linux/net/ipv6/ip6_output.c

1602 lines
40 KiB
C
Raw Normal View History

/*
* IPv6 output functions
* Linux INET6 implementation
*
* Authors:
* Pedro Roque <roque@di.fc.ul.pt>
*
* Based on linux/net/ipv4/ip_output.c
*
* 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.
*
* Changes:
* A.N.Kuznetsov : airthmetics in fragmentation.
* extension headers are implemented.
* route changes now work.
* ip6_forward does not confuse sniffers.
* etc.
*
* H. von Brand : Added missing #include <linux/string.h>
* Imran Patel : frag id should be in NBO
* Kazunori MIYAZAWA @USAGI
* : add ip6_append_data and related functions
* for datagram xmit
*/
#include <linux/errno.h>
#include <linux/kernel.h>
#include <linux/string.h>
#include <linux/socket.h>
#include <linux/net.h>
#include <linux/netdevice.h>
#include <linux/if_arp.h>
#include <linux/in6.h>
#include <linux/tcp.h>
#include <linux/route.h>
#include <linux/module.h>
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h percpu.h is included by sched.h and module.h and thus ends up being included when building most .c files. percpu.h includes slab.h which in turn includes gfp.h making everything defined by the two files universally available and complicating inclusion dependencies. percpu.h -> slab.h dependency is about to be removed. Prepare for this change by updating users of gfp and slab facilities include those headers directly instead of assuming availability. As this conversion needs to touch large number of source files, the following script is used as the basis of conversion. http://userweb.kernel.org/~tj/misc/slabh-sweep.py The script does the followings. * Scan files for gfp and slab usages and update includes such that only the necessary includes are there. ie. if only gfp is used, gfp.h, if slab is used, slab.h. * When the script inserts a new include, it looks at the include blocks and try to put the new include such that its order conforms to its surrounding. It's put in the include block which contains core kernel includes, in the same order that the rest are ordered - alphabetical, Christmas tree, rev-Xmas-tree or at the end if there doesn't seem to be any matching order. * If the script can't find a place to put a new include (mostly because the file doesn't have fitting include block), it prints out an error message indicating which .h file needs to be added to the file. The conversion was done in the following steps. 1. The initial automatic conversion of all .c files updated slightly over 4000 files, deleting around 700 includes and adding ~480 gfp.h and ~3000 slab.h inclusions. The script emitted errors for ~400 files. 2. Each error was manually checked. Some didn't need the inclusion, some needed manual addition while adding it to implementation .h or embedding .c file was more appropriate for others. This step added inclusions to around 150 files. 3. The script was run again and the output was compared to the edits from #2 to make sure no file was left behind. 4. Several build tests were done and a couple of problems were fixed. e.g. lib/decompress_*.c used malloc/free() wrappers around slab APIs requiring slab.h to be added manually. 5. The script was run on all .h files but without automatically editing them as sprinkling gfp.h and slab.h inclusions around .h files could easily lead to inclusion dependency hell. Most gfp.h inclusion directives were ignored as stuff from gfp.h was usually wildly available and often used in preprocessor macros. Each slab.h inclusion directive was examined and added manually as necessary. 6. percpu.h was updated not to include slab.h. 7. Build test were done on the following configurations and failures were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my distributed build env didn't work with gcov compiles) and a few more options had to be turned off depending on archs to make things build (like ipr on powerpc/64 which failed due to missing writeq). * x86 and x86_64 UP and SMP allmodconfig and a custom test config. * powerpc and powerpc64 SMP allmodconfig * sparc and sparc64 SMP allmodconfig * ia64 SMP allmodconfig * s390 SMP allmodconfig * alpha SMP allmodconfig * um on x86_64 SMP allmodconfig 8. percpu.h modifications were reverted so that it could be applied as a separate patch and serve as bisection point. Given the fact that I had only a couple of failures from tests on step 6, I'm fairly confident about the coverage of this conversion patch. If there is a breakage, it's likely to be something in one of the arch headers which should be easily discoverable easily on most builds of the specific arch. Signed-off-by: Tejun Heo <tj@kernel.org> Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 16:04:11 +08:00
#include <linux/slab.h>
#include <linux/netfilter.h>
#include <linux/netfilter_ipv6.h>
#include <net/sock.h>
#include <net/snmp.h>
#include <net/ipv6.h>
#include <net/ndisc.h>
#include <net/protocol.h>
#include <net/ip6_route.h>
#include <net/addrconf.h>
#include <net/rawv6.h>
#include <net/icmp.h>
#include <net/xfrm.h>
#include <net/checksum.h>
#include <linux/mroute6.h>
int ip6_fragment(struct sk_buff *skb, int (*output)(struct sk_buff *));
int __ip6_local_out(struct sk_buff *skb)
{
int len;
len = skb->len - sizeof(struct ipv6hdr);
if (len > IPV6_MAXPLEN)
len = 0;
ipv6_hdr(skb)->payload_len = htons(len);
return nf_hook(NFPROTO_IPV6, NF_INET_LOCAL_OUT, skb, NULL,
skb_dst(skb)->dev, dst_output);
}
int ip6_local_out(struct sk_buff *skb)
{
int err;
err = __ip6_local_out(skb);
if (likely(err == 1))
err = dst_output(skb);
return err;
}
EXPORT_SYMBOL_GPL(ip6_local_out);
static int ip6_finish_output2(struct sk_buff *skb)
{
struct dst_entry *dst = skb_dst(skb);
struct net_device *dev = dst->dev;
struct neighbour *neigh;
struct rt6_info *rt;
skb->protocol = htons(ETH_P_IPV6);
skb->dev = dev;
if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr)) {
struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb));
if (!(dev->flags & IFF_LOOPBACK) && sk_mc_loop(skb->sk) &&
((mroute6_socket(dev_net(dev), skb) &&
!(IP6CB(skb)->flags & IP6SKB_FORWARDED)) ||
ipv6_chk_mcast_addr(dev, &ipv6_hdr(skb)->daddr,
&ipv6_hdr(skb)->saddr))) {
struct sk_buff *newskb = skb_clone(skb, GFP_ATOMIC);
/* Do not check for IFF_ALLMULTI; multicast routing
is not supported in any case.
*/
if (newskb)
NF_HOOK(NFPROTO_IPV6, NF_INET_POST_ROUTING,
newskb, NULL, newskb->dev,
dev_loopback_xmit);
if (ipv6_hdr(skb)->hop_limit == 0) {
IP6_INC_STATS(dev_net(dev), idev,
IPSTATS_MIB_OUTDISCARDS);
kfree_skb(skb);
return 0;
}
}
IP6_UPD_PO_STATS(dev_net(dev), idev, IPSTATS_MIB_OUTMCAST,
skb->len);
}
rt = (struct rt6_info *) dst;
neigh = rt->n;
if (neigh)
return dst_neigh_output(dst, neigh, skb);
IP6_INC_STATS_BH(dev_net(dst->dev),
ip6_dst_idev(dst), IPSTATS_MIB_OUTNOROUTES);
kfree_skb(skb);
return -EINVAL;
}
static int ip6_finish_output(struct sk_buff *skb)
{
if ((skb->len > ip6_skb_dst_mtu(skb) && !skb_is_gso(skb)) ||
dst_allfrag(skb_dst(skb)))
return ip6_fragment(skb, ip6_finish_output2);
else
return ip6_finish_output2(skb);
}
int ip6_output(struct sk_buff *skb)
{
struct net_device *dev = skb_dst(skb)->dev;
struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb));
if (unlikely(idev->cnf.disable_ipv6)) {
IP6_INC_STATS(dev_net(dev), idev,
IPSTATS_MIB_OUTDISCARDS);
kfree_skb(skb);
return 0;
}
return NF_HOOK_COND(NFPROTO_IPV6, NF_INET_POST_ROUTING, skb, NULL, dev,
ip6_finish_output,
!(IP6CB(skb)->flags & IP6SKB_REROUTED));
}
/*
* xmit an sk_buff (used by TCP, SCTP and DCCP)
*/
int ip6_xmit(struct sock *sk, struct sk_buff *skb, struct flowi6 *fl6,
struct ipv6_txoptions *opt, int tclass)
{
struct net *net = sock_net(sk);
struct ipv6_pinfo *np = inet6_sk(sk);
struct in6_addr *first_hop = &fl6->daddr;
struct dst_entry *dst = skb_dst(skb);
struct ipv6hdr *hdr;
u8 proto = fl6->flowi6_proto;
int seg_len = skb->len;
int hlimit = -1;
u32 mtu;
if (opt) {
unsigned int head_room;
/* First: exthdrs may take lots of space (~8K for now)
MAX_HEADER is not enough.
*/
head_room = opt->opt_nflen + opt->opt_flen;
seg_len += head_room;
head_room += sizeof(struct ipv6hdr) + LL_RESERVED_SPACE(dst->dev);
if (skb_headroom(skb) < head_room) {
struct sk_buff *skb2 = skb_realloc_headroom(skb, head_room);
if (skb2 == NULL) {
IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
IPSTATS_MIB_OUTDISCARDS);
kfree_skb(skb);
return -ENOBUFS;
}
consume_skb(skb);
skb = skb2;
skb_set_owner_w(skb, sk);
}
if (opt->opt_flen)
ipv6_push_frag_opts(skb, opt, &proto);
if (opt->opt_nflen)
ipv6_push_nfrag_opts(skb, opt, &proto, &first_hop);
}
skb_push(skb, sizeof(struct ipv6hdr));
skb_reset_network_header(skb);
hdr = ipv6_hdr(skb);
/*
* Fill in the IPv6 header
*/
if (np)
hlimit = np->hop_limit;
if (hlimit < 0)
hlimit = ip6_dst_hoplimit(dst);
*(__be32 *)hdr = htonl(0x60000000 | (tclass << 20)) | fl6->flowlabel;
hdr->payload_len = htons(seg_len);
hdr->nexthdr = proto;
hdr->hop_limit = hlimit;
hdr->saddr = fl6->saddr;
hdr->daddr = *first_hop;
skb->priority = sk->sk_priority;
skb->mark = sk->sk_mark;
mtu = dst_mtu(dst);
if ((skb->len <= mtu) || skb->local_df || skb_is_gso(skb)) {
IP6_UPD_PO_STATS(net, ip6_dst_idev(skb_dst(skb)),
IPSTATS_MIB_OUT, skb->len);
return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, skb, NULL,
dst->dev, dst_output);
}
net_dbg_ratelimited("IPv6: sending pkt_too_big to self\n");
skb->dev = dst->dev;
icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_FRAGFAILS);
kfree_skb(skb);
return -EMSGSIZE;
}
EXPORT_SYMBOL(ip6_xmit);
/*
* To avoid extra problems ND packets are send through this
* routine. It's code duplication but I really want to avoid
* extra checks since ipv6_build_header is used by TCP (which
* is for us performance critical)
*/
int ip6_nd_hdr(struct sock *sk, struct sk_buff *skb, struct net_device *dev,
const struct in6_addr *saddr, const struct in6_addr *daddr,
int proto, int len)
{
struct ipv6_pinfo *np = inet6_sk(sk);
struct ipv6hdr *hdr;
skb->protocol = htons(ETH_P_IPV6);
skb->dev = dev;
skb_reset_network_header(skb);
skb_put(skb, sizeof(struct ipv6hdr));
hdr = ipv6_hdr(skb);
*(__be32*)hdr = htonl(0x60000000);
hdr->payload_len = htons(len);
hdr->nexthdr = proto;
hdr->hop_limit = np->hop_limit;
hdr->saddr = *saddr;
hdr->daddr = *daddr;
return 0;
}
static int ip6_call_ra_chain(struct sk_buff *skb, int sel)
{
struct ip6_ra_chain *ra;
struct sock *last = NULL;
read_lock(&ip6_ra_lock);
for (ra = ip6_ra_chain; ra; ra = ra->next) {
struct sock *sk = ra->sk;
if (sk && ra->sel == sel &&
(!sk->sk_bound_dev_if ||
sk->sk_bound_dev_if == skb->dev->ifindex)) {
if (last) {
struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
if (skb2)
rawv6_rcv(last, skb2);
}
last = sk;
}
}
if (last) {
rawv6_rcv(last, skb);
read_unlock(&ip6_ra_lock);
return 1;
}
read_unlock(&ip6_ra_lock);
return 0;
}
static int ip6_forward_proxy_check(struct sk_buff *skb)
{
struct ipv6hdr *hdr = ipv6_hdr(skb);
u8 nexthdr = hdr->nexthdr;
__be16 frag_off;
int offset;
if (ipv6_ext_hdr(nexthdr)) {
offset = ipv6_skip_exthdr(skb, sizeof(*hdr), &nexthdr, &frag_off);
if (offset < 0)
return 0;
} else
offset = sizeof(struct ipv6hdr);
if (nexthdr == IPPROTO_ICMPV6) {
struct icmp6hdr *icmp6;
if (!pskb_may_pull(skb, (skb_network_header(skb) +
offset + 1 - skb->data)))
return 0;
icmp6 = (struct icmp6hdr *)(skb_network_header(skb) + offset);
switch (icmp6->icmp6_type) {
case NDISC_ROUTER_SOLICITATION:
case NDISC_ROUTER_ADVERTISEMENT:
case NDISC_NEIGHBOUR_SOLICITATION:
case NDISC_NEIGHBOUR_ADVERTISEMENT:
case NDISC_REDIRECT:
/* For reaction involving unicast neighbor discovery
* message destined to the proxied address, pass it to
* input function.
*/
return 1;
default:
break;
}
}
/*
* The proxying router can't forward traffic sent to a link-local
* address, so signal the sender and discard the packet. This
* behavior is clarified by the MIPv6 specification.
*/
if (ipv6_addr_type(&hdr->daddr) & IPV6_ADDR_LINKLOCAL) {
dst_link_failure(skb);
return -1;
}
return 0;
}
static inline int ip6_forward_finish(struct sk_buff *skb)
{
return dst_output(skb);
}
int ip6_forward(struct sk_buff *skb)
{
struct dst_entry *dst = skb_dst(skb);
struct ipv6hdr *hdr = ipv6_hdr(skb);
struct inet6_skb_parm *opt = IP6CB(skb);
struct net *net = dev_net(dst->dev);
u32 mtu;
if (net->ipv6.devconf_all->forwarding == 0)
goto error;
if (skb_warn_if_lro(skb))
goto drop;
if (!xfrm6_policy_check(NULL, XFRM_POLICY_FWD, skb)) {
IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_INDISCARDS);
goto drop;
}
if (skb->pkt_type != PACKET_HOST)
goto drop;
skb_forward_csum(skb);
/*
* We DO NOT make any processing on
* RA packets, pushing them to user level AS IS
* without ane WARRANTY that application will be able
* to interpret them. The reason is that we
* cannot make anything clever here.
*
* We are not end-node, so that if packet contains
* AH/ESP, we cannot make anything.
* Defragmentation also would be mistake, RA packets
* cannot be fragmented, because there is no warranty
* that different fragments will go along one path. --ANK
*/
if (opt->ra) {
u8 *ptr = skb_network_header(skb) + opt->ra;
if (ip6_call_ra_chain(skb, (ptr[2]<<8) + ptr[3]))
return 0;
}
/*
* check and decrement ttl
*/
if (hdr->hop_limit <= 1) {
/* Force OUTPUT device used as source address */
skb->dev = dst->dev;
icmpv6_send(skb, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT, 0);
IP6_INC_STATS_BH(net,
ip6_dst_idev(dst), IPSTATS_MIB_INHDRERRORS);
kfree_skb(skb);
return -ETIMEDOUT;
}
/* XXX: idev->cnf.proxy_ndp? */
if (net->ipv6.devconf_all->proxy_ndp &&
pneigh_lookup(&nd_tbl, net, &hdr->daddr, skb->dev, 0)) {
int proxied = ip6_forward_proxy_check(skb);
if (proxied > 0)
return ip6_input(skb);
else if (proxied < 0) {
IP6_INC_STATS(net, ip6_dst_idev(dst),
IPSTATS_MIB_INDISCARDS);
goto drop;
}
}
if (!xfrm6_route_forward(skb)) {
IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_INDISCARDS);
goto drop;
}
dst = skb_dst(skb);
/* IPv6 specs say nothing about it, but it is clear that we cannot
send redirects to source routed frames.
We don't send redirects to frames decapsulated from IPsec.
*/
if (skb->dev == dst->dev && opt->srcrt == 0 && !skb_sec_path(skb)) {
struct in6_addr *target = NULL;
struct inet_peer *peer;
struct rt6_info *rt;
/*
* incoming and outgoing devices are the same
* send a redirect.
*/
rt = (struct rt6_info *) dst;
if (rt->rt6i_flags & RTF_GATEWAY)
target = &rt->rt6i_gateway;
else
target = &hdr->daddr;
peer = inet_getpeer_v6(net->ipv6.peers, &rt->rt6i_dst.addr, 1);
/* Limit redirects both by destination (here)
and by source (inside ndisc_send_redirect)
*/
if (inet_peer_xrlim_allow(peer, 1*HZ))
ndisc_send_redirect(skb, target);
if (peer)
inet_putpeer(peer);
} else {
int addrtype = ipv6_addr_type(&hdr->saddr);
/* This check is security critical. */
if (addrtype == IPV6_ADDR_ANY ||
addrtype & (IPV6_ADDR_MULTICAST | IPV6_ADDR_LOOPBACK))
goto error;
if (addrtype & IPV6_ADDR_LINKLOCAL) {
icmpv6_send(skb, ICMPV6_DEST_UNREACH,
ICMPV6_NOT_NEIGHBOUR, 0);
goto error;
}
}
mtu = dst_mtu(dst);
if (mtu < IPV6_MIN_MTU)
mtu = IPV6_MIN_MTU;
netfilter: nf_conntrack_ipv6: improve fragmentation handling The IPv6 conntrack fragmentation currently has a couple of shortcomings. Fragmentes are collected in PREROUTING/OUTPUT, are defragmented, the defragmented packet is then passed to conntrack, the resulting conntrack information is attached to each original fragment and the fragments then continue their way through the stack. Helper invocation occurs in the POSTROUTING hook, at which point only the original fragments are available. The result of this is that fragmented packets are never passed to helpers. This patch improves the situation in the following way: - If a reassembled packet belongs to a connection that has a helper assigned, the reassembled packet is passed through the stack instead of the original fragments. - During defragmentation, the largest received fragment size is stored. On output, the packet is refragmented if required. If the largest received fragment size exceeds the outgoing MTU, a "packet too big" message is generated, thus behaving as if the original fragments were passed through the stack from an outside point of view. - The ipv6_helper() hook function can't receive fragments anymore for connections using a helper, so it is switched to use ipv6_skip_exthdr() instead of the netfilter specific nf_ct_ipv6_skip_exthdr() and the reassembled packets are passed to connection tracking helpers. The result of this is that we can properly track fragmented packets, but still generate ICMPv6 Packet too big messages if we would have before. This patch is also required as a precondition for IPv6 NAT, where NAT helpers might enlarge packets up to a point that they require fragmentation. In that case we can't generate Packet too big messages since the proper MTU can't be calculated in all cases (f.i. when changing textual representation of a variable amount of addresses), so the packet is transparently fragmented iff the original packet or fragments would have fit the outgoing MTU. IPVS parts by Jesper Dangaard Brouer <brouer@redhat.com>. Signed-off-by: Patrick McHardy <kaber@trash.net>
2012-08-27 01:13:58 +08:00
if ((!skb->local_df && skb->len > mtu && !skb_is_gso(skb)) ||
(IP6CB(skb)->frag_max_size && IP6CB(skb)->frag_max_size > mtu)) {
/* Again, force OUTPUT device used as source address */
skb->dev = dst->dev;
icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
IP6_INC_STATS_BH(net,
ip6_dst_idev(dst), IPSTATS_MIB_INTOOBIGERRORS);
IP6_INC_STATS_BH(net,
ip6_dst_idev(dst), IPSTATS_MIB_FRAGFAILS);
kfree_skb(skb);
return -EMSGSIZE;
}
if (skb_cow(skb, dst->dev->hard_header_len)) {
IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTDISCARDS);
goto drop;
}
hdr = ipv6_hdr(skb);
/* Mangling hops number delayed to point after skb COW */
hdr->hop_limit--;
IP6_INC_STATS_BH(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTFORWDATAGRAMS);
IP6_ADD_STATS_BH(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTOCTETS, skb->len);
return NF_HOOK(NFPROTO_IPV6, NF_INET_FORWARD, skb, skb->dev, dst->dev,
ip6_forward_finish);
error:
IP6_INC_STATS_BH(net, ip6_dst_idev(dst), IPSTATS_MIB_INADDRERRORS);
drop:
kfree_skb(skb);
return -EINVAL;
}
static void ip6_copy_metadata(struct sk_buff *to, struct sk_buff *from)
{
to->pkt_type = from->pkt_type;
to->priority = from->priority;
to->protocol = from->protocol;
skb_dst_drop(to);
skb_dst_set(to, dst_clone(skb_dst(from)));
to->dev = from->dev;
to->mark = from->mark;
#ifdef CONFIG_NET_SCHED
to->tc_index = from->tc_index;
#endif
nf_copy(to, from);
#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
to->nf_trace = from->nf_trace;
#endif
skb_copy_secmark(to, from);
}
int ip6_fragment(struct sk_buff *skb, int (*output)(struct sk_buff *))
{
struct sk_buff *frag;
struct rt6_info *rt = (struct rt6_info*)skb_dst(skb);
struct ipv6_pinfo *np = skb->sk ? inet6_sk(skb->sk) : NULL;
struct ipv6hdr *tmp_hdr;
struct frag_hdr *fh;
unsigned int mtu, hlen, left, len;
int hroom, troom;
__be32 frag_id = 0;
int ptr, offset = 0, err=0;
u8 *prevhdr, nexthdr = 0;
struct net *net = dev_net(skb_dst(skb)->dev);
hlen = ip6_find_1stfragopt(skb, &prevhdr);
nexthdr = *prevhdr;
mtu = ip6_skb_dst_mtu(skb);
/* We must not fragment if the socket is set to force MTU discovery
* or if the skb it not generated by a local socket.
*/
netfilter: nf_conntrack_ipv6: improve fragmentation handling The IPv6 conntrack fragmentation currently has a couple of shortcomings. Fragmentes are collected in PREROUTING/OUTPUT, are defragmented, the defragmented packet is then passed to conntrack, the resulting conntrack information is attached to each original fragment and the fragments then continue their way through the stack. Helper invocation occurs in the POSTROUTING hook, at which point only the original fragments are available. The result of this is that fragmented packets are never passed to helpers. This patch improves the situation in the following way: - If a reassembled packet belongs to a connection that has a helper assigned, the reassembled packet is passed through the stack instead of the original fragments. - During defragmentation, the largest received fragment size is stored. On output, the packet is refragmented if required. If the largest received fragment size exceeds the outgoing MTU, a "packet too big" message is generated, thus behaving as if the original fragments were passed through the stack from an outside point of view. - The ipv6_helper() hook function can't receive fragments anymore for connections using a helper, so it is switched to use ipv6_skip_exthdr() instead of the netfilter specific nf_ct_ipv6_skip_exthdr() and the reassembled packets are passed to connection tracking helpers. The result of this is that we can properly track fragmented packets, but still generate ICMPv6 Packet too big messages if we would have before. This patch is also required as a precondition for IPv6 NAT, where NAT helpers might enlarge packets up to a point that they require fragmentation. In that case we can't generate Packet too big messages since the proper MTU can't be calculated in all cases (f.i. when changing textual representation of a variable amount of addresses), so the packet is transparently fragmented iff the original packet or fragments would have fit the outgoing MTU. IPVS parts by Jesper Dangaard Brouer <brouer@redhat.com>. Signed-off-by: Patrick McHardy <kaber@trash.net>
2012-08-27 01:13:58 +08:00
if (unlikely(!skb->local_df && skb->len > mtu) ||
(IP6CB(skb)->frag_max_size &&
IP6CB(skb)->frag_max_size > mtu)) {
if (skb->sk && dst_allfrag(skb_dst(skb)))
sk_nocaps_add(skb->sk, NETIF_F_GSO_MASK);
skb->dev = skb_dst(skb)->dev;
icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
IPSTATS_MIB_FRAGFAILS);
kfree_skb(skb);
return -EMSGSIZE;
}
if (np && np->frag_size < mtu) {
if (np->frag_size)
mtu = np->frag_size;
}
mtu -= hlen + sizeof(struct frag_hdr);
if (skb_has_frag_list(skb)) {
int first_len = skb_pagelen(skb);
struct sk_buff *frag2;
if (first_len - hlen > mtu ||
((first_len - hlen) & 7) ||
skb_cloned(skb))
goto slow_path;
skb_walk_frags(skb, frag) {
/* Correct geometry. */
if (frag->len > mtu ||
((frag->len & 7) && frag->next) ||
skb_headroom(frag) < hlen)
goto slow_path_clean;
/* Partially cloned skb? */
if (skb_shared(frag))
goto slow_path_clean;
BUG_ON(frag->sk);
if (skb->sk) {
frag->sk = skb->sk;
frag->destructor = sock_wfree;
}
skb->truesize -= frag->truesize;
}
err = 0;
offset = 0;
frag = skb_shinfo(skb)->frag_list;
skb_frag_list_init(skb);
/* BUILD HEADER */
*prevhdr = NEXTHDR_FRAGMENT;
tmp_hdr = kmemdup(skb_network_header(skb), hlen, GFP_ATOMIC);
if (!tmp_hdr) {
IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
IPSTATS_MIB_FRAGFAILS);
return -ENOMEM;
}
__skb_pull(skb, hlen);
fh = (struct frag_hdr*)__skb_push(skb, sizeof(struct frag_hdr));
__skb_push(skb, hlen);
skb_reset_network_header(skb);
memcpy(skb_network_header(skb), tmp_hdr, hlen);
ipv6_select_ident(fh, rt);
fh->nexthdr = nexthdr;
fh->reserved = 0;
fh->frag_off = htons(IP6_MF);
frag_id = fh->identification;
first_len = skb_pagelen(skb);
skb->data_len = first_len - skb_headlen(skb);
skb->len = first_len;
ipv6_hdr(skb)->payload_len = htons(first_len -
sizeof(struct ipv6hdr));
dst_hold(&rt->dst);
for (;;) {
/* Prepare header of the next frame,
* before previous one went down. */
if (frag) {
frag->ip_summed = CHECKSUM_NONE;
skb_reset_transport_header(frag);
fh = (struct frag_hdr*)__skb_push(frag, sizeof(struct frag_hdr));
__skb_push(frag, hlen);
skb_reset_network_header(frag);
memcpy(skb_network_header(frag), tmp_hdr,
hlen);
offset += skb->len - hlen - sizeof(struct frag_hdr);
fh->nexthdr = nexthdr;
fh->reserved = 0;
fh->frag_off = htons(offset);
if (frag->next != NULL)
fh->frag_off |= htons(IP6_MF);
fh->identification = frag_id;
ipv6_hdr(frag)->payload_len =
htons(frag->len -
sizeof(struct ipv6hdr));
ip6_copy_metadata(frag, skb);
}
err = output(skb);
[IPV6]: SNMPv2 "ipv6IfStatsOutFragCreates" counter error When I tested linux kernel 2.6.71.7 about statistics "ipv6IfStatsOutFragCreates", and found that it couldn't increase correctly. The criteria is RFC 2465: ipv6IfStatsOutFragCreates OBJECT-TYPE SYNTAX Counter32 MAX-ACCESS read-only STATUS current DESCRIPTION "The number of output datagram fragments that have been generated as a result of fragmentation at this output interface." ::= { ipv6IfStatsEntry 15 } I think there are two issues in Linux kernel. 1st: RFC2465 specifies the counter is "The number of output datagram fragments...". I think increasing this counter after output a fragment successfully is better. And it should not be increased even though a fragment is created but failed to output. 2nd: If we send a big ICMP/ICMPv6 echo request to a host, and receive ICMP/ICMPv6 echo reply consisted of some fragments. As we know that in Linux kernel first fragmentation occurs in ICMP layer(maybe saying transport layer is better), but this is not the "real" fragmentation,just do some "pre-fragment" -- allocate space for date, and form a frag_list, etc. The "real" fragmentation happens in IP layer -- set offset and MF flag and so on. So I think in "fast path" for ip_fragment/ip6_fragment, if we send a fragment which "pre-fragment" by upper layer we should also increase "ipv6IfStatsOutFragCreates". Signed-off-by: Wei Dong <weid@nanjing-fnst.com> Acked-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2006-08-03 04:41:21 +08:00
if(!err)
IP6_INC_STATS(net, ip6_dst_idev(&rt->dst),
IPSTATS_MIB_FRAGCREATES);
[IPV6]: SNMPv2 "ipv6IfStatsOutFragCreates" counter error When I tested linux kernel 2.6.71.7 about statistics "ipv6IfStatsOutFragCreates", and found that it couldn't increase correctly. The criteria is RFC 2465: ipv6IfStatsOutFragCreates OBJECT-TYPE SYNTAX Counter32 MAX-ACCESS read-only STATUS current DESCRIPTION "The number of output datagram fragments that have been generated as a result of fragmentation at this output interface." ::= { ipv6IfStatsEntry 15 } I think there are two issues in Linux kernel. 1st: RFC2465 specifies the counter is "The number of output datagram fragments...". I think increasing this counter after output a fragment successfully is better. And it should not be increased even though a fragment is created but failed to output. 2nd: If we send a big ICMP/ICMPv6 echo request to a host, and receive ICMP/ICMPv6 echo reply consisted of some fragments. As we know that in Linux kernel first fragmentation occurs in ICMP layer(maybe saying transport layer is better), but this is not the "real" fragmentation,just do some "pre-fragment" -- allocate space for date, and form a frag_list, etc. The "real" fragmentation happens in IP layer -- set offset and MF flag and so on. So I think in "fast path" for ip_fragment/ip6_fragment, if we send a fragment which "pre-fragment" by upper layer we should also increase "ipv6IfStatsOutFragCreates". Signed-off-by: Wei Dong <weid@nanjing-fnst.com> Acked-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2006-08-03 04:41:21 +08:00
if (err || !frag)
break;
skb = frag;
frag = skb->next;
skb->next = NULL;
}
kfree(tmp_hdr);
if (err == 0) {
IP6_INC_STATS(net, ip6_dst_idev(&rt->dst),
IPSTATS_MIB_FRAGOKS);
ip6_rt_put(rt);
return 0;
}
while (frag) {
skb = frag->next;
kfree_skb(frag);
frag = skb;
}
IP6_INC_STATS(net, ip6_dst_idev(&rt->dst),
IPSTATS_MIB_FRAGFAILS);
ip6_rt_put(rt);
return err;
slow_path_clean:
skb_walk_frags(skb, frag2) {
if (frag2 == frag)
break;
frag2->sk = NULL;
frag2->destructor = NULL;
skb->truesize += frag2->truesize;
}
}
slow_path:
if ((skb->ip_summed == CHECKSUM_PARTIAL) &&
skb_checksum_help(skb))
goto fail;
left = skb->len - hlen; /* Space per frame */
ptr = hlen; /* Where to start from */
/*
* Fragment the datagram.
*/
*prevhdr = NEXTHDR_FRAGMENT;
hroom = LL_RESERVED_SPACE(rt->dst.dev);
troom = rt->dst.dev->needed_tailroom;
/*
* Keep copying data until we run out.
*/
while(left > 0) {
len = left;
/* IF: it doesn't fit, use 'mtu' - the data space left */
if (len > mtu)
len = mtu;
/* IF: we are not sending up to and including the packet end
then align the next start on an eight byte boundary */
if (len < left) {
len &= ~7;
}
/*
* Allocate buffer.
*/
if ((frag = alloc_skb(len + hlen + sizeof(struct frag_hdr) +
hroom + troom, GFP_ATOMIC)) == NULL) {
NETDEBUG(KERN_INFO "IPv6: frag: no memory for new fragment!\n");
IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
IPSTATS_MIB_FRAGFAILS);
err = -ENOMEM;
goto fail;
}
/*
* Set up data on packet
*/
ip6_copy_metadata(frag, skb);
skb_reserve(frag, hroom);
skb_put(frag, len + hlen + sizeof(struct frag_hdr));
skb_reset_network_header(frag);
fh = (struct frag_hdr *)(skb_network_header(frag) + hlen);
frag->transport_header = (frag->network_header + hlen +
sizeof(struct frag_hdr));
/*
* Charge the memory for the fragment to any owner
* it might possess
*/
if (skb->sk)
skb_set_owner_w(frag, skb->sk);
/*
* Copy the packet header into the new buffer.
*/
skb_copy_from_linear_data(skb, skb_network_header(frag), hlen);
/*
* Build fragment header.
*/
fh->nexthdr = nexthdr;
fh->reserved = 0;
if (!frag_id) {
ipv6_select_ident(fh, rt);
frag_id = fh->identification;
} else
fh->identification = frag_id;
/*
* Copy a block of the IP datagram.
*/
[IPV6]: Fix kernel panic while send SCTP data with IP fragments If ICMP6 message with "Packet Too Big" is received after send SCTP DATA, kernel panic will occur when SCTP DATA is send again. This is because of a bad dest address when call to skb_copy_bits(). The messages sequence is like this: Endpoint A Endpoint B <------- SCTP DATA (size=1432) ICMP6 message -------> (Packet Too Big pmtu=1280) <------- Resend SCTP DATA (size=1432) ------------kernel panic--------------- printing eip: c05be62a *pde = 00000000 Oops: 0002 [#1] SMP Modules linked in: scomm l2cap bluetooth ipv6 dm_mirror dm_mod video output sbs battery lp floppy sg i2c_piix4 i2c_core pcnet32 mii button ac parport_pc parport ide_cd cdrom serio_raw mptspi mptscsih mptbase scsi_transport_spi sd_mod scsi_mod ext3 jbd ehci_hcd ohci_hcd uhci_hcd CPU: 0 EIP: 0060:[<c05be62a>] Not tainted VLI EFLAGS: 00010282 (2.6.23-rc2 #1) EIP is at skb_copy_bits+0x4f/0x1ef eax: 000004d0 ebx: ce12a980 ecx: 00000134 edx: cfd5a880 esi: c8246858 edi: 00000000 ebp: c0759b14 esp: c0759adc ds: 007b es: 007b fs: 00d8 gs: 0000 ss: 0068 Process swapper (pid: 0, ti=c0759000 task=c06d0340 task.ti=c0713000) Stack: c0759b88 c0405867 ce12a980 c8bff838 c789c084 00000000 00000028 cfd5a880 d09f1890 000005dc 0000007b ce12a980 cfd5a880 c8bff838 c0759b88 d09bc521 000004d0 fffff96c 00000200 00000100 c0759b50 cfd5a880 00000246 c0759bd4 Call Trace: [<c0405e1d>] show_trace_log_lvl+0x1a/0x2f [<c0405ecd>] show_stack_log_lvl+0x9b/0xa3 [<c040608d>] show_registers+0x1b8/0x289 [<c0406271>] die+0x113/0x246 [<c0625dbc>] do_page_fault+0x4ad/0x57e [<c0624642>] error_code+0x72/0x78 [<d09bc521>] ip6_output+0x8e5/0xab2 [ipv6] [<d09bcec1>] ip6_xmit+0x2ea/0x3a3 [ipv6] [<d0a3f2ca>] sctp_v6_xmit+0x248/0x253 [sctp] [<d0a3c934>] sctp_packet_transmit+0x53f/0x5ae [sctp] [<d0a34bf8>] sctp_outq_flush+0x555/0x587 [sctp] [<d0a34d3c>] sctp_retransmit+0xf8/0x10f [sctp] [<d0a3d183>] sctp_icmp_frag_needed+0x57/0x5b [sctp] [<d0a3ece2>] sctp_v6_err+0xcd/0x148 [sctp] [<d09cf1ce>] icmpv6_notify+0xe6/0x167 [ipv6] [<d09d009a>] icmpv6_rcv+0x7d7/0x849 [ipv6] [<d09be240>] ip6_input+0x1dc/0x310 [ipv6] [<d09be965>] ipv6_rcv+0x294/0x2df [ipv6] [<c05c3789>] netif_receive_skb+0x2d2/0x335 [<c05c5733>] process_backlog+0x7f/0xd0 [<c05c58f6>] net_rx_action+0x96/0x17e [<c042e722>] __do_softirq+0x64/0xcd [<c0406f37>] do_softirq+0x5c/0xac ======================= Code: 00 00 29 ca 89 d0 2b 45 e0 89 55 ec 85 c0 7e 35 39 45 08 8b 55 e4 0f 4e 45 08 8b 75 e0 8b 7d dc 89 c1 c1 e9 02 03 b2 a0 00 00 00 <f3> a5 89 c1 83 e1 03 74 02 f3 a4 29 45 08 0f 84 7b 01 00 00 01 EIP: [<c05be62a>] skb_copy_bits+0x4f/0x1ef SS:ESP 0068:c0759adc Kernel panic - not syncing: Fatal exception in interrupt Arnaldo says: ==================== Thanks! I'm to blame for this one, problem was introduced in: b0e380b1d8a8e0aca215df97702f99815f05c094 @@ -761,7 +762,7 @@ slow_path: /* * Copy a block of the IP datagram. */ - if (skb_copy_bits(skb, ptr, frag->h.raw, len)) + if (skb_copy_bits(skb, ptr, skb_transport_header(skb), len)) BUG(); left -= len; ==================== Signed-off-by: Wei Yongjun <yjwei@cn.fujitsu.com> Acked-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org> Signed-off-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-08-22 11:59:08 +08:00
if (skb_copy_bits(skb, ptr, skb_transport_header(frag), len))
BUG();
left -= len;
fh->frag_off = htons(offset);
if (left > 0)
fh->frag_off |= htons(IP6_MF);
ipv6_hdr(frag)->payload_len = htons(frag->len -
sizeof(struct ipv6hdr));
ptr += len;
offset += len;
/*
* Put this fragment into the sending queue.
*/
err = output(frag);
if (err)
goto fail;
[IPV6]: SNMPv2 "ipv6IfStatsOutFragCreates" counter error When I tested linux kernel 2.6.71.7 about statistics "ipv6IfStatsOutFragCreates", and found that it couldn't increase correctly. The criteria is RFC 2465: ipv6IfStatsOutFragCreates OBJECT-TYPE SYNTAX Counter32 MAX-ACCESS read-only STATUS current DESCRIPTION "The number of output datagram fragments that have been generated as a result of fragmentation at this output interface." ::= { ipv6IfStatsEntry 15 } I think there are two issues in Linux kernel. 1st: RFC2465 specifies the counter is "The number of output datagram fragments...". I think increasing this counter after output a fragment successfully is better. And it should not be increased even though a fragment is created but failed to output. 2nd: If we send a big ICMP/ICMPv6 echo request to a host, and receive ICMP/ICMPv6 echo reply consisted of some fragments. As we know that in Linux kernel first fragmentation occurs in ICMP layer(maybe saying transport layer is better), but this is not the "real" fragmentation,just do some "pre-fragment" -- allocate space for date, and form a frag_list, etc. The "real" fragmentation happens in IP layer -- set offset and MF flag and so on. So I think in "fast path" for ip_fragment/ip6_fragment, if we send a fragment which "pre-fragment" by upper layer we should also increase "ipv6IfStatsOutFragCreates". Signed-off-by: Wei Dong <weid@nanjing-fnst.com> Acked-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2006-08-03 04:41:21 +08:00
IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
IPSTATS_MIB_FRAGCREATES);
}
IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
IPSTATS_MIB_FRAGOKS);
consume_skb(skb);
return err;
fail:
IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)),
IPSTATS_MIB_FRAGFAILS);
kfree_skb(skb);
return err;
}
static inline int ip6_rt_check(const struct rt6key *rt_key,
const struct in6_addr *fl_addr,
const struct in6_addr *addr_cache)
{
return (rt_key->plen != 128 || !ipv6_addr_equal(fl_addr, &rt_key->addr)) &&
(addr_cache == NULL || !ipv6_addr_equal(fl_addr, addr_cache));
}
static struct dst_entry *ip6_sk_dst_check(struct sock *sk,
struct dst_entry *dst,
const struct flowi6 *fl6)
{
struct ipv6_pinfo *np = inet6_sk(sk);
struct rt6_info *rt = (struct rt6_info *)dst;
if (!dst)
goto out;
/* Yes, checking route validity in not connected
* case is not very simple. Take into account,
* that we do not support routing by source, TOS,
* and MSG_DONTROUTE --ANK (980726)
*
* 1. ip6_rt_check(): If route was host route,
* check that cached destination is current.
* If it is network route, we still may
* check its validity using saved pointer
* to the last used address: daddr_cache.
* We do not want to save whole address now,
* (because main consumer of this service
* is tcp, which has not this problem),
* so that the last trick works only on connected
* sockets.
* 2. oif also should be the same.
*/
if (ip6_rt_check(&rt->rt6i_dst, &fl6->daddr, np->daddr_cache) ||
#ifdef CONFIG_IPV6_SUBTREES
ip6_rt_check(&rt->rt6i_src, &fl6->saddr, np->saddr_cache) ||
#endif
(fl6->flowi6_oif && fl6->flowi6_oif != dst->dev->ifindex)) {
dst_release(dst);
dst = NULL;
}
out:
return dst;
}
static int ip6_dst_lookup_tail(struct sock *sk,
struct dst_entry **dst, struct flowi6 *fl6)
{
struct net *net = sock_net(sk);
#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
struct neighbour *n;
struct rt6_info *rt;
#endif
int err;
if (*dst == NULL)
*dst = ip6_route_output(net, sk, fl6);
if ((err = (*dst)->error))
goto out_err_release;
if (ipv6_addr_any(&fl6->saddr)) {
struct rt6_info *rt = (struct rt6_info *) *dst;
err = ip6_route_get_saddr(net, rt, &fl6->daddr,
sk ? inet6_sk(sk)->srcprefs : 0,
&fl6->saddr);
if (err)
goto out_err_release;
}
#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
/*
* Here if the dst entry we've looked up
* has a neighbour entry that is in the INCOMPLETE
* state and the src address from the flow is
* marked as OPTIMISTIC, we release the found
* dst entry and replace it instead with the
* dst entry of the nexthop router
*/
rt = (struct rt6_info *) *dst;
n = rt->n;
if (n && !(n->nud_state & NUD_VALID)) {
struct inet6_ifaddr *ifp;
struct flowi6 fl_gw6;
int redirect;
ifp = ipv6_get_ifaddr(net, &fl6->saddr,
(*dst)->dev, 1);
redirect = (ifp && ifp->flags & IFA_F_OPTIMISTIC);
if (ifp)
in6_ifa_put(ifp);
if (redirect) {
/*
* We need to get the dst entry for the
* default router instead
*/
dst_release(*dst);
memcpy(&fl_gw6, fl6, sizeof(struct flowi6));
memset(&fl_gw6.daddr, 0, sizeof(struct in6_addr));
*dst = ip6_route_output(net, sk, &fl_gw6);
if ((err = (*dst)->error))
goto out_err_release;
}
}
#endif
return 0;
out_err_release:
if (err == -ENETUNREACH)
IP6_INC_STATS_BH(net, NULL, IPSTATS_MIB_OUTNOROUTES);
dst_release(*dst);
*dst = NULL;
return err;
}
/**
* ip6_dst_lookup - perform route lookup on flow
* @sk: socket which provides route info
* @dst: pointer to dst_entry * for result
* @fl6: flow to lookup
*
* This function performs a route lookup on the given flow.
*
* It returns zero on success, or a standard errno code on error.
*/
int ip6_dst_lookup(struct sock *sk, struct dst_entry **dst, struct flowi6 *fl6)
{
*dst = NULL;
return ip6_dst_lookup_tail(sk, dst, fl6);
}
EXPORT_SYMBOL_GPL(ip6_dst_lookup);
/**
* ip6_dst_lookup_flow - perform route lookup on flow with ipsec
* @sk: socket which provides route info
* @fl6: flow to lookup
* @final_dst: final destination address for ipsec lookup
* @can_sleep: we are in a sleepable context
*
* This function performs a route lookup on the given flow.
*
* It returns a valid dst pointer on success, or a pointer encoded
* error code.
*/
struct dst_entry *ip6_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6,
const struct in6_addr *final_dst,
bool can_sleep)
{
struct dst_entry *dst = NULL;
int err;
err = ip6_dst_lookup_tail(sk, &dst, fl6);
if (err)
return ERR_PTR(err);
if (final_dst)
fl6->daddr = *final_dst;
if (can_sleep)
fl6->flowi6_flags |= FLOWI_FLAG_CAN_SLEEP;
return xfrm_lookup(sock_net(sk), dst, flowi6_to_flowi(fl6), sk, 0);
}
EXPORT_SYMBOL_GPL(ip6_dst_lookup_flow);
/**
* ip6_sk_dst_lookup_flow - perform socket cached route lookup on flow
* @sk: socket which provides the dst cache and route info
* @fl6: flow to lookup
* @final_dst: final destination address for ipsec lookup
* @can_sleep: we are in a sleepable context
*
* This function performs a route lookup on the given flow with the
* possibility of using the cached route in the socket if it is valid.
* It will take the socket dst lock when operating on the dst cache.
* As a result, this function can only be used in process context.
*
* It returns a valid dst pointer on success, or a pointer encoded
* error code.
*/
struct dst_entry *ip6_sk_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6,
const struct in6_addr *final_dst,
bool can_sleep)
{
struct dst_entry *dst = sk_dst_check(sk, inet6_sk(sk)->dst_cookie);
int err;
dst = ip6_sk_dst_check(sk, dst, fl6);
err = ip6_dst_lookup_tail(sk, &dst, fl6);
if (err)
return ERR_PTR(err);
if (final_dst)
fl6->daddr = *final_dst;
if (can_sleep)
fl6->flowi6_flags |= FLOWI_FLAG_CAN_SLEEP;
return xfrm_lookup(sock_net(sk), dst, flowi6_to_flowi(fl6), sk, 0);
}
EXPORT_SYMBOL_GPL(ip6_sk_dst_lookup_flow);
static inline int ip6_ufo_append_data(struct sock *sk,
[IPv4/IPv6]: UFO Scatter-gather approach Attached is kernel patch for UDP Fragmentation Offload (UFO) feature. 1. This patch incorporate the review comments by Jeff Garzik. 2. Renamed USO as UFO (UDP Fragmentation Offload) 3. udp sendfile support with UFO This patches uses scatter-gather feature of skb to generate large UDP datagram. Below is a "how-to" on changes required in network device driver to use the UFO interface. UDP Fragmentation Offload (UFO) Interface: ------------------------------------------- UFO is a feature wherein the Linux kernel network stack will offload the IP fragmentation functionality of large UDP datagram to hardware. This will reduce the overhead of stack in fragmenting the large UDP datagram to MTU sized packets 1) Drivers indicate their capability of UFO using dev->features |= NETIF_F_UFO | NETIF_F_HW_CSUM | NETIF_F_SG NETIF_F_HW_CSUM is required for UFO over ipv6. 2) UFO packet will be submitted for transmission using driver xmit routine. UFO packet will have a non-zero value for "skb_shinfo(skb)->ufo_size" skb_shinfo(skb)->ufo_size will indicate the length of data part in each IP fragment going out of the adapter after IP fragmentation by hardware. skb->data will contain MAC/IP/UDP header and skb_shinfo(skb)->frags[] contains the data payload. The skb->ip_summed will be set to CHECKSUM_HW indicating that hardware has to do checksum calculation. Hardware should compute the UDP checksum of complete datagram and also ip header checksum of each fragmented IP packet. For IPV6 the UFO provides the fragment identification-id in skb_shinfo(skb)->ip6_frag_id. The adapter should use this ID for generating IPv6 fragments. Signed-off-by: Ananda Raju <ananda.raju@neterion.com> Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> (forwarded) Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
2005-10-19 06:46:41 +08:00
int getfrag(void *from, char *to, int offset, int len,
int odd, struct sk_buff *skb),
void *from, int length, int hh_len, int fragheaderlen,
int transhdrlen, int mtu,unsigned int flags,
struct rt6_info *rt)
[IPv4/IPv6]: UFO Scatter-gather approach Attached is kernel patch for UDP Fragmentation Offload (UFO) feature. 1. This patch incorporate the review comments by Jeff Garzik. 2. Renamed USO as UFO (UDP Fragmentation Offload) 3. udp sendfile support with UFO This patches uses scatter-gather feature of skb to generate large UDP datagram. Below is a "how-to" on changes required in network device driver to use the UFO interface. UDP Fragmentation Offload (UFO) Interface: ------------------------------------------- UFO is a feature wherein the Linux kernel network stack will offload the IP fragmentation functionality of large UDP datagram to hardware. This will reduce the overhead of stack in fragmenting the large UDP datagram to MTU sized packets 1) Drivers indicate their capability of UFO using dev->features |= NETIF_F_UFO | NETIF_F_HW_CSUM | NETIF_F_SG NETIF_F_HW_CSUM is required for UFO over ipv6. 2) UFO packet will be submitted for transmission using driver xmit routine. UFO packet will have a non-zero value for "skb_shinfo(skb)->ufo_size" skb_shinfo(skb)->ufo_size will indicate the length of data part in each IP fragment going out of the adapter after IP fragmentation by hardware. skb->data will contain MAC/IP/UDP header and skb_shinfo(skb)->frags[] contains the data payload. The skb->ip_summed will be set to CHECKSUM_HW indicating that hardware has to do checksum calculation. Hardware should compute the UDP checksum of complete datagram and also ip header checksum of each fragmented IP packet. For IPV6 the UFO provides the fragment identification-id in skb_shinfo(skb)->ip6_frag_id. The adapter should use this ID for generating IPv6 fragments. Signed-off-by: Ananda Raju <ananda.raju@neterion.com> Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> (forwarded) Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
2005-10-19 06:46:41 +08:00
{
struct sk_buff *skb;
int err;
/* There is support for UDP large send offload by network
* device, so create one single skb packet containing complete
* udp datagram
*/
if ((skb = skb_peek_tail(&sk->sk_write_queue)) == NULL) {
skb = sock_alloc_send_skb(sk,
hh_len + fragheaderlen + transhdrlen + 20,
(flags & MSG_DONTWAIT), &err);
if (skb == NULL)
return err;
[IPv4/IPv6]: UFO Scatter-gather approach Attached is kernel patch for UDP Fragmentation Offload (UFO) feature. 1. This patch incorporate the review comments by Jeff Garzik. 2. Renamed USO as UFO (UDP Fragmentation Offload) 3. udp sendfile support with UFO This patches uses scatter-gather feature of skb to generate large UDP datagram. Below is a "how-to" on changes required in network device driver to use the UFO interface. UDP Fragmentation Offload (UFO) Interface: ------------------------------------------- UFO is a feature wherein the Linux kernel network stack will offload the IP fragmentation functionality of large UDP datagram to hardware. This will reduce the overhead of stack in fragmenting the large UDP datagram to MTU sized packets 1) Drivers indicate their capability of UFO using dev->features |= NETIF_F_UFO | NETIF_F_HW_CSUM | NETIF_F_SG NETIF_F_HW_CSUM is required for UFO over ipv6. 2) UFO packet will be submitted for transmission using driver xmit routine. UFO packet will have a non-zero value for "skb_shinfo(skb)->ufo_size" skb_shinfo(skb)->ufo_size will indicate the length of data part in each IP fragment going out of the adapter after IP fragmentation by hardware. skb->data will contain MAC/IP/UDP header and skb_shinfo(skb)->frags[] contains the data payload. The skb->ip_summed will be set to CHECKSUM_HW indicating that hardware has to do checksum calculation. Hardware should compute the UDP checksum of complete datagram and also ip header checksum of each fragmented IP packet. For IPV6 the UFO provides the fragment identification-id in skb_shinfo(skb)->ip6_frag_id. The adapter should use this ID for generating IPv6 fragments. Signed-off-by: Ananda Raju <ananda.raju@neterion.com> Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> (forwarded) Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
2005-10-19 06:46:41 +08:00
/* reserve space for Hardware header */
skb_reserve(skb, hh_len);
/* create space for UDP/IP header */
skb_put(skb,fragheaderlen + transhdrlen);
/* initialize network header pointer */
skb_reset_network_header(skb);
[IPv4/IPv6]: UFO Scatter-gather approach Attached is kernel patch for UDP Fragmentation Offload (UFO) feature. 1. This patch incorporate the review comments by Jeff Garzik. 2. Renamed USO as UFO (UDP Fragmentation Offload) 3. udp sendfile support with UFO This patches uses scatter-gather feature of skb to generate large UDP datagram. Below is a "how-to" on changes required in network device driver to use the UFO interface. UDP Fragmentation Offload (UFO) Interface: ------------------------------------------- UFO is a feature wherein the Linux kernel network stack will offload the IP fragmentation functionality of large UDP datagram to hardware. This will reduce the overhead of stack in fragmenting the large UDP datagram to MTU sized packets 1) Drivers indicate their capability of UFO using dev->features |= NETIF_F_UFO | NETIF_F_HW_CSUM | NETIF_F_SG NETIF_F_HW_CSUM is required for UFO over ipv6. 2) UFO packet will be submitted for transmission using driver xmit routine. UFO packet will have a non-zero value for "skb_shinfo(skb)->ufo_size" skb_shinfo(skb)->ufo_size will indicate the length of data part in each IP fragment going out of the adapter after IP fragmentation by hardware. skb->data will contain MAC/IP/UDP header and skb_shinfo(skb)->frags[] contains the data payload. The skb->ip_summed will be set to CHECKSUM_HW indicating that hardware has to do checksum calculation. Hardware should compute the UDP checksum of complete datagram and also ip header checksum of each fragmented IP packet. For IPV6 the UFO provides the fragment identification-id in skb_shinfo(skb)->ip6_frag_id. The adapter should use this ID for generating IPv6 fragments. Signed-off-by: Ananda Raju <ananda.raju@neterion.com> Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> (forwarded) Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
2005-10-19 06:46:41 +08:00
/* initialize protocol header pointer */
skb->transport_header = skb->network_header + fragheaderlen;
[IPv4/IPv6]: UFO Scatter-gather approach Attached is kernel patch for UDP Fragmentation Offload (UFO) feature. 1. This patch incorporate the review comments by Jeff Garzik. 2. Renamed USO as UFO (UDP Fragmentation Offload) 3. udp sendfile support with UFO This patches uses scatter-gather feature of skb to generate large UDP datagram. Below is a "how-to" on changes required in network device driver to use the UFO interface. UDP Fragmentation Offload (UFO) Interface: ------------------------------------------- UFO is a feature wherein the Linux kernel network stack will offload the IP fragmentation functionality of large UDP datagram to hardware. This will reduce the overhead of stack in fragmenting the large UDP datagram to MTU sized packets 1) Drivers indicate their capability of UFO using dev->features |= NETIF_F_UFO | NETIF_F_HW_CSUM | NETIF_F_SG NETIF_F_HW_CSUM is required for UFO over ipv6. 2) UFO packet will be submitted for transmission using driver xmit routine. UFO packet will have a non-zero value for "skb_shinfo(skb)->ufo_size" skb_shinfo(skb)->ufo_size will indicate the length of data part in each IP fragment going out of the adapter after IP fragmentation by hardware. skb->data will contain MAC/IP/UDP header and skb_shinfo(skb)->frags[] contains the data payload. The skb->ip_summed will be set to CHECKSUM_HW indicating that hardware has to do checksum calculation. Hardware should compute the UDP checksum of complete datagram and also ip header checksum of each fragmented IP packet. For IPV6 the UFO provides the fragment identification-id in skb_shinfo(skb)->ip6_frag_id. The adapter should use this ID for generating IPv6 fragments. Signed-off-by: Ananda Raju <ananda.raju@neterion.com> Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> (forwarded) Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
2005-10-19 06:46:41 +08:00
skb->ip_summed = CHECKSUM_PARTIAL;
[IPv4/IPv6]: UFO Scatter-gather approach Attached is kernel patch for UDP Fragmentation Offload (UFO) feature. 1. This patch incorporate the review comments by Jeff Garzik. 2. Renamed USO as UFO (UDP Fragmentation Offload) 3. udp sendfile support with UFO This patches uses scatter-gather feature of skb to generate large UDP datagram. Below is a "how-to" on changes required in network device driver to use the UFO interface. UDP Fragmentation Offload (UFO) Interface: ------------------------------------------- UFO is a feature wherein the Linux kernel network stack will offload the IP fragmentation functionality of large UDP datagram to hardware. This will reduce the overhead of stack in fragmenting the large UDP datagram to MTU sized packets 1) Drivers indicate their capability of UFO using dev->features |= NETIF_F_UFO | NETIF_F_HW_CSUM | NETIF_F_SG NETIF_F_HW_CSUM is required for UFO over ipv6. 2) UFO packet will be submitted for transmission using driver xmit routine. UFO packet will have a non-zero value for "skb_shinfo(skb)->ufo_size" skb_shinfo(skb)->ufo_size will indicate the length of data part in each IP fragment going out of the adapter after IP fragmentation by hardware. skb->data will contain MAC/IP/UDP header and skb_shinfo(skb)->frags[] contains the data payload. The skb->ip_summed will be set to CHECKSUM_HW indicating that hardware has to do checksum calculation. Hardware should compute the UDP checksum of complete datagram and also ip header checksum of each fragmented IP packet. For IPV6 the UFO provides the fragment identification-id in skb_shinfo(skb)->ip6_frag_id. The adapter should use this ID for generating IPv6 fragments. Signed-off-by: Ananda Raju <ananda.raju@neterion.com> Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> (forwarded) Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
2005-10-19 06:46:41 +08:00
skb->csum = 0;
}
err = skb_append_datato_frags(sk,skb, getfrag, from,
(length - transhdrlen));
if (!err) {
struct frag_hdr fhdr;
/* Specify the length of each IPv6 datagram fragment.
* It has to be a multiple of 8.
*/
skb_shinfo(skb)->gso_size = (mtu - fragheaderlen -
sizeof(struct frag_hdr)) & ~7;
[IPV6]: Added GSO support for TCPv6 This patch adds GSO support for IPv6 and TCPv6. This is based on a patch by Ananda Raju <Ananda.Raju@neterion.com>. His original description is: This patch enables TSO over IPv6. Currently Linux network stacks restricts TSO over IPv6 by clearing of the NETIF_F_TSO bit from "dev->features". This patch will remove this restriction. This patch will introduce a new flag NETIF_F_TSO6 which will be used to check whether device supports TSO over IPv6. If device support TSO over IPv6 then we don't clear of NETIF_F_TSO and which will make the TCP layer to create TSO packets. Any device supporting TSO over IPv6 will set NETIF_F_TSO6 flag in "dev->features" along with NETIF_F_TSO. In case when user disables TSO using ethtool, NETIF_F_TSO will get cleared from "dev->features". So even if we have NETIF_F_TSO6 we don't get TSO packets created by TCP layer. SKB_GSO_TCPV4 renamed to SKB_GSO_TCP to make it generic GSO packet. SKB_GSO_UDPV4 renamed to SKB_GSO_UDP as UFO is not a IPv4 feature. UFO is supported over IPv6 also The following table shows there is significant improvement in throughput with normal frames and CPU usage for both normal and jumbo. -------------------------------------------------- | | 1500 | 9600 | | ------------------|-------------------| | | thru CPU | thru CPU | -------------------------------------------------- | TSO OFF | 2.00 5.5% id | 5.66 20.0% id | -------------------------------------------------- | TSO ON | 2.63 78.0 id | 5.67 39.0% id | -------------------------------------------------- Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2006-07-01 04:37:03 +08:00
skb_shinfo(skb)->gso_type = SKB_GSO_UDP;
ipv6_select_ident(&fhdr, rt);
[IPv4/IPv6]: UFO Scatter-gather approach Attached is kernel patch for UDP Fragmentation Offload (UFO) feature. 1. This patch incorporate the review comments by Jeff Garzik. 2. Renamed USO as UFO (UDP Fragmentation Offload) 3. udp sendfile support with UFO This patches uses scatter-gather feature of skb to generate large UDP datagram. Below is a "how-to" on changes required in network device driver to use the UFO interface. UDP Fragmentation Offload (UFO) Interface: ------------------------------------------- UFO is a feature wherein the Linux kernel network stack will offload the IP fragmentation functionality of large UDP datagram to hardware. This will reduce the overhead of stack in fragmenting the large UDP datagram to MTU sized packets 1) Drivers indicate their capability of UFO using dev->features |= NETIF_F_UFO | NETIF_F_HW_CSUM | NETIF_F_SG NETIF_F_HW_CSUM is required for UFO over ipv6. 2) UFO packet will be submitted for transmission using driver xmit routine. UFO packet will have a non-zero value for "skb_shinfo(skb)->ufo_size" skb_shinfo(skb)->ufo_size will indicate the length of data part in each IP fragment going out of the adapter after IP fragmentation by hardware. skb->data will contain MAC/IP/UDP header and skb_shinfo(skb)->frags[] contains the data payload. The skb->ip_summed will be set to CHECKSUM_HW indicating that hardware has to do checksum calculation. Hardware should compute the UDP checksum of complete datagram and also ip header checksum of each fragmented IP packet. For IPV6 the UFO provides the fragment identification-id in skb_shinfo(skb)->ip6_frag_id. The adapter should use this ID for generating IPv6 fragments. Signed-off-by: Ananda Raju <ananda.raju@neterion.com> Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> (forwarded) Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
2005-10-19 06:46:41 +08:00
skb_shinfo(skb)->ip6_frag_id = fhdr.identification;
__skb_queue_tail(&sk->sk_write_queue, skb);
return 0;
}
/* There is not enough support do UPD LSO,
* so follow normal path
*/
kfree_skb(skb);
return err;
}
static inline struct ipv6_opt_hdr *ip6_opt_dup(struct ipv6_opt_hdr *src,
gfp_t gfp)
{
return src ? kmemdup(src, (src->hdrlen + 1) * 8, gfp) : NULL;
}
static inline struct ipv6_rt_hdr *ip6_rthdr_dup(struct ipv6_rt_hdr *src,
gfp_t gfp)
{
return src ? kmemdup(src, (src->hdrlen + 1) * 8, gfp) : NULL;
}
static void ip6_append_data_mtu(int *mtu,
int *maxfraglen,
unsigned int fragheaderlen,
struct sk_buff *skb,
struct rt6_info *rt)
{
if (!(rt->dst.flags & DST_XFRM_TUNNEL)) {
if (skb == NULL) {
/* first fragment, reserve header_len */
*mtu = *mtu - rt->dst.header_len;
} else {
/*
* this fragment is not first, the headers
* space is regarded as data space.
*/
*mtu = dst_mtu(rt->dst.path);
}
*maxfraglen = ((*mtu - fragheaderlen) & ~7)
+ fragheaderlen - sizeof(struct frag_hdr);
}
}
int ip6_append_data(struct sock *sk, int getfrag(void *from, char *to,
int offset, int len, int odd, struct sk_buff *skb),
void *from, int length, int transhdrlen,
int hlimit, int tclass, struct ipv6_txoptions *opt, struct flowi6 *fl6,
struct rt6_info *rt, unsigned int flags, int dontfrag)
{
struct inet_sock *inet = inet_sk(sk);
struct ipv6_pinfo *np = inet6_sk(sk);
struct inet_cork *cork;
struct sk_buff *skb, *skb_prev = NULL;
unsigned int maxfraglen, fragheaderlen;
int exthdrlen;
int dst_exthdrlen;
int hh_len;
int mtu;
int copy;
int err;
int offset = 0;
__u8 tx_flags = 0;
if (flags&MSG_PROBE)
return 0;
cork = &inet->cork.base;
if (skb_queue_empty(&sk->sk_write_queue)) {
/*
* setup for corking
*/
if (opt) {
if (WARN_ON(np->cork.opt))
return -EINVAL;
np->cork.opt = kmalloc(opt->tot_len, sk->sk_allocation);
if (unlikely(np->cork.opt == NULL))
return -ENOBUFS;
np->cork.opt->tot_len = opt->tot_len;
np->cork.opt->opt_flen = opt->opt_flen;
np->cork.opt->opt_nflen = opt->opt_nflen;
np->cork.opt->dst0opt = ip6_opt_dup(opt->dst0opt,
sk->sk_allocation);
if (opt->dst0opt && !np->cork.opt->dst0opt)
return -ENOBUFS;
np->cork.opt->dst1opt = ip6_opt_dup(opt->dst1opt,
sk->sk_allocation);
if (opt->dst1opt && !np->cork.opt->dst1opt)
return -ENOBUFS;
np->cork.opt->hopopt = ip6_opt_dup(opt->hopopt,
sk->sk_allocation);
if (opt->hopopt && !np->cork.opt->hopopt)
return -ENOBUFS;
np->cork.opt->srcrt = ip6_rthdr_dup(opt->srcrt,
sk->sk_allocation);
if (opt->srcrt && !np->cork.opt->srcrt)
return -ENOBUFS;
/* need source address above miyazawa*/
}
dst_hold(&rt->dst);
cork->dst = &rt->dst;
inet->cork.fl.u.ip6 = *fl6;
np->cork.hop_limit = hlimit;
np->cork.tclass = tclass;
if (rt->dst.flags & DST_XFRM_TUNNEL)
mtu = np->pmtudisc == IPV6_PMTUDISC_PROBE ?
rt->dst.dev->mtu : dst_mtu(&rt->dst);
else
mtu = np->pmtudisc == IPV6_PMTUDISC_PROBE ?
rt->dst.dev->mtu : dst_mtu(rt->dst.path);
if (np->frag_size < mtu) {
if (np->frag_size)
mtu = np->frag_size;
}
cork->fragsize = mtu;
if (dst_allfrag(rt->dst.path))
cork->flags |= IPCORK_ALLFRAG;
cork->length = 0;
exthdrlen = (opt ? opt->opt_flen : 0) - rt->rt6i_nfheader_len;
length += exthdrlen;
transhdrlen += exthdrlen;
dst_exthdrlen = rt->dst.header_len;
} else {
rt = (struct rt6_info *)cork->dst;
fl6 = &inet->cork.fl.u.ip6;
opt = np->cork.opt;
transhdrlen = 0;
exthdrlen = 0;
dst_exthdrlen = 0;
mtu = cork->fragsize;
}
hh_len = LL_RESERVED_SPACE(rt->dst.dev);
fragheaderlen = sizeof(struct ipv6hdr) + rt->rt6i_nfheader_len +
(opt ? opt->opt_nflen : 0);
maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen - sizeof(struct frag_hdr);
if (mtu <= sizeof(struct ipv6hdr) + IPV6_MAXPLEN) {
if (cork->length + length > sizeof(struct ipv6hdr) + IPV6_MAXPLEN - fragheaderlen) {
ipv6_local_error(sk, EMSGSIZE, fl6, mtu-exthdrlen);
return -EMSGSIZE;
}
}
/* For UDP, check if TX timestamp is enabled */
if (sk->sk_type == SOCK_DGRAM) {
err = sock_tx_timestamp(sk, &tx_flags);
if (err)
goto error;
}
/*
* Let's try using as much space as possible.
* Use MTU if total length of the message fits into the MTU.
* Otherwise, we need to reserve fragment header and
* fragment alignment (= 8-15 octects, in total).
*
* Note that we may need to "move" the data from the tail of
* of the buffer to the new fragment when we split
* the message.
*
* FIXME: It may be fragmented into multiple chunks
* at once if non-fragmentable extension headers
* are too large.
* --yoshfuji
*/
cork->length += length;
if (length > mtu) {
int proto = sk->sk_protocol;
if (dontfrag && (proto == IPPROTO_UDP || proto == IPPROTO_RAW)){
ipv6_local_rxpmtu(sk, fl6, mtu-exthdrlen);
return -EMSGSIZE;
}
[IPv4/IPv6]: UFO Scatter-gather approach Attached is kernel patch for UDP Fragmentation Offload (UFO) feature. 1. This patch incorporate the review comments by Jeff Garzik. 2. Renamed USO as UFO (UDP Fragmentation Offload) 3. udp sendfile support with UFO This patches uses scatter-gather feature of skb to generate large UDP datagram. Below is a "how-to" on changes required in network device driver to use the UFO interface. UDP Fragmentation Offload (UFO) Interface: ------------------------------------------- UFO is a feature wherein the Linux kernel network stack will offload the IP fragmentation functionality of large UDP datagram to hardware. This will reduce the overhead of stack in fragmenting the large UDP datagram to MTU sized packets 1) Drivers indicate their capability of UFO using dev->features |= NETIF_F_UFO | NETIF_F_HW_CSUM | NETIF_F_SG NETIF_F_HW_CSUM is required for UFO over ipv6. 2) UFO packet will be submitted for transmission using driver xmit routine. UFO packet will have a non-zero value for "skb_shinfo(skb)->ufo_size" skb_shinfo(skb)->ufo_size will indicate the length of data part in each IP fragment going out of the adapter after IP fragmentation by hardware. skb->data will contain MAC/IP/UDP header and skb_shinfo(skb)->frags[] contains the data payload. The skb->ip_summed will be set to CHECKSUM_HW indicating that hardware has to do checksum calculation. Hardware should compute the UDP checksum of complete datagram and also ip header checksum of each fragmented IP packet. For IPV6 the UFO provides the fragment identification-id in skb_shinfo(skb)->ip6_frag_id. The adapter should use this ID for generating IPv6 fragments. Signed-off-by: Ananda Raju <ananda.raju@neterion.com> Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> (forwarded) Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
2005-10-19 06:46:41 +08:00
if (proto == IPPROTO_UDP &&
(rt->dst.dev->features & NETIF_F_UFO)) {
err = ip6_ufo_append_data(sk, getfrag, from, length,
hh_len, fragheaderlen,
transhdrlen, mtu, flags, rt);
if (err)
goto error;
return 0;
}
[IPv4/IPv6]: UFO Scatter-gather approach Attached is kernel patch for UDP Fragmentation Offload (UFO) feature. 1. This patch incorporate the review comments by Jeff Garzik. 2. Renamed USO as UFO (UDP Fragmentation Offload) 3. udp sendfile support with UFO This patches uses scatter-gather feature of skb to generate large UDP datagram. Below is a "how-to" on changes required in network device driver to use the UFO interface. UDP Fragmentation Offload (UFO) Interface: ------------------------------------------- UFO is a feature wherein the Linux kernel network stack will offload the IP fragmentation functionality of large UDP datagram to hardware. This will reduce the overhead of stack in fragmenting the large UDP datagram to MTU sized packets 1) Drivers indicate their capability of UFO using dev->features |= NETIF_F_UFO | NETIF_F_HW_CSUM | NETIF_F_SG NETIF_F_HW_CSUM is required for UFO over ipv6. 2) UFO packet will be submitted for transmission using driver xmit routine. UFO packet will have a non-zero value for "skb_shinfo(skb)->ufo_size" skb_shinfo(skb)->ufo_size will indicate the length of data part in each IP fragment going out of the adapter after IP fragmentation by hardware. skb->data will contain MAC/IP/UDP header and skb_shinfo(skb)->frags[] contains the data payload. The skb->ip_summed will be set to CHECKSUM_HW indicating that hardware has to do checksum calculation. Hardware should compute the UDP checksum of complete datagram and also ip header checksum of each fragmented IP packet. For IPV6 the UFO provides the fragment identification-id in skb_shinfo(skb)->ip6_frag_id. The adapter should use this ID for generating IPv6 fragments. Signed-off-by: Ananda Raju <ananda.raju@neterion.com> Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> (forwarded) Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
2005-10-19 06:46:41 +08:00
}
if ((skb = skb_peek_tail(&sk->sk_write_queue)) == NULL)
goto alloc_new_skb;
while (length > 0) {
/* Check if the remaining data fits into current packet. */
copy = (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - skb->len;
if (copy < length)
copy = maxfraglen - skb->len;
if (copy <= 0) {
char *data;
unsigned int datalen;
unsigned int fraglen;
unsigned int fraggap;
unsigned int alloclen;
alloc_new_skb:
/* There's no room in the current skb */
if (skb)
fraggap = skb->len - maxfraglen;
else
fraggap = 0;
/* update mtu and maxfraglen if necessary */
if (skb == NULL || skb_prev == NULL)
ip6_append_data_mtu(&mtu, &maxfraglen,
fragheaderlen, skb, rt);
skb_prev = skb;
/*
* If remaining data exceeds the mtu,
* we know we need more fragment(s).
*/
datalen = length + fraggap;
if (datalen > (cork->length <= mtu && !(cork->flags & IPCORK_ALLFRAG) ? mtu : maxfraglen) - fragheaderlen)
datalen = maxfraglen - fragheaderlen - rt->dst.trailer_len;
if ((flags & MSG_MORE) &&
!(rt->dst.dev->features&NETIF_F_SG))
alloclen = mtu;
else
alloclen = datalen + fragheaderlen;
alloclen += dst_exthdrlen;
if (datalen != length + fraggap) {
/*
* this is not the last fragment, the trailer
* space is regarded as data space.
*/
datalen += rt->dst.trailer_len;
}
alloclen += rt->dst.trailer_len;
fraglen = datalen + fragheaderlen;
/*
* We just reserve space for fragment header.
* Note: this may be overallocation if the message
* (without MSG_MORE) fits into the MTU.
*/
alloclen += sizeof(struct frag_hdr);
if (transhdrlen) {
skb = sock_alloc_send_skb(sk,
alloclen + hh_len,
(flags & MSG_DONTWAIT), &err);
} else {
skb = NULL;
if (atomic_read(&sk->sk_wmem_alloc) <=
2 * sk->sk_sndbuf)
skb = sock_wmalloc(sk,
alloclen + hh_len, 1,
sk->sk_allocation);
if (unlikely(skb == NULL))
err = -ENOBUFS;
else {
/* Only the initial fragment
* is time stamped.
*/
tx_flags = 0;
}
}
if (skb == NULL)
goto error;
/*
* Fill in the control structures
*/
skb->ip_summed = CHECKSUM_NONE;
skb->csum = 0;
/* reserve for fragmentation and ipsec header */
skb_reserve(skb, hh_len + sizeof(struct frag_hdr) +
dst_exthdrlen);
if (sk->sk_type == SOCK_DGRAM)
skb_shinfo(skb)->tx_flags = tx_flags;
/*
* Find where to start putting bytes
*/
data = skb_put(skb, fraglen);
skb_set_network_header(skb, exthdrlen);
data += fragheaderlen;
skb->transport_header = (skb->network_header +
fragheaderlen);
if (fraggap) {
skb->csum = skb_copy_and_csum_bits(
skb_prev, maxfraglen,
data + transhdrlen, fraggap, 0);
skb_prev->csum = csum_sub(skb_prev->csum,
skb->csum);
data += fraggap;
pskb_trim_unique(skb_prev, maxfraglen);
}
copy = datalen - transhdrlen - fraggap;
if (copy < 0) {
err = -EINVAL;
kfree_skb(skb);
goto error;
} else if (copy > 0 && getfrag(from, data + transhdrlen, offset, copy, fraggap, skb) < 0) {
err = -EFAULT;
kfree_skb(skb);
goto error;
}
offset += copy;
length -= datalen - fraggap;
transhdrlen = 0;
exthdrlen = 0;
dst_exthdrlen = 0;
/*
* Put the packet on the pending queue
*/
__skb_queue_tail(&sk->sk_write_queue, skb);
continue;
}
if (copy > length)
copy = length;
if (!(rt->dst.dev->features&NETIF_F_SG)) {
unsigned int off;
off = skb->len;
if (getfrag(from, skb_put(skb, copy),
offset, copy, off, skb) < 0) {
__skb_trim(skb, off);
err = -EFAULT;
goto error;
}
} else {
int i = skb_shinfo(skb)->nr_frags;
net: use a per task frag allocator We currently use a per socket order-0 page cache for tcp_sendmsg() operations. This page is used to build fragments for skbs. Its done to increase probability of coalescing small write() into single segments in skbs still in write queue (not yet sent) But it wastes a lot of memory for applications handling many mostly idle sockets, since each socket holds one page in sk->sk_sndmsg_page Its also quite inefficient to build TSO 64KB packets, because we need about 16 pages per skb on arches where PAGE_SIZE = 4096, so we hit page allocator more than wanted. This patch adds a per task frag allocator and uses bigger pages, if available. An automatic fallback is done in case of memory pressure. (up to 32768 bytes per frag, thats order-3 pages on x86) This increases TCP stream performance by 20% on loopback device, but also benefits on other network devices, since 8x less frags are mapped on transmit and unmapped on tx completion. Alexander Duyck mentioned a probable performance win on systems with IOMMU enabled. Its possible some SG enabled hardware cant cope with bigger fragments, but their ndo_start_xmit() should already handle this, splitting a fragment in sub fragments, since some arches have PAGE_SIZE=65536 Successfully tested on various ethernet devices. (ixgbe, igb, bnx2x, tg3, mellanox mlx4) Signed-off-by: Eric Dumazet <edumazet@google.com> Cc: Ben Hutchings <bhutchings@solarflare.com> Cc: Vijay Subramanian <subramanian.vijay@gmail.com> Cc: Alexander Duyck <alexander.h.duyck@intel.com> Tested-by: Vijay Subramanian <subramanian.vijay@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2012-09-24 07:04:42 +08:00
struct page_frag *pfrag = sk_page_frag(sk);
net: use a per task frag allocator We currently use a per socket order-0 page cache for tcp_sendmsg() operations. This page is used to build fragments for skbs. Its done to increase probability of coalescing small write() into single segments in skbs still in write queue (not yet sent) But it wastes a lot of memory for applications handling many mostly idle sockets, since each socket holds one page in sk->sk_sndmsg_page Its also quite inefficient to build TSO 64KB packets, because we need about 16 pages per skb on arches where PAGE_SIZE = 4096, so we hit page allocator more than wanted. This patch adds a per task frag allocator and uses bigger pages, if available. An automatic fallback is done in case of memory pressure. (up to 32768 bytes per frag, thats order-3 pages on x86) This increases TCP stream performance by 20% on loopback device, but also benefits on other network devices, since 8x less frags are mapped on transmit and unmapped on tx completion. Alexander Duyck mentioned a probable performance win on systems with IOMMU enabled. Its possible some SG enabled hardware cant cope with bigger fragments, but their ndo_start_xmit() should already handle this, splitting a fragment in sub fragments, since some arches have PAGE_SIZE=65536 Successfully tested on various ethernet devices. (ixgbe, igb, bnx2x, tg3, mellanox mlx4) Signed-off-by: Eric Dumazet <edumazet@google.com> Cc: Ben Hutchings <bhutchings@solarflare.com> Cc: Vijay Subramanian <subramanian.vijay@gmail.com> Cc: Alexander Duyck <alexander.h.duyck@intel.com> Tested-by: Vijay Subramanian <subramanian.vijay@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2012-09-24 07:04:42 +08:00
err = -ENOMEM;
if (!sk_page_frag_refill(sk, pfrag))
goto error;
net: use a per task frag allocator We currently use a per socket order-0 page cache for tcp_sendmsg() operations. This page is used to build fragments for skbs. Its done to increase probability of coalescing small write() into single segments in skbs still in write queue (not yet sent) But it wastes a lot of memory for applications handling many mostly idle sockets, since each socket holds one page in sk->sk_sndmsg_page Its also quite inefficient to build TSO 64KB packets, because we need about 16 pages per skb on arches where PAGE_SIZE = 4096, so we hit page allocator more than wanted. This patch adds a per task frag allocator and uses bigger pages, if available. An automatic fallback is done in case of memory pressure. (up to 32768 bytes per frag, thats order-3 pages on x86) This increases TCP stream performance by 20% on loopback device, but also benefits on other network devices, since 8x less frags are mapped on transmit and unmapped on tx completion. Alexander Duyck mentioned a probable performance win on systems with IOMMU enabled. Its possible some SG enabled hardware cant cope with bigger fragments, but their ndo_start_xmit() should already handle this, splitting a fragment in sub fragments, since some arches have PAGE_SIZE=65536 Successfully tested on various ethernet devices. (ixgbe, igb, bnx2x, tg3, mellanox mlx4) Signed-off-by: Eric Dumazet <edumazet@google.com> Cc: Ben Hutchings <bhutchings@solarflare.com> Cc: Vijay Subramanian <subramanian.vijay@gmail.com> Cc: Alexander Duyck <alexander.h.duyck@intel.com> Tested-by: Vijay Subramanian <subramanian.vijay@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2012-09-24 07:04:42 +08:00
if (!skb_can_coalesce(skb, i, pfrag->page,
pfrag->offset)) {
err = -EMSGSIZE;
if (i == MAX_SKB_FRAGS)
goto error;
__skb_fill_page_desc(skb, i, pfrag->page,
pfrag->offset, 0);
skb_shinfo(skb)->nr_frags = ++i;
get_page(pfrag->page);
}
net: use a per task frag allocator We currently use a per socket order-0 page cache for tcp_sendmsg() operations. This page is used to build fragments for skbs. Its done to increase probability of coalescing small write() into single segments in skbs still in write queue (not yet sent) But it wastes a lot of memory for applications handling many mostly idle sockets, since each socket holds one page in sk->sk_sndmsg_page Its also quite inefficient to build TSO 64KB packets, because we need about 16 pages per skb on arches where PAGE_SIZE = 4096, so we hit page allocator more than wanted. This patch adds a per task frag allocator and uses bigger pages, if available. An automatic fallback is done in case of memory pressure. (up to 32768 bytes per frag, thats order-3 pages on x86) This increases TCP stream performance by 20% on loopback device, but also benefits on other network devices, since 8x less frags are mapped on transmit and unmapped on tx completion. Alexander Duyck mentioned a probable performance win on systems with IOMMU enabled. Its possible some SG enabled hardware cant cope with bigger fragments, but their ndo_start_xmit() should already handle this, splitting a fragment in sub fragments, since some arches have PAGE_SIZE=65536 Successfully tested on various ethernet devices. (ixgbe, igb, bnx2x, tg3, mellanox mlx4) Signed-off-by: Eric Dumazet <edumazet@google.com> Cc: Ben Hutchings <bhutchings@solarflare.com> Cc: Vijay Subramanian <subramanian.vijay@gmail.com> Cc: Alexander Duyck <alexander.h.duyck@intel.com> Tested-by: Vijay Subramanian <subramanian.vijay@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2012-09-24 07:04:42 +08:00
copy = min_t(int, copy, pfrag->size - pfrag->offset);
if (getfrag(from,
net: use a per task frag allocator We currently use a per socket order-0 page cache for tcp_sendmsg() operations. This page is used to build fragments for skbs. Its done to increase probability of coalescing small write() into single segments in skbs still in write queue (not yet sent) But it wastes a lot of memory for applications handling many mostly idle sockets, since each socket holds one page in sk->sk_sndmsg_page Its also quite inefficient to build TSO 64KB packets, because we need about 16 pages per skb on arches where PAGE_SIZE = 4096, so we hit page allocator more than wanted. This patch adds a per task frag allocator and uses bigger pages, if available. An automatic fallback is done in case of memory pressure. (up to 32768 bytes per frag, thats order-3 pages on x86) This increases TCP stream performance by 20% on loopback device, but also benefits on other network devices, since 8x less frags are mapped on transmit and unmapped on tx completion. Alexander Duyck mentioned a probable performance win on systems with IOMMU enabled. Its possible some SG enabled hardware cant cope with bigger fragments, but their ndo_start_xmit() should already handle this, splitting a fragment in sub fragments, since some arches have PAGE_SIZE=65536 Successfully tested on various ethernet devices. (ixgbe, igb, bnx2x, tg3, mellanox mlx4) Signed-off-by: Eric Dumazet <edumazet@google.com> Cc: Ben Hutchings <bhutchings@solarflare.com> Cc: Vijay Subramanian <subramanian.vijay@gmail.com> Cc: Alexander Duyck <alexander.h.duyck@intel.com> Tested-by: Vijay Subramanian <subramanian.vijay@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2012-09-24 07:04:42 +08:00
page_address(pfrag->page) + pfrag->offset,
offset, copy, skb->len, skb) < 0)
goto error_efault;
pfrag->offset += copy;
skb_frag_size_add(&skb_shinfo(skb)->frags[i - 1], copy);
skb->len += copy;
skb->data_len += copy;
skb->truesize += copy;
atomic_add(copy, &sk->sk_wmem_alloc);
}
offset += copy;
length -= copy;
}
net: use a per task frag allocator We currently use a per socket order-0 page cache for tcp_sendmsg() operations. This page is used to build fragments for skbs. Its done to increase probability of coalescing small write() into single segments in skbs still in write queue (not yet sent) But it wastes a lot of memory for applications handling many mostly idle sockets, since each socket holds one page in sk->sk_sndmsg_page Its also quite inefficient to build TSO 64KB packets, because we need about 16 pages per skb on arches where PAGE_SIZE = 4096, so we hit page allocator more than wanted. This patch adds a per task frag allocator and uses bigger pages, if available. An automatic fallback is done in case of memory pressure. (up to 32768 bytes per frag, thats order-3 pages on x86) This increases TCP stream performance by 20% on loopback device, but also benefits on other network devices, since 8x less frags are mapped on transmit and unmapped on tx completion. Alexander Duyck mentioned a probable performance win on systems with IOMMU enabled. Its possible some SG enabled hardware cant cope with bigger fragments, but their ndo_start_xmit() should already handle this, splitting a fragment in sub fragments, since some arches have PAGE_SIZE=65536 Successfully tested on various ethernet devices. (ixgbe, igb, bnx2x, tg3, mellanox mlx4) Signed-off-by: Eric Dumazet <edumazet@google.com> Cc: Ben Hutchings <bhutchings@solarflare.com> Cc: Vijay Subramanian <subramanian.vijay@gmail.com> Cc: Alexander Duyck <alexander.h.duyck@intel.com> Tested-by: Vijay Subramanian <subramanian.vijay@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2012-09-24 07:04:42 +08:00
return 0;
net: use a per task frag allocator We currently use a per socket order-0 page cache for tcp_sendmsg() operations. This page is used to build fragments for skbs. Its done to increase probability of coalescing small write() into single segments in skbs still in write queue (not yet sent) But it wastes a lot of memory for applications handling many mostly idle sockets, since each socket holds one page in sk->sk_sndmsg_page Its also quite inefficient to build TSO 64KB packets, because we need about 16 pages per skb on arches where PAGE_SIZE = 4096, so we hit page allocator more than wanted. This patch adds a per task frag allocator and uses bigger pages, if available. An automatic fallback is done in case of memory pressure. (up to 32768 bytes per frag, thats order-3 pages on x86) This increases TCP stream performance by 20% on loopback device, but also benefits on other network devices, since 8x less frags are mapped on transmit and unmapped on tx completion. Alexander Duyck mentioned a probable performance win on systems with IOMMU enabled. Its possible some SG enabled hardware cant cope with bigger fragments, but their ndo_start_xmit() should already handle this, splitting a fragment in sub fragments, since some arches have PAGE_SIZE=65536 Successfully tested on various ethernet devices. (ixgbe, igb, bnx2x, tg3, mellanox mlx4) Signed-off-by: Eric Dumazet <edumazet@google.com> Cc: Ben Hutchings <bhutchings@solarflare.com> Cc: Vijay Subramanian <subramanian.vijay@gmail.com> Cc: Alexander Duyck <alexander.h.duyck@intel.com> Tested-by: Vijay Subramanian <subramanian.vijay@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2012-09-24 07:04:42 +08:00
error_efault:
err = -EFAULT;
error:
cork->length -= length;
IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
return err;
}
EXPORT_SYMBOL_GPL(ip6_append_data);
static void ip6_cork_release(struct inet_sock *inet, struct ipv6_pinfo *np)
{
if (np->cork.opt) {
kfree(np->cork.opt->dst0opt);
kfree(np->cork.opt->dst1opt);
kfree(np->cork.opt->hopopt);
kfree(np->cork.opt->srcrt);
kfree(np->cork.opt);
np->cork.opt = NULL;
}
if (inet->cork.base.dst) {
dst_release(inet->cork.base.dst);
inet->cork.base.dst = NULL;
inet->cork.base.flags &= ~IPCORK_ALLFRAG;
}
memset(&inet->cork.fl, 0, sizeof(inet->cork.fl));
}
int ip6_push_pending_frames(struct sock *sk)
{
struct sk_buff *skb, *tmp_skb;
struct sk_buff **tail_skb;
struct in6_addr final_dst_buf, *final_dst = &final_dst_buf;
struct inet_sock *inet = inet_sk(sk);
struct ipv6_pinfo *np = inet6_sk(sk);
struct net *net = sock_net(sk);
struct ipv6hdr *hdr;
struct ipv6_txoptions *opt = np->cork.opt;
struct rt6_info *rt = (struct rt6_info *)inet->cork.base.dst;
struct flowi6 *fl6 = &inet->cork.fl.u.ip6;
unsigned char proto = fl6->flowi6_proto;
int err = 0;
if ((skb = __skb_dequeue(&sk->sk_write_queue)) == NULL)
goto out;
tail_skb = &(skb_shinfo(skb)->frag_list);
/* move skb->data to ip header from ext header */
if (skb->data < skb_network_header(skb))
__skb_pull(skb, skb_network_offset(skb));
while ((tmp_skb = __skb_dequeue(&sk->sk_write_queue)) != NULL) {
__skb_pull(tmp_skb, skb_network_header_len(skb));
*tail_skb = tmp_skb;
tail_skb = &(tmp_skb->next);
skb->len += tmp_skb->len;
skb->data_len += tmp_skb->len;
skb->truesize += tmp_skb->truesize;
tmp_skb->destructor = NULL;
tmp_skb->sk = NULL;
}
/* Allow local fragmentation. */
if (np->pmtudisc < IPV6_PMTUDISC_DO)
skb->local_df = 1;
*final_dst = fl6->daddr;
__skb_pull(skb, skb_network_header_len(skb));
if (opt && opt->opt_flen)
ipv6_push_frag_opts(skb, opt, &proto);
if (opt && opt->opt_nflen)
ipv6_push_nfrag_opts(skb, opt, &proto, &final_dst);
skb_push(skb, sizeof(struct ipv6hdr));
skb_reset_network_header(skb);
hdr = ipv6_hdr(skb);
*(__be32*)hdr = fl6->flowlabel |
htonl(0x60000000 | ((int)np->cork.tclass << 20));
hdr->hop_limit = np->cork.hop_limit;
hdr->nexthdr = proto;
hdr->saddr = fl6->saddr;
hdr->daddr = *final_dst;
skb->priority = sk->sk_priority;
skb->mark = sk->sk_mark;
skb_dst_set(skb, dst_clone(&rt->dst));
IP6_UPD_PO_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
if (proto == IPPROTO_ICMPV6) {
struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb));
ICMP6MSGOUT_INC_STATS_BH(net, idev, icmp6_hdr(skb)->icmp6_type);
ICMP6_INC_STATS_BH(net, idev, ICMP6_MIB_OUTMSGS);
}
err = ip6_local_out(skb);
if (err) {
if (err > 0)
err = net_xmit_errno(err);
if (err)
goto error;
}
out:
ip6_cork_release(inet, np);
return err;
error:
IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
goto out;
}
EXPORT_SYMBOL_GPL(ip6_push_pending_frames);
void ip6_flush_pending_frames(struct sock *sk)
{
struct sk_buff *skb;
while ((skb = __skb_dequeue_tail(&sk->sk_write_queue)) != NULL) {
if (skb_dst(skb))
IP6_INC_STATS(sock_net(sk), ip6_dst_idev(skb_dst(skb)),
IPSTATS_MIB_OUTDISCARDS);
kfree_skb(skb);
}
ip6_cork_release(inet_sk(sk), inet6_sk(sk));
}
EXPORT_SYMBOL_GPL(ip6_flush_pending_frames);