X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=net%2Fipv6%2Fipcomp6.c;h=511730b67e97fb0ba1f9d167c07e120ea3b34f6a;hb=refs%2Fheads%2Fvserver;hp=b357db4cc84bc2d2d0d4680844879edcbbfdc609;hpb=9213980e6a70d8473e0ffd4b39ab5b6caaba9ff5;p=linux-2.6.git

diff --git a/net/ipv6/ipcomp6.c b/net/ipv6/ipcomp6.c
index b357db4cc..511730b67 100644
--- a/net/ipv6/ipcomp6.c
+++ b/net/ipv6/ipcomp6.c
@@ -30,240 +30,361 @@
  *  The decompression of IP datagram MUST be done after the reassembly, 
  *  AH/ESP processing.
  */
-#include <linux/config.h>
 #include <linux/module.h>
-#include <net/inet_ecn.h>
 #include <net/ip.h>
 #include <net/xfrm.h>
 #include <net/ipcomp.h>
 #include <asm/scatterlist.h>
+#include <asm/semaphore.h>
 #include <linux/crypto.h>
 #include <linux/pfkeyv2.h>
 #include <linux/random.h>
+#include <linux/percpu.h>
+#include <linux/smp.h>
+#include <linux/list.h>
+#include <linux/vmalloc.h>
+#include <linux/rtnetlink.h>
 #include <net/icmp.h>
 #include <net/ipv6.h>
+#include <net/protocol.h>
 #include <linux/ipv6.h>
 #include <linux/icmpv6.h>
+#include <linux/mutex.h>
 
-static int ipcomp6_input(struct xfrm_state *x, struct xfrm_decap_state *decap, struct sk_buff *skb)
+struct ipcomp6_tfms {
+	struct list_head list;
+	struct crypto_comp **tfms;
+	int users;
+};
+
+static DEFINE_MUTEX(ipcomp6_resource_mutex);
+static void **ipcomp6_scratches;
+static int ipcomp6_scratch_users;
+static LIST_HEAD(ipcomp6_tfms_list);
+
+static int ipcomp6_input(struct xfrm_state *x, struct sk_buff *skb)
 {
-	int err = 0;
-	u8 nexthdr = 0;
-	u8 *prevhdr;
-	int hdr_len = skb->h.raw - skb->nh.raw;
-	unsigned char *tmp_hdr = NULL;
+	int err = -ENOMEM;
 	struct ipv6hdr *iph;
+	struct ipv6_comp_hdr *ipch;
 	int plen, dlen;
 	struct ipcomp_data *ipcd = x->data;
-	u8 *start, *scratch = ipcd->scratch;
+	u8 *start, *scratch;
+	struct crypto_comp *tfm;
+	int cpu;
 
-	if ((skb_is_nonlinear(skb) || skb_cloned(skb)) &&
-		skb_linearize(skb, GFP_ATOMIC) != 0) {
-		err = -ENOMEM;
+	if (skb_linearize_cow(skb))
 		goto out;
-	}
 
 	skb->ip_summed = CHECKSUM_NONE;
 
 	/* Remove ipcomp header and decompress original payload */
 	iph = skb->nh.ipv6h;
-	tmp_hdr = kmalloc(hdr_len, GFP_ATOMIC);
-	if (!tmp_hdr)
-		goto out;
-	memcpy(tmp_hdr, iph, hdr_len);
-	nexthdr = *(u8 *)skb->data;
-	skb_pull(skb, sizeof(struct ipv6_comp_hdr)); 
-	skb->nh.raw += sizeof(struct ipv6_comp_hdr);
-	memcpy(skb->nh.raw, tmp_hdr, hdr_len);
-	iph = skb->nh.ipv6h;
-	iph->payload_len = htons(ntohs(iph->payload_len) - sizeof(struct ipv6_comp_hdr));
-	skb->h.raw = skb->data;
+	ipch = (void *)skb->data;
+	skb->h.raw = skb->nh.raw + sizeof(*ipch);
+	__skb_pull(skb, sizeof(*ipch));
 
 	/* decompression */
 	plen = skb->len;
 	dlen = IPCOMP_SCRATCH_SIZE;
 	start = skb->data;
 
-	err = crypto_comp_decompress(ipcd->tfm, start, plen, scratch, &dlen);
+	cpu = get_cpu();
+	scratch = *per_cpu_ptr(ipcomp6_scratches, cpu);
+	tfm = *per_cpu_ptr(ipcd->tfms, cpu);
+
+	err = crypto_comp_decompress(tfm, start, plen, scratch, &dlen);
 	if (err) {
 		err = -EINVAL;
-		goto out;
+		goto out_put_cpu;
 	}
 
 	if (dlen < (plen + sizeof(struct ipv6_comp_hdr))) {
 		err = -EINVAL;
-		goto out;
+		goto out_put_cpu;
 	}
 
 	err = pskb_expand_head(skb, 0, dlen - plen, GFP_ATOMIC);
 	if (err) {
-		goto out;
+		goto out_put_cpu;
 	}
 
-	skb_put(skb, dlen - plen);
+	skb->truesize += dlen - plen;
+	__skb_put(skb, dlen - plen);
 	memcpy(skb->data, scratch, dlen);
+	err = ipch->nexthdr;
 
-	iph = skb->nh.ipv6h;
-	iph->payload_len = htons(skb->len);
-	
-	ip6_find_1stfragopt(skb, &prevhdr);
-	*prevhdr = nexthdr;
+out_put_cpu:
+	put_cpu();
 out:
-	if (tmp_hdr)
-		kfree(tmp_hdr);
-	if (err)
-		goto error_out;
-	return nexthdr;
-error_out:
 	return err;
 }
 
-static int ipcomp6_output(struct sk_buff **pskb)
+static int ipcomp6_output(struct xfrm_state *x, struct sk_buff *skb)
 {
 	int err;
-	struct dst_entry *dst = (*pskb)->dst;
-	struct xfrm_state *x = dst->xfrm;
-	struct ipv6hdr *tmp_iph = NULL, *iph, *top_iph;
-	int hdr_len = 0;
+	struct ipv6hdr *top_iph;
+	int hdr_len;
 	struct ipv6_comp_hdr *ipch;
 	struct ipcomp_data *ipcd = x->data;
-	u8 *prevhdr;
-	u8 nexthdr = 0;
 	int plen, dlen;
-	u8 *start, *scratch = ipcd->scratch;
+	u8 *start, *scratch;
+	struct crypto_comp *tfm;
+	int cpu;
 
-	if ((*pskb)->ip_summed == CHECKSUM_HW) {
-		err = skb_checksum_help(pskb, 0);
-		if (err)
-			goto error_nolock;
-	}
-
-	spin_lock_bh(&x->lock);
-
-	err = xfrm_check_output(x, *pskb, AF_INET6);
-	if (err)
-		goto error;
-
-	if (x->props.mode) {
-		hdr_len = sizeof(struct ipv6hdr);
-		nexthdr = IPPROTO_IPV6;
-		iph = (*pskb)->nh.ipv6h;
-		top_iph = (struct ipv6hdr *)skb_push(*pskb, sizeof(struct ipv6hdr));
-		top_iph->version = 6;
-		top_iph->priority = iph->priority;
-		top_iph->flow_lbl[0] = iph->flow_lbl[0];
-		top_iph->flow_lbl[1] = iph->flow_lbl[1];
-		top_iph->flow_lbl[2] = iph->flow_lbl[2];
-		top_iph->nexthdr = IPPROTO_IPV6; /* initial */
-		top_iph->payload_len = htons((*pskb)->len - sizeof(struct ipv6hdr));
-		top_iph->hop_limit = iph->hop_limit;
-		memcpy(&top_iph->saddr, (struct in6_addr *)&x->props.saddr, sizeof(struct in6_addr));
-		memcpy(&top_iph->daddr, (struct in6_addr *)&x->id.daddr, sizeof(struct in6_addr));
-		(*pskb)->nh.raw = (*pskb)->data; /* == top_iph */
-		(*pskb)->h.raw = (*pskb)->nh.raw + hdr_len;
-	} else {
-		hdr_len = ip6_find_1stfragopt(*pskb, &prevhdr);
-		nexthdr = *prevhdr;
-	}
+	hdr_len = skb->h.raw - skb->data;
 
 	/* check whether datagram len is larger than threshold */
-	if (((*pskb)->len - hdr_len) < ipcd->threshold) {
+	if ((skb->len - hdr_len) < ipcd->threshold) {
 		goto out_ok;
 	}
 
-	if ((skb_is_nonlinear(*pskb) || skb_cloned(*pskb)) &&
-		skb_linearize(*pskb, GFP_ATOMIC) != 0) {
-		err = -ENOMEM;
-		goto error;
-	}
+	if (skb_linearize_cow(skb))
+		goto out_ok;
 
 	/* compression */
-	plen = (*pskb)->len - hdr_len;
+	plen = skb->len - hdr_len;
 	dlen = IPCOMP_SCRATCH_SIZE;
-	start = (*pskb)->data + hdr_len;
+	start = skb->h.raw;
 
-	err = crypto_comp_compress(ipcd->tfm, start, plen, scratch, &dlen);
-	if (err) {
-		goto error;
-	}
-	if ((dlen + sizeof(struct ipv6_comp_hdr)) >= plen) {
+	cpu = get_cpu();
+	scratch = *per_cpu_ptr(ipcomp6_scratches, cpu);
+	tfm = *per_cpu_ptr(ipcd->tfms, cpu);
+
+	err = crypto_comp_compress(tfm, start, plen, scratch, &dlen);
+	if (err || (dlen + sizeof(struct ipv6_comp_hdr)) >= plen) {
+		put_cpu();
 		goto out_ok;
 	}
-	memcpy(start, scratch, dlen);
-	pskb_trim(*pskb, hdr_len+dlen);
+	memcpy(start + sizeof(struct ip_comp_hdr), scratch, dlen);
+	put_cpu();
+	pskb_trim(skb, hdr_len + dlen + sizeof(struct ip_comp_hdr));
 
 	/* insert ipcomp header and replace datagram */
-	tmp_iph = kmalloc(hdr_len, GFP_ATOMIC);
-	if (!tmp_iph) {
-		err = -ENOMEM;
-		goto error;
-	}
-	memcpy(tmp_iph, (*pskb)->nh.raw, hdr_len);
-	top_iph = (struct ipv6hdr*)skb_push(*pskb, sizeof(struct ipv6_comp_hdr));
-	memcpy(top_iph, tmp_iph, hdr_len);
-	kfree(tmp_iph);
-
-	if (x->props.mode && (x->props.flags & XFRM_STATE_NOECN))
-		IP6_ECN_clear(top_iph);
-	top_iph->payload_len = htons((*pskb)->len - sizeof(struct ipv6hdr));
-	(*pskb)->nh.raw = (*pskb)->data; /* top_iph */
-	ip6_find_1stfragopt(*pskb, &prevhdr); 
-	*prevhdr = IPPROTO_COMP;
-
-	ipch = (struct ipv6_comp_hdr *)((unsigned char *)top_iph + hdr_len);
-	ipch->nexthdr = nexthdr;
+	top_iph = (struct ipv6hdr *)skb->data;
+
+	top_iph->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
+
+	ipch = (struct ipv6_comp_hdr *)start;
+	ipch->nexthdr = *skb->nh.raw;
 	ipch->flags = 0;
 	ipch->cpi = htons((u16 )ntohl(x->id.spi));
+	*skb->nh.raw = IPPROTO_COMP;
 
-	(*pskb)->h.raw = (unsigned char*)ipch;
 out_ok:
-	x->curlft.bytes += (*pskb)->len;
-	x->curlft.packets++;
-	spin_unlock_bh(&x->lock);
-
-	if (((*pskb)->dst = dst_pop(dst)) == NULL) {
-		err = -EHOSTUNREACH;
-		goto error_nolock;
-	}
-	err = NET_XMIT_BYPASS;
-
-out_exit:
-	return err;
-error:
-	spin_unlock_bh(&x->lock);
-error_nolock:
-	kfree_skb(*pskb);
-	goto out_exit;
+	return 0;
 }
 
 static void ipcomp6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
-		                int type, int code, int offset, __u32 info)
+		                int type, int code, int offset, __be32 info)
 {
-	u32 spi;
+	__be32 spi;
 	struct ipv6hdr *iph = (struct ipv6hdr*)skb->data;
 	struct ipv6_comp_hdr *ipcomph = (struct ipv6_comp_hdr*)(skb->data+offset);
 	struct xfrm_state *x;
 
-	if (type != ICMPV6_DEST_UNREACH || type != ICMPV6_PKT_TOOBIG)
+	if (type != ICMPV6_DEST_UNREACH && type != ICMPV6_PKT_TOOBIG)
 		return;
 
-	spi = ntohl(ntohs(ipcomph->cpi));
+	spi = htonl(ntohs(ipcomph->cpi));
 	x = xfrm_state_lookup((xfrm_address_t *)&iph->daddr, spi, IPPROTO_COMP, AF_INET6);
 	if (!x)
 		return;
 
-	printk(KERN_DEBUG "pmtu discovery on SA IPCOMP/%08x/"
-			"%04x:%04x:%04x:%04x:%04x:%04x:%04x:%04x\n",
+	printk(KERN_DEBUG "pmtu discovery on SA IPCOMP/%08x/" NIP6_FMT "\n",
 			spi, NIP6(iph->daddr));
 	xfrm_state_put(x);
 }
 
+static struct xfrm_state *ipcomp6_tunnel_create(struct xfrm_state *x)
+{
+	struct xfrm_state *t = NULL;
+	u8 mode = XFRM_MODE_TUNNEL;
+
+	t = xfrm_state_alloc();
+	if (!t)
+		goto out;
+
+	t->id.proto = IPPROTO_IPV6;
+	t->id.spi = xfrm6_tunnel_alloc_spi((xfrm_address_t *)&x->props.saddr);
+	if (!t->id.spi)
+		goto error;
+
+	memcpy(t->id.daddr.a6, x->id.daddr.a6, sizeof(struct in6_addr));
+	memcpy(&t->sel, &x->sel, sizeof(t->sel));
+	t->props.family = AF_INET6;
+	if (x->props.mode == XFRM_MODE_BEET)
+		mode = x->props.mode;
+	t->props.mode = mode;
+	memcpy(t->props.saddr.a6, x->props.saddr.a6, sizeof(struct in6_addr));
+
+	if (xfrm_init_state(t))
+		goto error;
+
+	atomic_set(&t->tunnel_users, 1);
+
+out:
+	return t;
+
+error:
+	t->km.state = XFRM_STATE_DEAD;
+	xfrm_state_put(t);
+	t = NULL;
+	goto out;
+}
+
+static int ipcomp6_tunnel_attach(struct xfrm_state *x)
+{
+	int err = 0;
+	struct xfrm_state *t = NULL;
+	__be32 spi;
+
+	spi = xfrm6_tunnel_spi_lookup((xfrm_address_t *)&x->props.saddr);
+	if (spi)
+		t = xfrm_state_lookup((xfrm_address_t *)&x->id.daddr,
+					      spi, IPPROTO_IPV6, AF_INET6);
+	if (!t) {
+		t = ipcomp6_tunnel_create(x);
+		if (!t) {
+			err = -EINVAL;
+			goto out;
+		}
+		xfrm_state_insert(t);
+		xfrm_state_hold(t);
+	}
+	x->tunnel = t;
+	atomic_inc(&t->tunnel_users);
+
+out:
+	return err;
+}
+
+static void ipcomp6_free_scratches(void)
+{
+	int i;
+	void **scratches;
+
+	if (--ipcomp6_scratch_users)
+		return;
+
+	scratches = ipcomp6_scratches;
+	if (!scratches)
+		return;
+
+	for_each_possible_cpu(i) {
+		void *scratch = *per_cpu_ptr(scratches, i);
+
+		vfree(scratch);
+	}
+
+	free_percpu(scratches);
+}
+
+static void **ipcomp6_alloc_scratches(void)
+{
+	int i;
+	void **scratches;
+
+	if (ipcomp6_scratch_users++)
+		return ipcomp6_scratches;
+
+	scratches = alloc_percpu(void *);
+	if (!scratches)
+		return NULL;
+
+	ipcomp6_scratches = scratches;
+
+	for_each_possible_cpu(i) {
+		void *scratch = vmalloc(IPCOMP_SCRATCH_SIZE);
+		if (!scratch)
+			return NULL;
+		*per_cpu_ptr(scratches, i) = scratch;
+	}
+
+	return scratches;
+}
+
+static void ipcomp6_free_tfms(struct crypto_comp **tfms)
+{
+	struct ipcomp6_tfms *pos;
+	int cpu;
+
+	list_for_each_entry(pos, &ipcomp6_tfms_list, list) {
+		if (pos->tfms == tfms)
+			break;
+	}
+
+	BUG_TRAP(pos);
+
+	if (--pos->users)
+		return;
+
+	list_del(&pos->list);
+	kfree(pos);
+
+	if (!tfms)
+		return;
+
+	for_each_possible_cpu(cpu) {
+		struct crypto_comp *tfm = *per_cpu_ptr(tfms, cpu);
+		crypto_free_comp(tfm);
+	}
+	free_percpu(tfms);
+}
+
+static struct crypto_comp **ipcomp6_alloc_tfms(const char *alg_name)
+{
+	struct ipcomp6_tfms *pos;
+	struct crypto_comp **tfms;
+	int cpu;
+
+	/* This can be any valid CPU ID so we don't need locking. */
+	cpu = raw_smp_processor_id();
+
+	list_for_each_entry(pos, &ipcomp6_tfms_list, list) {
+		struct crypto_comp *tfm;
+
+		tfms = pos->tfms;
+		tfm = *per_cpu_ptr(tfms, cpu);
+
+		if (!strcmp(crypto_comp_name(tfm), alg_name)) {
+			pos->users++;
+			return tfms;
+		}
+	}
+
+	pos = kmalloc(sizeof(*pos), GFP_KERNEL);
+	if (!pos)
+		return NULL;
+
+	pos->users = 1;
+	INIT_LIST_HEAD(&pos->list);
+	list_add(&pos->list, &ipcomp6_tfms_list);
+
+	pos->tfms = tfms = alloc_percpu(struct crypto_comp *);
+	if (!tfms)
+		goto error;
+
+	for_each_possible_cpu(cpu) {
+		struct crypto_comp *tfm = crypto_alloc_comp(alg_name, 0,
+							    CRYPTO_ALG_ASYNC);
+		if (!tfm)
+			goto error;
+		*per_cpu_ptr(tfms, cpu) = tfm;
+	}
+
+	return tfms;
+
+error:
+	ipcomp6_free_tfms(tfms);
+	return NULL;
+}
+
 static void ipcomp6_free_data(struct ipcomp_data *ipcd)
 {
-	if (ipcd->tfm)
-		crypto_free_tfm(ipcd->tfm);
-	if (ipcd->scratch)
-		kfree(ipcd->scratch);
+	if (ipcd->tfms)
+		ipcomp6_free_tfms(ipcd->tfms);
+	ipcomp6_free_scratches();
 }
 
 static void ipcomp6_destroy(struct xfrm_state *x)
@@ -271,11 +392,16 @@ static void ipcomp6_destroy(struct xfrm_state *x)
 	struct ipcomp_data *ipcd = x->data;
 	if (!ipcd)
 		return;
+	xfrm_state_delete_tunnel(x);
+	mutex_lock(&ipcomp6_resource_mutex);
 	ipcomp6_free_data(ipcd);
+	mutex_unlock(&ipcomp6_resource_mutex);
 	kfree(ipcd);
+
+	xfrm6_tunnel_free_spi((xfrm_address_t *)&x->props.saddr);
 }
 
-static int ipcomp6_init_state(struct xfrm_state *x, void *args)
+static int ipcomp6_init_state(struct xfrm_state *x)
 {
 	int err;
 	struct ipcomp_data *ipcd;
@@ -285,36 +411,46 @@ static int ipcomp6_init_state(struct xfrm_state *x, void *args)
 	if (!x->calg)
 		goto out;
 
+	if (x->encap)
+		goto out;
+
 	err = -ENOMEM;
-	ipcd = kmalloc(sizeof(*ipcd), GFP_KERNEL);
+	ipcd = kzalloc(sizeof(*ipcd), GFP_KERNEL);
 	if (!ipcd)
-		goto error;
+		goto out;
 
-	memset(ipcd, 0, sizeof(*ipcd));
-	x->props.header_len = sizeof(struct ipv6_comp_hdr);
-	if (x->props.mode)
+	x->props.header_len = 0;
+	if (x->props.mode == XFRM_MODE_TUNNEL)
 		x->props.header_len += sizeof(struct ipv6hdr);
 	
-	ipcd->scratch = kmalloc(IPCOMP_SCRATCH_SIZE, GFP_KERNEL);
-	if (!ipcd->scratch)
+	mutex_lock(&ipcomp6_resource_mutex);
+	if (!ipcomp6_alloc_scratches())
 		goto error;
 
-	ipcd->tfm = crypto_alloc_tfm(x->calg->alg_name, 0);
-	if (!ipcd->tfm)
+	ipcd->tfms = ipcomp6_alloc_tfms(x->calg->alg_name);
+	if (!ipcd->tfms)
 		goto error;
+	mutex_unlock(&ipcomp6_resource_mutex);
+
+	if (x->props.mode == XFRM_MODE_TUNNEL) {
+		err = ipcomp6_tunnel_attach(x);
+		if (err)
+			goto error_tunnel;
+	}
 
-	calg_desc = xfrm_calg_get_byname(x->calg->alg_name);
+	calg_desc = xfrm_calg_get_byname(x->calg->alg_name, 0);
 	BUG_ON(!calg_desc);
 	ipcd->threshold = calg_desc->uinfo.comp.threshold;
 	x->data = ipcd;
 	err = 0;
 out:
 	return err;
+error_tunnel:
+	mutex_lock(&ipcomp6_resource_mutex);
 error:
-	if (ipcd) {
-		ipcomp6_free_data(ipcd);
-		kfree(ipcd);
-	}
+	ipcomp6_free_data(ipcd);
+	mutex_unlock(&ipcomp6_resource_mutex);
+	kfree(ipcd);
 
 	goto out;
 }
@@ -328,6 +464,7 @@ static struct xfrm_type ipcomp6_type =
 	.destructor	= ipcomp6_destroy,
 	.input		= ipcomp6_input,
 	.output		= ipcomp6_output,
+	.hdr_offset	= xfrm6_find_1stfragopt,
 };
 
 static struct inet6_protocol ipcomp6_protocol =