[NET]: Make device event notification network namespace safe
[safe/jmp/linux-2.6] / net / ipv4 / ipmr.c
index 51528d3..0365988 100644 (file)
@@ -49,6 +49,7 @@
 #include <linux/mroute.h>
 #include <linux/init.h>
 #include <linux/if_ether.h>
+#include <net/net_namespace.h>
 #include <net/ip.h>
 #include <net/protocol.h>
 #include <linux/skbuff.h>
@@ -62,6 +63,7 @@
 #include <linux/netfilter_ipv4.h>
 #include <net/ipip.h>
 #include <net/checksum.h>
+#include <net/netlink.h>
 
 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
 #define CONFIG_IP_PIMSM        1
@@ -151,9 +153,11 @@ struct net_device *ipmr_new_tunnel(struct vifctl *v)
                        dev->flags |= IFF_MULTICAST;
 
                        in_dev = __in_dev_get_rtnl(dev);
-                       if (in_dev == NULL && (in_dev = inetdev_init(dev)) == NULL)
+                       if (in_dev == NULL)
                                goto failure;
-                       in_dev->cnf.rp_filter = 0;
+
+                       ipv4_devconf_setall(in_dev);
+                       IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
 
                        if (dev_open(dev))
                                goto failure;
@@ -217,10 +221,15 @@ static struct net_device *ipmr_reg_vif(void)
        }
        dev->iflink = 0;
 
-       if ((in_dev = inetdev_init(dev)) == NULL)
+       rcu_read_lock();
+       if ((in_dev = __in_dev_get_rcu(dev)) == NULL) {
+               rcu_read_unlock();
                goto failure;
+       }
 
-       in_dev->cnf.rp_filter = 0;
+       ipv4_devconf_setall(in_dev);
+       IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
+       rcu_read_unlock();
 
        if (dev_open(dev))
                goto failure;
@@ -280,7 +289,7 @@ static int vif_delete(int vifi)
        dev_set_allmulti(dev, -1);
 
        if ((in_dev = __in_dev_get_rtnl(dev)) != NULL) {
-               in_dev->cnf.mc_forwarding--;
+               IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
                ip_rt_multicast_event(in_dev);
        }
 
@@ -303,7 +312,7 @@ static void ipmr_destroy_unres(struct mfc_cache *c)
        atomic_dec(&cache_resolve_queue_len);
 
        while ((skb=skb_dequeue(&c->mfc_un.unres.unresolved))) {
-               if (skb->nh.iph->version == 0) {
+               if (ip_hdr(skb)->version == 0) {
                        struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));
                        nlh->nlmsg_type = NLMSG_ERROR;
                        nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr));
@@ -425,7 +434,7 @@ static int vif_add(struct vifctl *vifc, int mrtsock)
 
        if ((in_dev = __in_dev_get_rtnl(dev)) == NULL)
                return -EADDRNOTAVAIL;
-       in_dev->cnf.mc_forwarding++;
+       IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
        dev_set_allmulti(dev, +1);
        ip_rt_multicast_event(in_dev);
 
@@ -509,11 +518,12 @@ static void ipmr_cache_resolve(struct mfc_cache *uc, struct mfc_cache *c)
         */
 
        while ((skb=__skb_dequeue(&uc->mfc_un.unres.unresolved))) {
-               if (skb->nh.iph->version == 0) {
+               if (ip_hdr(skb)->version == 0) {
                        struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));
 
                        if (ipmr_fill_mroute(skb, c, NLMSG_DATA(nlh)) > 0) {
-                               nlh->nlmsg_len = skb->tail - (u8*)nlh;
+                               nlh->nlmsg_len = (skb_tail_pointer(skb) -
+                                                 (u8 *)nlh);
                        } else {
                                nlh->nlmsg_type = NLMSG_ERROR;
                                nlh->nlmsg_len = NLMSG_LENGTH(sizeof(struct nlmsgerr));
@@ -539,7 +549,7 @@ static void ipmr_cache_resolve(struct mfc_cache *uc, struct mfc_cache *c)
 static int ipmr_cache_report(struct sk_buff *pkt, vifi_t vifi, int assert)
 {
        struct sk_buff *skb;
-       int ihl = pkt->nh.iph->ihl<<2;
+       const int ihl = ip_hdrlen(pkt);
        struct igmphdr *igmp;
        struct igmpmsg *msg;
        int ret;
@@ -561,14 +571,17 @@ static int ipmr_cache_report(struct sk_buff *pkt, vifi_t vifi, int assert)
                   And all this only to mangle msg->im_msgtype and
                   to set msg->im_mbz to "mbz" :-)
                 */
-               msg = (struct igmpmsg*)skb_push(skb, sizeof(struct iphdr));
-               skb->nh.raw = skb->h.raw = (u8*)msg;
-               memcpy(msg, pkt->nh.raw, sizeof(struct iphdr));
+               skb_push(skb, sizeof(struct iphdr));
+               skb_reset_network_header(skb);
+               skb_reset_transport_header(skb);
+               msg = (struct igmpmsg *)skb_network_header(skb);
+               memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
                msg->im_msgtype = IGMPMSG_WHOLEPKT;
                msg->im_mbz = 0;
                msg->im_vif = reg_vif_num;
-               skb->nh.iph->ihl = sizeof(struct iphdr) >> 2;
-               skb->nh.iph->tot_len = htons(ntohs(pkt->nh.iph->tot_len) + sizeof(struct iphdr));
+               ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
+               ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
+                                            sizeof(struct iphdr));
        } else
 #endif
        {
@@ -577,10 +590,11 @@ static int ipmr_cache_report(struct sk_buff *pkt, vifi_t vifi, int assert)
         *      Copy the IP header
         */
 
-       skb->nh.iph = (struct iphdr *)skb_put(skb, ihl);
-       memcpy(skb->data,pkt->data,ihl);
-       skb->nh.iph->protocol = 0;                      /* Flag to the kernel this is a route add */
-       msg = (struct igmpmsg*)skb->nh.iph;
+       skb->network_header = skb->tail;
+       skb_put(skb, ihl);
+       skb_copy_to_linear_data(skb, pkt->data, ihl);
+       ip_hdr(skb)->protocol = 0;                      /* Flag to the kernel this is a route add */
+       msg = (struct igmpmsg *)skb_network_header(skb);
        msg->im_vif = vifi;
        skb->dst = dst_clone(pkt->dst);
 
@@ -592,8 +606,8 @@ static int ipmr_cache_report(struct sk_buff *pkt, vifi_t vifi, int assert)
        igmp->type      =
        msg->im_msgtype = assert;
        igmp->code      =       0;
-       skb->nh.iph->tot_len=htons(skb->len);                   /* Fix the length */
-       skb->h.raw = skb->nh.raw;
+       ip_hdr(skb)->tot_len = htons(skb->len);                 /* Fix the length */
+       skb->transport_header = skb->network_header;
        }
 
        if (mroute_socket == NULL) {
@@ -622,11 +636,12 @@ ipmr_cache_unresolved(vifi_t vifi, struct sk_buff *skb)
 {
        int err;
        struct mfc_cache *c;
+       const struct iphdr *iph = ip_hdr(skb);
 
        spin_lock_bh(&mfc_unres_lock);
        for (c=mfc_unres_queue; c; c=c->next) {
-               if (c->mfc_mcastgrp == skb->nh.iph->daddr &&
-                   c->mfc_origin == skb->nh.iph->saddr)
+               if (c->mfc_mcastgrp == iph->daddr &&
+                   c->mfc_origin == iph->saddr)
                        break;
        }
 
@@ -646,9 +661,9 @@ ipmr_cache_unresolved(vifi_t vifi, struct sk_buff *skb)
                /*
                 *      Fill in the new cache entry
                 */
-               c->mfc_parent=-1;
-               c->mfc_origin=skb->nh.iph->saddr;
-               c->mfc_mcastgrp=skb->nh.iph->daddr;
+               c->mfc_parent   = -1;
+               c->mfc_origin   = iph->saddr;
+               c->mfc_mcastgrp = iph->daddr;
 
                /*
                 *      Reflect first query at mrouted.
@@ -834,7 +849,7 @@ static void mrtsock_destruct(struct sock *sk)
 {
        rtnl_lock();
        if (sk == mroute_socket) {
-               ipv4_devconf.mc_forwarding--;
+               IPV4_DEVCONF_ALL(MC_FORWARDING)--;
 
                write_lock_bh(&mrt_lock);
                mroute_socket=NULL;
@@ -883,7 +898,7 @@ int ip_mroute_setsockopt(struct sock *sk,int optname,char __user *optval,int opt
                        mroute_socket=sk;
                        write_unlock_bh(&mrt_lock);
 
-                       ipv4_devconf.mc_forwarding++;
+                       IPV4_DEVCONF_ALL(MC_FORWARDING)++;
                }
                rtnl_unlock();
                return ret;
@@ -1068,13 +1083,18 @@ int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg)
 
 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
 {
+       struct net_device *dev = ptr;
        struct vif_device *v;
        int ct;
+
+       if (dev->nd_net != &init_net)
+               return NOTIFY_DONE;
+
        if (event != NETDEV_UNREGISTER)
                return NOTIFY_DONE;
        v=&vif_table[0];
        for (ct=0;ct<maxvif;ct++,v++) {
-               if (v->dev==ptr)
+               if (v->dev==dev)
                        vif_delete(ct);
        }
        return NOTIFY_DONE;
@@ -1094,15 +1114,16 @@ static struct notifier_block ip_mr_notifier={
 static void ip_encap(struct sk_buff *skb, __be32 saddr, __be32 daddr)
 {
        struct iphdr *iph;
+       struct iphdr *old_iph = ip_hdr(skb);
 
        skb_push(skb, sizeof(struct iphdr));
-       skb->h.ipiph = skb->nh.iph;
+       skb->transport_header = skb->network_header;
        skb_reset_network_header(skb);
-       iph = skb->nh.iph;
+       iph = ip_hdr(skb);
 
        iph->version    =       4;
-       iph->tos        =       skb->nh.iph->tos;
-       iph->ttl        =       skb->nh.iph->ttl;
+       iph->tos        =       old_iph->tos;
+       iph->ttl        =       old_iph->ttl;
        iph->frag_off   =       0;
        iph->daddr      =       daddr;
        iph->saddr      =       saddr;
@@ -1134,7 +1155,7 @@ static inline int ipmr_forward_finish(struct sk_buff *skb)
 
 static void ipmr_queue_xmit(struct sk_buff *skb, struct mfc_cache *c, int vifi)
 {
-       struct iphdr *iph = skb->nh.iph;
+       const struct iphdr *iph = ip_hdr(skb);
        struct vif_device *vif = &vif_table[vifi];
        struct net_device *dev;
        struct rtable *rt;
@@ -1200,8 +1221,7 @@ static void ipmr_queue_xmit(struct sk_buff *skb, struct mfc_cache *c, int vifi)
 
        dst_release(skb->dst);
        skb->dst = &rt->u.dst;
-       iph = skb->nh.iph;
-       ip_decrease_ttl(iph);
+       ip_decrease_ttl(ip_hdr(skb));
 
        /* FIXME: forward and output firewalls used to be called here.
         * What do we do with netfilter? -- RR */
@@ -1301,7 +1321,7 @@ static int ip_mr_forward(struct sk_buff *skb, struct mfc_cache *cache, int local
         *      Forward the frame
         */
        for (ct = cache->mfc_un.res.maxvif-1; ct >= cache->mfc_un.res.minvif; ct--) {
-               if (skb->nh.iph->ttl > cache->mfc_un.res.ttls[ct]) {
+               if (ip_hdr(skb)->ttl > cache->mfc_un.res.ttls[ct]) {
                        if (psend != -1) {
                                struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
                                if (skb2)
@@ -1347,7 +1367,7 @@ int ip_mr_input(struct sk_buff *skb)
                    if (IPCB(skb)->opt.router_alert) {
                            if (ip_call_ra_chain(skb))
                                    return 0;
-                   } else if (skb->nh.iph->protocol == IPPROTO_IGMP){
+                   } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP){
                            /* IGMPv1 (and broken IGMPv2 implementations sort of
                               Cisco IOS <= 11.2(8)) do not put router alert
                               option to IGMP packets destined to routable
@@ -1366,7 +1386,7 @@ int ip_mr_input(struct sk_buff *skb)
        }
 
        read_lock(&mrt_lock);
-       cache = ipmr_cache_find(skb->nh.iph->saddr, skb->nh.iph->daddr);
+       cache = ipmr_cache_find(ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
 
        /*
         *      No usable cache entry
@@ -1426,14 +1446,15 @@ int pim_rcv_v1(struct sk_buff * skb)
        if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(*encap)))
                goto drop;
 
-       pim = (struct igmphdr*)skb->h.raw;
+       pim = igmp_hdr(skb);
 
        if (!mroute_do_pim ||
            skb->len < sizeof(*pim) + sizeof(*encap) ||
            pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
                goto drop;
 
-       encap = (struct iphdr*)(skb->h.raw + sizeof(struct igmphdr));
+       encap = (struct iphdr *)(skb_transport_header(skb) +
+                                sizeof(struct igmphdr));
        /*
           Check that:
           a. packet is really destinted to a multicast group
@@ -1455,7 +1476,7 @@ int pim_rcv_v1(struct sk_buff * skb)
        if (reg_dev == NULL)
                goto drop;
 
-       skb->mac.raw = skb->nh.raw;
+       skb->mac_header = skb->network_header;
        skb_pull(skb, (u8*)encap - skb->data);
        skb_reset_network_header(skb);
        skb->dev = reg_dev;
@@ -1486,7 +1507,7 @@ static int pim_rcv(struct sk_buff * skb)
        if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(*encap)))
                goto drop;
 
-       pim = (struct pimreghdr*)skb->h.raw;
+       pim = (struct pimreghdr *)skb_transport_header(skb);
        if (pim->type != ((PIM_VERSION<<4)|(PIM_REGISTER)) ||
            (pim->flags&PIM_NULL_REGISTER) ||
            (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
@@ -1494,7 +1515,8 @@ static int pim_rcv(struct sk_buff * skb)
                goto drop;
 
        /* check if the inner packet is destined to mcast group */
-       encap = (struct iphdr*)(skb->h.raw + sizeof(struct pimreghdr));
+       encap = (struct iphdr *)(skb_transport_header(skb) +
+                                sizeof(struct pimreghdr));
        if (!MULTICAST(encap->daddr) ||
            encap->tot_len == 0 ||
            ntohs(encap->tot_len) + sizeof(*pim) > skb->len)
@@ -1510,7 +1532,7 @@ static int pim_rcv(struct sk_buff * skb)
        if (reg_dev == NULL)
                goto drop;
 
-       skb->mac.raw = skb->nh.raw;
+       skb->mac_header = skb->network_header;
        skb_pull(skb, (u8*)encap - skb->data);
        skb_reset_network_header(skb);
        skb->dev = reg_dev;
@@ -1537,7 +1559,7 @@ ipmr_fill_mroute(struct sk_buff *skb, struct mfc_cache *c, struct rtmsg *rtm)
        int ct;
        struct rtnexthop *nhp;
        struct net_device *dev = vif_table[c->mfc_parent].dev;
-       u8 *b = skb->tail;
+       u8 *b = skb_tail_pointer(skb);
        struct rtattr *mp_head;
 
        if (dev)
@@ -1557,12 +1579,12 @@ ipmr_fill_mroute(struct sk_buff *skb, struct mfc_cache *c, struct rtmsg *rtm)
                }
        }
        mp_head->rta_type = RTA_MULTIPATH;
-       mp_head->rta_len = skb->tail - (u8*)mp_head;
+       mp_head->rta_len = skb_tail_pointer(skb) - (u8 *)mp_head;
        rtm->rtm_type = RTN_MULTICAST;
        return 1;
 
 rtattr_failure:
-       skb_trim(skb, b - skb->data);
+       nlmsg_trim(skb, b);
        return -EMSGSIZE;
 }
 
@@ -1577,6 +1599,7 @@ int ipmr_get_route(struct sk_buff *skb, struct rtmsg *rtm, int nowait)
 
        if (cache==NULL) {
                struct sk_buff *skb2;
+               struct iphdr *iph;
                struct net_device *dev;
                int vif;
 
@@ -1598,10 +1621,11 @@ int ipmr_get_route(struct sk_buff *skb, struct rtmsg *rtm, int nowait)
 
                skb_push(skb2, sizeof(struct iphdr));
                skb_reset_network_header(skb2);
-               skb2->nh.iph->ihl = sizeof(struct iphdr)>>2;
-               skb2->nh.iph->saddr = rt->rt_src;
-               skb2->nh.iph->daddr = rt->rt_dst;
-               skb2->nh.iph->version = 0;
+               iph = ip_hdr(skb2);
+               iph->ihl = sizeof(struct iphdr) >> 2;
+               iph->saddr = rt->rt_src;
+               iph->daddr = rt->rt_dst;
+               iph->version = 0;
                err = ipmr_cache_unresolved(vif, skb2);
                read_unlock(&mrt_lock);
                return err;
@@ -1681,7 +1705,7 @@ static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
        return 0;
 }
 
-static struct seq_operations ipmr_vif_seq_ops = {
+static const struct seq_operations ipmr_vif_seq_ops = {
        .start = ipmr_vif_seq_start,
        .next  = ipmr_vif_seq_next,
        .stop  = ipmr_vif_seq_stop,
@@ -1844,7 +1868,7 @@ static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
        return 0;
 }
 
-static struct seq_operations ipmr_mfc_seq_ops = {
+static const struct seq_operations ipmr_mfc_seq_ops = {
        .start = ipmr_mfc_seq_start,
        .next  = ipmr_mfc_seq_next,
        .stop  = ipmr_mfc_seq_stop,
@@ -1899,12 +1923,12 @@ void __init ip_mr_init(void)
        mrt_cachep = kmem_cache_create("ip_mrt_cache",
                                       sizeof(struct mfc_cache),
                                       0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
-                                      NULL, NULL);
+                                      NULL);
        init_timer(&ipmr_expire_timer);
        ipmr_expire_timer.function=ipmr_expire_process;
        register_netdevice_notifier(&ip_mr_notifier);
 #ifdef CONFIG_PROC_FS
-       proc_net_fops_create("ip_mr_vif", 0, &ipmr_vif_fops);
-       proc_net_fops_create("ip_mr_cache", 0, &ipmr_mfc_fops);
+       proc_net_fops_create(&init_net, "ip_mr_vif", 0, &ipmr_vif_fops);
+       proc_net_fops_create(&init_net, "ip_mr_cache", 0, &ipmr_mfc_fops);
 #endif
 }