/*
* Check and stop ack storm.
* Return 0 if ack storm is found.
*/
static int syn_proxy_is_ack_storm(struct tcphdr *tcph, struct ip_vs_conn *cp)
{
/* only for syn-proxy sessions */
if (!(cp->flags & IP_VS_CONN_F_SYNPROXY) || !tcph->ack)
return 1;
if (unlikely(sysctl_ip_vs_synproxy_dup_ack_thresh == 0))
return 1;
if (unlikely(tcph->seq == cp->last_seq &&
tcph->ack_seq == cp->last_ack_seq)) {
atomic_inc(&cp->dup_ack_cnt);
if (atomic_read(&cp->dup_ack_cnt) >=
sysctl_ip_vs_synproxy_dup_ack_thresh) {
atomic_set(&cp->dup_ack_cnt,
sysctl_ip_vs_synproxy_dup_ack_thresh);
/* update statistics */
IP_VS_INC_ESTATS(ip_vs_esmib, SYNPROXY_ACK_STORM);
return 0;
}
return 1;
}
cp->last_seq = tcph->seq;
cp->last_ack_seq = tcph->ack_seq;
atomic_set(&cp->dup_ack_cnt, 0);
return 1;
}
int
ip_vs_synproxy_filter_ack(struct sk_buff *skb, struct ip_vs_conn *cp,
struct ip_vs_protocol *pp,
struct ip_vs_iphdr *iph, int *verdict)
{
struct tcphdr _tcph, *th;
th = skb_header_pointer(skb, iph->len, sizeof(_tcph), &_tcph);
if (unlikely(NULL == th)) {
IP_VS_ERR_RL("skb has a invalid tcp header\n");
*verdict = NF_DROP;
return 0;
}
spin_lock(&cp->lock);
if ((cp->flags & IP_VS_CONN_F_SYNPROXY) &&
cp->state == IP_VS_TCP_S_SYN_SENT) {
/*
* Not a ack packet, drop it.
*/
if (!th->ack) {
spin_unlock(&cp->lock);
*verdict = NF_DROP;
return 0;
}
if (sysctl_ip_vs_synproxy_skb_store_thresh <
skb_queue_len(&cp->ack_skb)) {
spin_unlock(&cp->lock);
/* update statistics */
IP_VS_INC_ESTATS(ip_vs_esmib, SYNPROXY_SYNSEND_QLEN);
*verdict = NF_DROP;
return 0;
}
/*
* Still some space left, store it.
*/
skb_queue_tail(&cp->ack_skb, skb);
spin_unlock(&cp->lock);
*verdict = NF_STOLEN;
return 0;
}
spin_unlock(&cp->lock);
return 1;
}
/*
* Syn-proxy session reuse function.
* Update syn_proxy_seq struct and clean syn-proxy related
* members.
*/
int
ip_vs_synproxy_reuse_conn(int af, struct sk_buff *skb,
struct ip_vs_conn *cp,
struct ip_vs_protocol *pp,
struct ip_vs_iphdr *iph, int *verdict)
{
struct tcphdr _tcph, *th = NULL;
struct ip_vs_synproxy_opt opt;
int res_cookie_check;
u32 tcp_conn_reuse_states = 0;
th = skb_header_pointer(skb, iph->len, sizeof(_tcph), &_tcph);
if (unlikely(NULL == th)) {
IP_VS_ERR_RL("skb has a invalid tcp header\n");
*verdict = NF_DROP;
return 0;
}
tcp_conn_reuse_states =
((sysctl_ip_vs_synproxy_conn_reuse_cl << IP_VS_TCP_S_CLOSE) |
(sysctl_ip_vs_synproxy_conn_reuse_tw << IP_VS_TCP_S_TIME_WAIT) |
(sysctl_ip_vs_synproxy_conn_reuse_fw << IP_VS_TCP_S_FIN_WAIT) |
(sysctl_ip_vs_synproxy_conn_reuse_cw << IP_VS_TCP_S_CLOSE_WAIT) |
(sysctl_ip_vs_synproxy_conn_reuse_la << IP_VS_TCP_S_LAST_ACK));
if (((1 << (cp->state)) & tcp_conn_reuse_states) &&
(cp->flags & IP_VS_CONN_F_SYNPROXY) &&
(!th->syn && th->ack && !th->rst && !th->fin) &&
(cp->syn_proxy_seq.init_seq !=
htonl((__u32) ((ntohl(th->ack_seq) - 1))))) {
/*
* Import: set tcp hdr before cookie check, as it
* will be used in cookie_check funcs.
*/
skb_set_transport_header(skb, iph->len);
#ifdef CONFIG_IP_VS_IPV6
if (af == AF_INET6) {
res_cookie_check = ip_vs_synproxy_v6_cookie_check(skb,
ntohl
(th->
ack_seq)
- 1,
&opt);
} else
#endif
{
res_cookie_check = ip_vs_synproxy_v4_cookie_check(skb,
ntohl
(th->
ack_seq)
- 1,
&opt);
}
if (!res_cookie_check) {
/* update statistics */
IP_VS_INC_ESTATS(ip_vs_esmib, SYNPROXY_BAD_ACK);
/*
* Cookie check fail, let it go.
*/
return 1;
}
/* update statistics */
IP_VS_INC_ESTATS(ip_vs_esmib, SYNPROXY_OK_ACK);
IP_VS_INC_ESTATS(ip_vs_esmib, SYNPROXY_CONN_REUSED);
switch (cp->old_state) {
case IP_VS_TCP_S_CLOSE:
IP_VS_INC_ESTATS(ip_vs_esmib,
SYNPROXY_CONN_REUSED_CLOSE);
break;
case IP_VS_TCP_S_TIME_WAIT:
IP_VS_INC_ESTATS(ip_vs_esmib,
SYNPROXY_CONN_REUSED_TIMEWAIT);
break;
case IP_VS_TCP_S_FIN_WAIT:
IP_VS_INC_ESTATS(ip_vs_esmib,
SYNPROXY_CONN_REUSED_FINWAIT);
break;
case IP_VS_TCP_S_CLOSE_WAIT:
IP_VS_INC_ESTATS(ip_vs_esmib,
SYNPROXY_CONN_REUSED_CLOSEWAIT);
break;
case IP_VS_TCP_S_LAST_ACK:
IP_VS_INC_ESTATS(ip_vs_esmib,
SYNPROXY_CONN_REUSED_LASTACK);
break;
}
spin_lock(&cp->lock);
__syn_proxy_reuse_conn(cp, skb, th, pp);
spin_unlock(&cp->lock);
if (unlikely(!syn_proxy_send_rs_syn(af, th, cp, skb, pp, &opt))) {
IP_VS_ERR_RL
("syn_proxy_send_rs_syn failed when reuse conn!\n");
/* release conn immediately */
spin_lock(&cp->lock);
cp->timeout = 0;
spin_unlock(&cp->lock);
}
*verdict = NF_STOLEN;
return 0;
}
return 1;
}
/*
* Syn-proxy step 2 logic
* Receive client's 3-handshakes Ack packet, do cookie check
* and then send syn to rs after creating a session.
*
*/
int
ip_vs_synproxy_ack_rcv(int af, struct sk_buff *skb, struct tcphdr *th,
struct ip_vs_protocol *pp, struct ip_vs_conn **cpp,
struct ip_vs_iphdr *iph, int *verdict)
{
struct ip_vs_synproxy_opt opt;
struct ip_vs_service *svc;
int res_cookie_check;
/*
* Don't check svc syn-proxy flag, as it may
* be changed after syn-proxy step 1.
*/
if (!th->syn && th->ack && !th->rst && !th->fin &&
(svc =
ip_vs_service_get(af, skb->mark, iph->protocol, &iph->daddr,
th->dest))) {
if (ip_vs_todrop()) {
/*
* It seems that we are very loaded.
* We have to drop this packet :(
*/
ip_vs_service_put(svc);
*verdict = NF_DROP;
return 0;
}
if (sysctl_ip_vs_synproxy_defer &&
!syn_proxy_ack_has_data(skb, iph, th)) {
/* update statistics */
IP_VS_INC_ESTATS(ip_vs_esmib, SYNPROXY_NULL_ACK);
/*
* When expecting ack packet with payload,
* we get a pure ack, so have to drop it.
*/
ip_vs_service_put(svc);
*verdict = NF_DROP;
return 0;
}
/*
* Import: set tcp hdr before cookie check, as it
* will be used in cookie_check funcs.
*/
skb_set_transport_header(skb, iph->len);
#ifdef CONFIG_IP_VS_IPV6
if (af == AF_INET6) {
res_cookie_check = ip_vs_synproxy_v6_cookie_check(skb,
ntohl
(th->
ack_seq)
- 1,
&opt);
} else
#endif
{
res_cookie_check = ip_vs_synproxy_v4_cookie_check(skb,
ntohl
(th->
ack_seq)
- 1,
&opt);
}
if (!res_cookie_check) {
/* update statistics */
IP_VS_INC_ESTATS(ip_vs_esmib, SYNPROXY_BAD_ACK);
/*
* Cookie check fail, drop it.
*/
IP_VS_DBG(6, "syn_cookie check failed seq=%u\n",
ntohl(th->ack_seq) - 1);
ip_vs_service_put(svc);
*verdict = NF_DROP;
return 0;
}
/* update statistics */
IP_VS_INC_ESTATS(ip_vs_esmib, SYNPROXY_OK_ACK);
/*
* Let the virtual server select a real server for the
* incoming connection, and create a connection entry.
*/
*cpp = ip_vs_schedule(svc, skb, 1);
if (!*cpp) {
IP_VS_DBG(6, "ip_vs_schedule failed\n");
*verdict = ip_vs_leave(svc, skb, pp);
return 0;
}
/*
* Release service, we don't need it any more.
*/
ip_vs_service_put(svc);
/*
* Do anything but print a error msg when fail.
* Because session will be correctly freed in ip_vs_conn_expire.
*/
if (!syn_proxy_send_rs_syn(af, th, *cpp, skb, pp, &opt)) {
IP_VS_ERR_RL("syn_proxy_send_rs_syn failed!\n");
}
/* count in the ack packet (STOLEN by synproxy) */
ip_vs_in_stats(*cpp, skb);
/*
* Active sesion timer, and dec refcnt.
* Also stole the skb, and let caller return immediately.
*/
ip_vs_conn_put(*cpp);
*verdict = NF_STOLEN;
return 0;
}
return 1;
}
/*
* Create syn packet and send it to rs.
* ATTENTION: we also store syn skb in cp if syn retransimition
* is tured on.
*/
static int
syn_proxy_send_rs_syn(int af, const struct tcphdr *th,
struct ip_vs_conn *cp, struct sk_buff *skb,
struct ip_vs_protocol *pp, struct ip_vs_synproxy_opt *opt)
{
struct sk_buff *syn_skb;
int tcp_hdr_size;
__u8 tcp_flags = TCPCB_FLAG_SYN;
unsigned int tcphoff;
struct tcphdr *new_th;
if (!cp->packet_xmit) {
IP_VS_ERR_RL("warning: packet_xmit is null");
return 0;
}
syn_skb = alloc_skb(MAX_TCP_HEADER + 15, GFP_ATOMIC);
if (unlikely(syn_skb == NULL)) {
IP_VS_ERR_RL("alloc skb failed when send rs syn packet\n");
return 0;
}
/* Reserve space for headers */
skb_reserve(syn_skb, MAX_TCP_HEADER);
tcp_hdr_size = (sizeof(struct tcphdr) + TCPOLEN_MSS +
(opt->tstamp_ok ? TCPOLEN_TSTAMP_ALIGNED : 0) +
(opt->wscale_ok ? TCPOLEN_WSCALE_ALIGNED : 0) +
/* SACK_PERM is in the place of NOP NOP of TS */
((opt->sack_ok
&& !opt->tstamp_ok) ? TCPOLEN_SACKPERM_ALIGNED : 0));
new_th = (struct tcphdr *)skb_push(syn_skb, tcp_hdr_size);
/* Compose tcp header */
skb_reset_transport_header(syn_skb);
syn_skb->csum = 0;
/* Set tcp hdr */
new_th->source = th->source;
new_th->dest = th->dest;
new_th->seq = htonl(ntohl(th->seq) - 1);
new_th->ack_seq = 0;
*(((__u16 *) new_th) + 6) =
htons(((tcp_hdr_size >> 2) << 12) | tcp_flags);
/* FIX_ME: what window should we use */
new_th->window = htons(5000);
new_th->check = 0;
new_th->urg_ptr = 0;
new_th->urg = 0;
new_th->ece = 0;
new_th->cwr = 0;
syn_proxy_syn_build_options((__be32 *) (new_th + 1), opt);
/*
* Set ip hdr
* Attention: set source and dest addr to ack skb's.
* we rely on packet_xmit func to do NATs thing.
*/
#ifdef CONFIG_IP_VS_IPV6
if (af == AF_INET6) {
struct ipv6hdr *ack_iph = ipv6_hdr(skb);
struct ipv6hdr *iph =
(struct ipv6hdr *)skb_push(syn_skb, sizeof(struct ipv6hdr));
tcphoff = sizeof(struct ipv6hdr);
skb_reset_network_header(syn_skb);
memcpy(&iph->saddr, &ack_iph->saddr, sizeof(struct in6_addr));
memcpy(&iph->daddr, &ack_iph->daddr, sizeof(struct in6_addr));
iph->version = 6;
iph->nexthdr = NEXTHDR_TCP;
iph->payload_len = htons(tcp_hdr_size);
iph->hop_limit = IPV6_DEFAULT_HOPLIMIT;
new_th->check = 0;
syn_skb->csum =
skb_checksum(syn_skb, tcphoff, syn_skb->len - tcphoff, 0);
new_th->check =
csum_ipv6_magic(&iph->saddr, &iph->daddr,
syn_skb->len - tcphoff, IPPROTO_TCP,
syn_skb->csum);
} else
#endif
{
struct iphdr *ack_iph = ip_hdr(skb);
u32 rtos = RT_TOS(ack_iph->tos);
struct iphdr *iph =
(struct iphdr *)skb_push(syn_skb, sizeof(struct iphdr));
tcphoff = sizeof(struct iphdr);
skb_reset_network_header(syn_skb);
*((__u16 *) iph) = htons((4 << 12) | (5 << 8) | (rtos & 0xff));
iph->tot_len = htons(syn_skb->len);
iph->frag_off = htons(IP_DF);
/* FIX_ME: what ttl shoule we use */
iph->ttl = IPDEFTTL;
iph->protocol = IPPROTO_TCP;
iph->saddr = ack_iph->saddr;
iph->daddr = ack_iph->daddr;
ip_send_check(iph);
new_th->check = 0;
syn_skb->csum =
skb_checksum(syn_skb, tcphoff, syn_skb->len - tcphoff, 0);
new_th->check =
csum_tcpudp_magic(iph->saddr, iph->daddr,
syn_skb->len - tcphoff, IPPROTO_TCP,
syn_skb->csum);
}
/* Save syn_skb if syn retransmission is on */
if (sysctl_ip_vs_synproxy_syn_retry > 0) {
cp->syn_skb = skb_copy(syn_skb, GFP_ATOMIC);
atomic_set(&cp->syn_retry_max, sysctl_ip_vs_synproxy_syn_retry);
}
/* Save info for fast_response_xmit */
if(sysctl_ip_vs_fast_xmit && skb->dev &&
likely(skb->dev->type == ARPHRD_ETHER) &&
skb_mac_header_was_set(skb)) {
struct ethhdr *eth = (struct ethhdr *)skb_mac_header(skb);
if(likely(cp->indev == NULL)) {
cp->indev = skb->dev;
dev_hold(cp->indev);
}
if (unlikely(cp->indev != skb->dev)) {
dev_put(cp->indev);
cp->indev = skb->dev;
dev_hold(cp->indev);
}
memcpy(cp->src_hwaddr, eth->h_source, ETH_ALEN);
memcpy(cp->dst_hwaddr, eth->h_dest, ETH_ALEN);
IP_VS_INC_ESTATS(ip_vs_esmib, FAST_XMIT_SYNPROXY_SAVE);
IP_VS_DBG_RL("syn_proxy_send_rs_syn netdevice:%s\n",
netdev_name(skb->dev));
}
/* count in the syn packet */
ip_vs_in_stats(cp, skb);
/* If xmit failed, syn_skb will be freed correctly. */
cp->packet_xmit(syn_skb, cp, pp);
return 1;
}
/*
* syn-proxy step 1 logic:
* Check if synproxy is enabled for this skb, and
* send Syn/Ack back.
*
* Synproxy is enabled when:
* 1) skb is a Syn packet.
* 2) And the service is synproxy-enable.
* 3) And ip_vs_todrop return false.
*
* @return 0 means the caller should return at once and use
* verdict as return value, return 1 for nothing.
*/
int
ip_vs_synproxy_syn_rcv(int af, struct sk_buff *skb,
struct ip_vs_iphdr *iph, int *verdict)
{
struct ip_vs_service *svc = NULL;
struct tcphdr _tcph, *th;
struct ip_vs_synproxy_opt tcp_opt;
th = skb_header_pointer(skb, iph->len, sizeof(_tcph), &_tcph);
if (unlikely(th == NULL)) {
goto syn_rcv_out;
}
if (th->syn && !th->ack && !th->rst && !th->fin &&
(svc =
ip_vs_service_get(af, skb->mark, iph->protocol, &iph->daddr,
th->dest))
&& (svc->flags & IP_VS_CONN_F_SYNPROXY)) {
// release service here, because don't use it any all.
ip_vs_service_put(svc);
if (ip_vs_todrop()) {
/*
* It seems that we are very loaded.
* We have to drop this packet :(
*/
goto syn_rcv_out;
}
} else {
/*
* release service.
*/
if (svc != NULL) {
ip_vs_service_put(svc);
}
return 1;
}
/* update statistics */
IP_VS_INC_ESTATS(ip_vs_esmib, SYNPROXY_SYN_CNT);
/* Try to reuse skb if possible */
if (unlikely(skb_shared(skb) || skb_cloned(skb))) {
struct sk_buff *new_skb = skb_copy(skb, GFP_ATOMIC);
if (unlikely(new_skb == NULL)) {
goto syn_rcv_out;
}
/* Drop old skb */
kfree_skb(skb);
skb = new_skb;
}
/* reuse skb here: deal with tcp options, exchage ip, port. */
syn_proxy_reuse_skb(af, skb, &tcp_opt);
if (unlikely(skb->dev == NULL)) {
IP_VS_ERR_RL("%s: skb->dev is null !!!\n", __func__);
goto syn_rcv_out;
}
/* Send the packet out */
if (likely(skb->dev->type == ARPHRD_ETHER)) {
unsigned char t_hwaddr[ETH_ALEN];
/* Move the data pointer to point to the link layer header */
struct ethhdr *eth = (struct ethhdr *)skb_mac_header(skb);
skb->data = (unsigned char *)skb_mac_header(skb);
skb->len += ETH_HLEN; //sizeof(skb->mac.ethernet);
memcpy(t_hwaddr, (eth->h_dest), ETH_ALEN);
memcpy((eth->h_dest), (eth->h_source), ETH_ALEN);
memcpy((eth->h_source), t_hwaddr, ETH_ALEN);
skb->pkt_type = PACKET_OUTGOING;
} else if (skb->dev->type == ARPHRD_LOOPBACK) {
/* set link layer */
if (likely(skb_mac_header_was_set(skb))) {
skb->data = skb_mac_header(skb);
skb->len += sizeof(struct ethhdr);
} else {
skb_push(skb, sizeof(struct ethhdr));
skb_reset_mac_header(skb);
}
}
dev_queue_xmit(skb);
*verdict = NF_STOLEN;
return 0;
syn_rcv_out:
/* Drop the packet when all things are right also,
* then we needn't to kfree_skb() */
*verdict = NF_DROP;
return 0;
}
static int
tcp_conn_schedule(int af, struct sk_buff *skb, struct ip_vs_protocol *pp,
int *verdict, struct ip_vs_conn **cpp)
{
struct ip_vs_service *svc;
struct tcphdr _tcph, *th;
struct ip_vs_iphdr iph;
ip_vs_fill_iphdr(af, skb_network_header(skb), &iph);
th = skb_header_pointer(skb, iph.len, sizeof(_tcph), &_tcph);
if (th == NULL) {
*verdict = NF_DROP;
return 0;
}
/*
* Syn-proxy step 2 logic: receive client's
* 3-handshake Ack packet
*/
if (ip_vs_synproxy_ack_rcv(af, skb, th, pp, cpp, &iph, verdict) == 0) {
return 0;
}
if (th->syn && !th->ack && !th->fin && !th->rst &&
(svc = ip_vs_service_get(af, skb->mark, iph.protocol, &iph.daddr,
th->dest))) {
if (ip_vs_todrop()) {
/*
* It seems that we are very loaded.
* We have to drop this packet :(
*/
ip_vs_service_put(svc);
*verdict = NF_DROP;
return 0;
}
/*
* Let the virtual server select a real server for the
* incoming connection, and create a connection entry.
*/
*cpp = ip_vs_schedule(svc, skb, 0);
if (!*cpp) {
*verdict = ip_vs_leave(svc, skb, pp);
return 0;
}
/*
* Set private establish state timeout into cp from svc,
* due cp may use its user establish state timeout
* different from sysctl_ip_vs_tcp_timeouts
*/
(*cpp)->est_timeout = svc->est_timeout;
ip_vs_service_put(svc);
return 1;
}
/* drop tcp packet which send to vip and !vport */
if (sysctl_ip_vs_tcp_drop_entry &&
(svc = ip_vs_lookup_vip(af, iph.protocol, &iph.daddr))) {
IP_VS_INC_ESTATS(ip_vs_esmib, DEFENCE_TCP_DROP);
*verdict = NF_DROP;
return 0;
}
return 1;
}