;
		}

		skb_push(skb2, skb2->data - skb_mac_header(skb2));

skip_push:
		skb2 = nskb;
	} while (skb2);

	return skb;
}
EXPORT_SYMBOL_GPL(validate_xmit_xfrm);

int xfrm_dev_state_add(struct net *net, struct xfrm_state *x,
		       struct xfrm_user_offload *xuo)
{
	int err;
	struct dst_entry *dst;
	struct net_device *dev;
	struct xfrm_state_offload *xso = &x->xso;
	xfrm_address_t *saddr;
	xfrm_address_t *daddr;

	if (!x->type_offload)
		return -EINVAL;

	/* We don't yet support UDP encapsulation and TFC padding. */
	if (x->encap || x->tfcpad)
		return -EINVAL;

	dev = dev_get_by_index(net, xuo->ifindex);
	if (!dev) {
		if (!(xuo->flags & XFRM_OFFLOAD_INBOUND)) {
			saddr = &x->props.saddr;
			daddr = &x->id.daddr;
		} else {
			saddr = &x->id.daddr;
			daddr = &x->props.saddr;
		}

		dst = __xfrm_dst_lookup(net, 0, 0, saddr, daddr,
					x->props.family,
					xfrm_smark_get(0, x));
		if (IS_ERR(dst))
			return 0;

		dev = dst->dev;

		dev_hold(dev);
		dst_release(dst);
	}

	if (!dev->xfrmdev_ops || !dev->xfrmdev_ops->xdo_dev_state_add) {
		xso->dev = NULL;
		dev_put(dev);
		return 0;
	}

	if (x->props.flags & XFRM_STATE_ESN &&
	    !dev->xfrmdev_ops->xdo_dev_state_advance_esn) {
		xso->dev = NULL;
		dev_put(dev);
		return -EINVAL;
	}

	xso->dev = dev;
	xso->num_exthdrs = 1;
	xso->flags = xuo->flags;

	err = dev->xfrmdev_ops->xdo_dev_state_add(x);
	if (err) {
		xso->dev = NULL;
		dev_put(dev);
		return err;
	}

	return 0;
}
EXPORT_SYMBOL_GPL(xfrm_dev_state_add);

bool xfrm_dev_offload_ok(struct sk_buff *skb, struct xfrm_state *x)
{
	int mtu;
	struct dst_entry *dst = skb_dst(skb);
	struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
	struct net_device *dev = x->xso.dev;

	if (!x->type_offload || x->encap)
		return false;

	if ((!dev || (dev == xfrm_dst_path(dst)->dev)) &&
	    (!xdst->child->xfrm && x->type->get_mtu)) {
		mtu = x->type->get_mtu(x, xdst->child_mtu_cached);

		if (skb->len <= mtu)
			goto ok;

		if (skb_is_gso(skb) && skb_gso_validate_network_len(skb, mtu))
			goto ok;
	}

	return false;

ok:
	if (dev && dev->xfrmdev_ops && dev->xfrmdev_ops->xdo_dev_offload_ok)
		return x->xso.dev->xfrmdev_ops->xdo_dev_offload_ok(skb, x);

	return true;
}
EXPORT_SYMBOL_GPL(xfrm_dev_offload_ok);

void xfrm_dev_resume(struct sk_buff *skb)
{
	struct net_device *dev = skb->dev;
	int ret = NETDEV_TX_BUSY;
	struct netdev_queue *txq;
	struct softnet_data *sd;
	unsigned long flags;

	rcu_read_lock();
	txq = netdev_pick_tx(dev, skb, NULL);

	HARD_TX_LOCK(dev, txq, smp_processor_id());
	if (!netif_xmit_frozen_or_stopped(txq))
		skb = dev_hard_start_xmit(skb, dev, txq, &ret);
	HARD_TX_UNLOCK(dev, txq);

	if (!dev_xmit_complete(ret)) {
		local_irq_save(flags);
		sd = this_cpu_ptr(&softnet_data);
		skb_queue_tail(&sd->xfrm_backlog, skb);
		raise_softirq_irqoff(NET_TX_SOFTIRQ);
		local_irq_restore(flags);
	}
	rcu_read_unlock();
}
EXPORT_SYMBOL_GPL(xfrm_dev_resume);

void xfrm_dev_backlog(struct softnet_data *sd)
{
	struct sk_buff_head *xfrm_backlog = &sd->xfrm_backlog;
	struct sk_buff_head list;
	struct sk_buff *skb;

	if (skb_queue_empty(xfrm_backlog))
		return;

	__skb_queue_head_init(&list);

	spin_lock(&xfrm_backlog->lock);
	skb_queue_splice_init(xfrm_backlog, &list);
	spin_unlock(&xfrm_backlog->lock);

	while (!skb_queue_empty(&list)) {
		skb = __skb_dequeue(&list);
		xfrm_dev_resume(skb);
	}

}
#endif

static int xfrm_api_check(struct net_device *dev)
{
#ifdef CONFIG_XFRM_OFFLOAD
	if ((dev->features & NETIF_F_HW_ESP_TX_CSUM) &&
	    !(dev->features & NETIF_F_HW_ESP))
		return NOTIFY_BAD;

	if ((dev->features & NETIF_F_HW_ESP) &&
	    (!(dev->xfrmdev_ops &&
	       dev->xfrmdev_ops->xdo_dev_state_add &&
	       dev->xfrmdev_ops->xdo_dev_state_delete)))
		return NOTIFY_BAD;
#else
	if (dev->features & (NETIF_F_HW_ESP | NETIF_F_HW_ESP_TX_CSUM))
		return NOTIFY_BAD;
#endif

	return NOTIFY_DONE;
}

static int xfrm_dev_register(struct net_device *dev)
{
	return xfrm_api_check(dev);
}

static int xfrm_dev_feat_change(struct net_device *dev)
{
	return xfrm_api_check(dev);
}

static int xfrm_dev_down(struct net_device *dev)
{
	if (dev->features & NETIF_F_HW_ESP)
		xfrm_dev_state_flush(dev_net(dev), dev, true);

	return NOTIFY_DONE;
}

static int xfrm_dev_event(struct notifier_block *this, unsigned long event, void *ptr)
{
	struct net_device *dev = netdev_notifier_info_to_dev(ptr);

	switch (event) {
	case NETDEV_REGISTER:
		return xfrm_dev_register(dev);

	case NETDEV_FEAT_CHANGE:
		return xfrm_dev_feat_change(dev);

	case NETDEV_DOWN:
		return xfrm_dev_down(dev);
	}
	return NOTIFY_DONE;
}

static struct notifier_block xfrm_dev_notifier = {
	.notifier_call	= xfrm_dev_event,
};

void __init xfrm_dev_init(void)
{
	register_netdevice_notifier(&xfrm_dev_notifier);
}
Commit	Line	Data
21f42cc9 SK	1	/*
	2	* xfrm_device.c - IPsec device offloading code.
	3	*
	4	* Copyright (c) 2015 secunet Security Networks AG
	5	*
	6	* Author:
	7	* Steffen Klassert <steffen.klassert@secunet.com>
	8	*
	9	* This program is free software; you can redistribute it and/or
	10	* modify it under the terms of the GNU General Public License
	11	* as published by the Free Software Foundation; either version
	12	* 2 of the License, or (at your option) any later version.
	13	*/
	14
	15	#include <linux/errno.h>
	16	#include <linux/module.h>
	17	#include <linux/netdevice.h>
	18	#include <linux/skbuff.h>
	19	#include <linux/slab.h>
	20	#include <linux/spinlock.h>
	21	#include <net/dst.h>
	22	#include <net/xfrm.h>
	23	#include <linux/notifier.h>
	24
b81f884a	25	#ifdef CONFIG_XFRM_OFFLOAD
f53c7239	26	struct sk_buff validate_xmit_xfrm(struct sk_buff skb, netdev_features_t features, bool *again)
f6e27114 SK	27	{
f6e27114 SK	28	int err;
f53c7239	29	unsigned long flags;
f6e27114	30	struct xfrm_state *x;
3dca3f38	31	struct sk_buff *skb2;
f53c7239	32	struct softnet_data *sd;
3dca3f38	33	netdev_features_t esp_features = features;
f6e27114 SK	34	struct xfrm_offload *xo = xfrm_offload(skb);
f6e27114 SK	35
3dca3f38 SK	36	if (!xo)
3dca3f38 SK	37	return skb;
f6e27114	38
3dca3f38 SK	39	if (!(features & NETIF_F_HW_ESP))
3dca3f38 SK	40	esp_features = features & ~(NETIF_F_SG \| NETIF_F_CSUM_MASK);
f6e27114	41
3dca3f38 SK	42	x = skb->sp->xvec[skb->sp->len - 1];
	43	if (xo->flags & XFRM_GRO \|\| x->xso.flags & XFRM_OFFLOAD_INBOUND)
	44	return skb;
	45
f53c7239 SK	46	local_irq_save(flags);
	47	sd = this_cpu_ptr(&softnet_data);
	48	err = !skb_queue_empty(&sd->xfrm_backlog);
	49	local_irq_restore(flags);
	50
	51	if (err) {
	52	*again = true;
	53	return skb;
	54	}
	55
3dca3f38 SK	56	if (skb_is_gso(skb)) {
	57	struct net_device *dev = skb->dev;
	58
fcb662de	59	if (unlikely(x->xso.dev != dev)) {
3dca3f38 SK	60	struct sk_buff *segs;
	61
	62	/* Packet got rerouted, fixup features and segment it. */
	63	esp_features = esp_features & ~(NETIF_F_HW_ESP
	64	\| NETIF_F_GSO_ESP);
	65
	66	segs = skb_gso_segment(skb, esp_features);
	67	if (IS_ERR(segs)) {
3dca3f38	68	kfree_skb(skb);
f53c7239	69	atomic_long_inc(&dev->tx_dropped);
3dca3f38 SK	70	return NULL;
	71	} else {
	72	consume_skb(skb);
	73	skb = segs;
	74	}
3dca3f38 SK	75	}
	76	}
	77
	78	if (!skb->next) {
f6e27114 SK	79	x->outer_mode->xmit(x, skb);
f6e27114 SK	80
f53c7239 SK	81	xo->flags \|= XFRM_DEV_RESUME;
f53c7239 SK	82
3dca3f38	83	err = x->type_offload->xmit(x, skb, esp_features);
f6e27114	84	if (err) {
f53c7239 SK	85	if (err == -EINPROGRESS)
	86	return NULL;
	87
f6e27114	88	XFRM_INC_STATS(xs_net(x), LINUX_MIB_XFRMOUTSTATEPROTOERROR);
3dca3f38 SK	89	kfree_skb(skb);
3dca3f38 SK	90	return NULL;
f6e27114 SK	91	}
	92
	93	skb_push(skb, skb->data - skb_mac_header(skb));
3dca3f38 SK	94
3dca3f38 SK	95	return skb;
f6e27114 SK	96	}
f6e27114 SK	97
3dca3f38	98	skb2 = skb;
3dca3f38 SK	99
	100	do {
	101	struct sk_buff *nskb = skb2->next;
f53c7239	102	skb2->next = NULL;
3dca3f38 SK	103
3dca3f38 SK	104	xo = xfrm_offload(skb2);
f53c7239	105	xo->flags \|= XFRM_DEV_RESUME;
3dca3f38 SK	106
	107	x->outer_mode->xmit(x, skb2);
	108
	109	err = x->type_offload->xmit(x, skb2, esp_features);
f53c7239 SK	110	if (!err) {
	111	skb2->next = nskb;
	112	} else if (err != -EINPROGRESS) {
3dca3f38 SK	113	XFRM_INC_STATS(xs_net(x), LINUX_MIB_XFRMOUTSTATEPROTOERROR);
	114	skb2->next = nskb;
	115	kfree_skb_list(skb2);
	116	return NULL;
f53c7239 SK	117	} else {
	118	if (skb == skb2)
	119	skb = nskb;
	120
	121	if (!skb)
	122	return NULL;
3dca3f38	123
f53c7239 SK	124	goto skip_push;
f53c7239 SK	125	}
3dca3f38 SK	126
	127	skb_push(skb2, skb2->data - skb_mac_header(skb2));
	128
f53c7239	129	skip_push:
3dca3f38 SK	130	skb2 = nskb;
	131	} while (skb2);
	132
	133	return skb;
f6e27114 SK	134	}
	135	EXPORT_SYMBOL_GPL(validate_xmit_xfrm);
	136
d77e38e6 SK	137	int xfrm_dev_state_add(struct net net, struct xfrm_state x,
	138	struct xfrm_user_offload *xuo)
	139	{
	140	int err;
	141	struct dst_entry *dst;
	142	struct net_device *dev;
	143	struct xfrm_state_offload *xso = &x->xso;
	144	xfrm_address_t *saddr;
	145	xfrm_address_t *daddr;
	146
	147	if (!x->type_offload)
ffdb5211	148	return -EINVAL;
d77e38e6	149
50bd870a YE	150	/* We don't yet support UDP encapsulation and TFC padding. */
50bd870a YE	151	if (x->encap \|\| x->tfcpad)
43024b9c	152	return -EINVAL;
d77e38e6 SK	153
	154	dev = dev_get_by_index(net, xuo->ifindex);
	155	if (!dev) {
	156	if (!(xuo->flags & XFRM_OFFLOAD_INBOUND)) {
	157	saddr = &x->props.saddr;
	158	daddr = &x->id.daddr;
	159	} else {
	160	saddr = &x->id.daddr;
	161	daddr = &x->props.saddr;
	162	}
	163
077fbac4	164	dst = __xfrm_dst_lookup(net, 0, 0, saddr, daddr,
9b42c1f1 SK	165	x->props.family,
9b42c1f1 SK	166	xfrm_smark_get(0, x));
d77e38e6 SK	167	if (IS_ERR(dst))
	168	return 0;
	169
	170	dev = dst->dev;
	171
	172	dev_hold(dev);
	173	dst_release(dst);
	174	}
	175
	176	if (!dev->xfrmdev_ops \|\| !dev->xfrmdev_ops->xdo_dev_state_add) {
67a63387	177	xso->dev = NULL;
d77e38e6 SK	178	dev_put(dev);
	179	return 0;
	180	}
	181
50bd870a YE	182	if (x->props.flags & XFRM_STATE_ESN &&
	183	!dev->xfrmdev_ops->xdo_dev_state_advance_esn) {
	184	xso->dev = NULL;
	185	dev_put(dev);
	186	return -EINVAL;
	187	}
	188
d77e38e6 SK	189	xso->dev = dev;
	190	xso->num_exthdrs = 1;
	191	xso->flags = xuo->flags;
	192
	193	err = dev->xfrmdev_ops->xdo_dev_state_add(x);
	194	if (err) {
aa5dd6fa	195	xso->dev = NULL;
d77e38e6 SK	196	dev_put(dev);
	197	return err;
	198	}
	199
	200	return 0;
	201	}
	202	EXPORT_SYMBOL_GPL(xfrm_dev_state_add);
	203
	204	bool xfrm_dev_offload_ok(struct sk_buff skb, struct xfrm_state x)
	205	{
	206	int mtu;
	207	struct dst_entry *dst = skb_dst(skb);
	208	struct xfrm_dst xdst = (struct xfrm_dst )dst;
	209	struct net_device *dev = x->xso.dev;
	210
	211	if (!x->type_offload \|\| x->encap)
	212	return false;
	213
fcb662de SN	214	if ((!dev \|\| (dev == xfrm_dst_path(dst)->dev)) &&
fcb662de SN	215	(!xdst->child->xfrm && x->type->get_mtu)) {
d77e38e6 SK	216	mtu = x->type->get_mtu(x, xdst->child_mtu_cached);
	217
	218	if (skb->len <= mtu)
	219	goto ok;
	220
779b7931	221	if (skb_is_gso(skb) && skb_gso_validate_network_len(skb, mtu))
d77e38e6 SK	222	goto ok;
	223	}
	224
	225	return false;
	226
	227	ok:
	228	if (dev && dev->xfrmdev_ops && dev->xfrmdev_ops->xdo_dev_offload_ok)
	229	return x->xso.dev->xfrmdev_ops->xdo_dev_offload_ok(skb, x);
	230
	231	return true;
	232	}
	233	EXPORT_SYMBOL_GPL(xfrm_dev_offload_ok);
f53c7239 SK	234
	235	void xfrm_dev_resume(struct sk_buff *skb)
	236	{
	237	struct net_device *dev = skb->dev;
	238	int ret = NETDEV_TX_BUSY;
	239	struct netdev_queue *txq;
	240	struct softnet_data *sd;
	241	unsigned long flags;
	242
	243	rcu_read_lock();
	244	txq = netdev_pick_tx(dev, skb, NULL);
	245
	246	HARD_TX_LOCK(dev, txq, smp_processor_id());
	247	if (!netif_xmit_frozen_or_stopped(txq))
	248	skb = dev_hard_start_xmit(skb, dev, txq, &ret);
	249	HARD_TX_UNLOCK(dev, txq);
	250
	251	if (!dev_xmit_complete(ret)) {
	252	local_irq_save(flags);
	253	sd = this_cpu_ptr(&softnet_data);
	254	skb_queue_tail(&sd->xfrm_backlog, skb);
	255	raise_softirq_irqoff(NET_TX_SOFTIRQ);
	256	local_irq_restore(flags);
	257	}
	258	rcu_read_unlock();
	259	}
	260	EXPORT_SYMBOL_GPL(xfrm_dev_resume);
	261
	262	void xfrm_dev_backlog(struct softnet_data *sd)
	263	{
	264	struct sk_buff_head *xfrm_backlog = &sd->xfrm_backlog;
	265	struct sk_buff_head list;
	266	struct sk_buff *skb;
	267
	268	if (skb_queue_empty(xfrm_backlog))
	269	return;
	270
	271	__skb_queue_head_init(&list);
	272
	273	spin_lock(&xfrm_backlog->lock);
	274	skb_queue_splice_init(xfrm_backlog, &list);
	275	spin_unlock(&xfrm_backlog->lock);
	276
	277	while (!skb_queue_empty(&list)) {
	278	skb = __skb_dequeue(&list);
	279	xfrm_dev_resume(skb);
	280	}
	281
	282	}
b81f884a	283	#endif
d77e38e6	284
92a23206	285	static int xfrm_api_check(struct net_device *dev)
d77e38e6	286	{
92a23206	287	#ifdef CONFIG_XFRM_OFFLOAD
d77e38e6 SK	288	if ((dev->features & NETIF_F_HW_ESP_TX_CSUM) &&
	289	!(dev->features & NETIF_F_HW_ESP))
	290	return NOTIFY_BAD;
	291
92a23206 SN	292	if ((dev->features & NETIF_F_HW_ESP) &&
	293	(!(dev->xfrmdev_ops &&
	294	dev->xfrmdev_ops->xdo_dev_state_add &&
	295	dev->xfrmdev_ops->xdo_dev_state_delete)))
	296	return NOTIFY_BAD;
	297	#else
	298	if (dev->features & (NETIF_F_HW_ESP \| NETIF_F_HW_ESP_TX_CSUM))
	299	return NOTIFY_BAD;
	300	#endif
	301
d77e38e6 SK	302	return NOTIFY_DONE;
	303	}
	304
92a23206 SN	305	static int xfrm_dev_register(struct net_device *dev)
	306	{
	307	return xfrm_api_check(dev);
	308	}
	309
d77e38e6 SK	310	static int xfrm_dev_feat_change(struct net_device *dev)
d77e38e6 SK	311	{
92a23206	312	return xfrm_api_check(dev);
d77e38e6 SK	313	}
	314
	315	static int xfrm_dev_down(struct net_device *dev)
	316	{
2c1497bb	317	if (dev->features & NETIF_F_HW_ESP)
d77e38e6 SK	318	xfrm_dev_state_flush(dev_net(dev), dev, true);
d77e38e6 SK	319
d77e38e6 SK	320	return NOTIFY_DONE;
	321	}
	322
21f42cc9 SK	323	static int xfrm_dev_event(struct notifier_block this, unsigned long event, void ptr)
	324	{
	325	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
	326
	327	switch (event) {
d77e38e6 SK	328	case NETDEV_REGISTER:
	329	return xfrm_dev_register(dev);
	330
d77e38e6 SK	331	case NETDEV_FEAT_CHANGE:
	332	return xfrm_dev_feat_change(dev);
	333
21f42cc9	334	case NETDEV_DOWN:
d77e38e6	335	return xfrm_dev_down(dev);
21f42cc9 SK	336	}
	337	return NOTIFY_DONE;
	338	}
	339
	340	static struct notifier_block xfrm_dev_notifier = {
	341	.notifier_call = xfrm_dev_event,
	342	};
	343
e9a441b6	344	void __init xfrm_dev_init(void)
21f42cc9 SK	345	{
	346	register_netdevice_notifier(&xfrm_dev_notifier);
	347	}