diff options
author | Ville Nuorvala <vnuorval@tcs.hut.fi> | 2006-11-25 04:08:32 +0300 |
---|---|---|
committer | David S. Miller <davem@sunset.davemloft.net> | 2006-12-03 08:30:26 +0300 |
commit | 6fb32ddeb2502292bb89e17f58600ce265598ed7 (patch) | |
tree | 4870680606e4dad418c52de2cd5a842ce53fed2c | |
parent | 305d4b3ce8c2f00643edc3bb19f005f72e8f84fc (diff) | |
download | linux-6fb32ddeb2502292bb89e17f58600ce265598ed7.tar.xz |
[IPV6]: Don't allocate memory for Tunnel Encapsulation Limit Option
Signed-off-by: Ville Nuorvala <vnuorval@tcs.hut.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
-rw-r--r-- | net/ipv6/ip6_tunnel.c | 58 |
1 files changed, 22 insertions, 36 deletions
diff --git a/net/ipv6/ip6_tunnel.c b/net/ipv6/ip6_tunnel.c index 3944ea24c38c..daad1e5a2a85 100644 --- a/net/ipv6/ip6_tunnel.c +++ b/net/ipv6/ip6_tunnel.c @@ -572,31 +572,23 @@ discard: return 0; } -static inline struct ipv6_txoptions *create_tel(__u8 encap_limit) -{ - struct ipv6_tlv_tnl_enc_lim *tel; - struct ipv6_txoptions *opt; - __u8 *raw; - - int opt_len = sizeof(*opt) + 8; - - if (!(opt = kzalloc(opt_len, GFP_ATOMIC))) { - return NULL; - } - opt->tot_len = opt_len; - opt->dst0opt = (struct ipv6_opt_hdr *) (opt + 1); - opt->opt_nflen = 8; +struct ipv6_tel_txoption { + struct ipv6_txoptions ops; + __u8 dst_opt[8]; +}; - tel = (struct ipv6_tlv_tnl_enc_lim *) (opt->dst0opt + 1); - tel->type = IPV6_TLV_TNL_ENCAP_LIMIT; - tel->length = 1; - tel->encap_limit = encap_limit; +static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit) +{ + memset(opt, 0, sizeof(struct ipv6_tel_txoption)); - raw = (__u8 *) opt->dst0opt; - raw[5] = IPV6_TLV_PADN; - raw[6] = 1; + opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT; + opt->dst_opt[3] = 1; + opt->dst_opt[4] = encap_limit; + opt->dst_opt[5] = IPV6_TLV_PADN; + opt->dst_opt[6] = 1; - return opt; + opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt; + opt->ops.opt_nflen = 8; } /** @@ -666,8 +658,8 @@ ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) struct ip6_tnl *t = netdev_priv(dev); struct net_device_stats *stats = &t->stat; struct ipv6hdr *ipv6h = skb->nh.ipv6h; - struct ipv6_txoptions *opt = NULL; int encap_limit = -1; + struct ipv6_tel_txoption opt; __u16 offset; struct flowi fl; struct dst_entry *dst; @@ -696,9 +688,9 @@ ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) goto tx_err; } encap_limit = tel->encap_limit - 1; - } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) { + } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) encap_limit = t->parms.encap_limit; - } + memcpy(&fl, &t->fl, sizeof (fl)); proto = fl.proto; @@ -708,9 +700,6 @@ ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) if ((t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)) fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK); - if (encap_limit >= 0 && (opt = create_tel(encap_limit)) == NULL) - goto tx_err; - if ((dst = ip6_tnl_dst_check(t)) != NULL) dst_hold(dst); else { @@ -731,7 +720,7 @@ ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) goto tx_err_dst_release; } mtu = dst_mtu(dst) - sizeof (*ipv6h); - if (opt) { + if (encap_limit >= 0) { max_headroom += 8; mtu -= 8; } @@ -769,9 +758,10 @@ ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) skb->h.raw = skb->nh.raw; - if (opt) - ipv6_push_nfrag_opts(skb, opt, &proto, NULL); - + if (encap_limit >= 0) { + init_tel_txopt(&opt, encap_limit); + ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL); + } skb->nh.raw = skb_push(skb, sizeof(struct ipv6hdr)); ipv6h = skb->nh.ipv6h; *(__be32*)ipv6h = fl.fl6_flowlabel | htonl(0x60000000); @@ -795,9 +785,6 @@ ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev) stats->tx_aborted_errors++; } ip6_tnl_dst_store(t, dst); - - kfree(opt); - t->recursion--; return 0; tx_err_link_failure: @@ -805,7 +792,6 @@ tx_err_link_failure: dst_link_failure(skb); tx_err_dst_release: dst_release(dst); - kfree(opt); tx_err: stats->tx_errors++; stats->tx_dropped++; |