int udp_chkaddr(struct sk_buff *skb)
{
struct iphdr *iph = skb->h.iph;
struct udphdr *uh = (struct udphdr *)(skb->h.raw + iph->ihl*4);
struct sock *sk;
sk = udp_v4_lookup(iph->saddr, uh->source, iph->daddr, uh->dest,
skb->dev);
if (!sk)
return 0;
/* 0 means accept all LOCAL addresses here, not all the world... */
if (sk->rcv_saddr == 0)
return 0;
return 1;
}
int icmp_chkaddr(struct sk_buff *skb)
{
struct icmphdr *icmph=(struct icmphdr *)(skb->h.raw + skb->h.iph->ihl*4);
struct iphdr *iph = (struct iphdr *) (icmph + 1);
void (*handler)(struct icmphdr *icmph, struct sk_buff *skb, struct device *dev, __u32 saddr, __u32 daddr, int len) = icmp_pointers[icmph->type].handler;
if (handler == icmp_unreach || handler == icmp_redirect) {
struct sock *sk;
switch (iph->protocol) {
case IPPROTO_TCP:
{
struct tcphdr *th = (struct tcphdr *)(((unsigned char *)iph)+(iph->ihl<<2));
sk = tcp_v4_lookup(iph->saddr, th->source, iph->daddr, th->dest, skb->dev);
if (!sk) return 0;
if (sk->saddr != iph->saddr) return 0;
if (sk->daddr != iph->daddr) return 0;
if (sk->dummy_th.dest != th->dest) return 0;
/*
* This packet came from us.
*/
return 1;
}
case IPPROTO_UDP:
{
struct udphdr *uh = (struct udphdr *)(((unsigned char *)iph)+(iph->ihl<<2));
sk = udp_v4_lookup(iph->saddr, uh->source, iph->daddr, uh->dest, skb->dev);
if (!sk) return 0;
if (sk->saddr != iph->saddr && ip_chk_addr(iph->saddr) != IS_MYADDR)
return 0;
/*
* This packet may have come from us.
* Assume it did.
*/
return 1;
}
}
}
return 0;
}
int icmp_chkaddr(struct sk_buff *skb)
{
struct icmphdr *icmph=(struct icmphdr *)(skb->nh.raw + skb->nh.iph->ihl*4);
struct iphdr *iph = (struct iphdr *) (icmph + 1);
void (*handler)(struct icmphdr *icmph, struct sk_buff *skb, int len) = icmp_pointers[icmph->type].handler;
if (handler == icmp_unreach || handler == icmp_redirect) {
struct sock *sk;
switch (iph->protocol) {
case IPPROTO_TCP:
{
struct tcphdr *th = (struct tcphdr *)(((unsigned char *)iph)+(iph->ihl<<2));
sk = tcp_v4_lookup(iph->daddr, th->dest, iph->saddr, th->source, skb->dev->ifindex);
if (!sk || (sk->state == TCP_LISTEN))
return 0;
/*
* This packet came from us.
*/
return 1;
}
case IPPROTO_UDP:
{
struct udphdr *uh = (struct udphdr *)(((unsigned char *)iph)+(iph->ihl<<2));
sk = udp_v4_lookup(iph->daddr, uh->dest, iph->saddr, uh->source, skb->dev->ifindex);
if (!sk) return 0;
if (sk->saddr != iph->saddr && inet_addr_type(iph->saddr) != RTN_LOCAL)
return 0;
/*
* This packet may have come from us.
* Assume it did.
*/
return 1;
}
}
}
return 0;
}
int udp_rcv(struct sk_buff *skb, struct device *dev, struct options *opt,
__u32 daddr, unsigned short len,
__u32 saddr, int redo, struct inet_protocol *protocol)
{
struct sock *sk;
struct udphdr *uh;
unsigned short ulen;
int addr_type;
/*
* If we're doing a "redo" (the socket was busy last time
* around), we can just queue the packet now..
*/
if (redo) {
udp_queue_rcv_skb(skb->sk, skb);
return 0;
}
/*
* First time through the loop.. Do all the setup stuff
* (including finding out the socket we go to etc)
*/
addr_type = IS_MYADDR;
if(!dev || dev->pa_addr!=daddr)
addr_type=ip_chk_addr(daddr);
/*
* Get the header.
*/
uh = (struct udphdr *) skb->h.uh;
ip_statistics.IpInDelivers++;
/*
* Validate the packet and the UDP length.
*/
ulen = ntohs(uh->len);
if (ulen > len || len < sizeof(*uh) || ulen < sizeof(*uh))
{
NETDEBUG(printk("UDP: short packet: %d/%d\n", ulen, len));
udp_statistics.UdpInErrors++;
kfree_skb(skb, FREE_WRITE);
return(0);
}
/* RFC1122 warning: According to 4.1.3.6, we MUST discard any */
/* datagram which has an invalid source address, either here or */
/* in IP. */
/* Right now, IP isn't doing it, and neither is UDP. It's on the */
/* FIXME list for IP, though, so I wouldn't worry about it. */
/* (That's the Right Place to do it, IMHO.) -- MS */
if (uh->check && (
( (skb->ip_summed == CHECKSUM_HW) && udp_check(uh, len, saddr, daddr, skb->csum ) ) ||
( (skb->ip_summed == CHECKSUM_NONE) && udp_check(uh, len, saddr, daddr,csum_partial((char*)uh, len, 0)))
/* skip if CHECKSUM_UNNECESSARY */
)
)
{
/* <*****@*****.**> wants to know, who sent it, to
go and stomp on the garbage sender... */
/* RFC1122: OK. Discards the bad packet silently (as far as */
/* the network is concerned, anyway) as per 4.1.3.4 (MUST). */
NETDEBUG(printk("UDP: bad checksum. From %08lX:%d to %08lX:%d ulen %d\n",
ntohl(saddr),ntohs(uh->source),
ntohl(daddr),ntohs(uh->dest),
ulen));
udp_statistics.UdpInErrors++;
kfree_skb(skb, FREE_WRITE);
return(0);
}
/*
* These are supposed to be switched.
*/
skb->daddr = saddr;
skb->saddr = daddr;
len=ulen;
skb->dev = dev;
skb_trim(skb,len);
#ifdef CONFIG_IP_MULTICAST
if (addr_type==IS_BROADCAST || addr_type==IS_MULTICAST)
return udp_v4_mcast_deliver(skb, uh, saddr, daddr);
#endif
#ifdef CONFIG_IP_TRANSPARENT_PROXY
if(skb->redirport)
sk = udp_v4_proxy_lookup(saddr, uh->source, daddr, uh->dest,
dev->pa_addr, skb->redirport, dev);
else
#endif
sk = udp_v4_lookup(saddr, uh->source, daddr, uh->dest, dev);
if (sk == NULL)
{
udp_statistics.UdpNoPorts++;
if (addr_type != IS_BROADCAST && addr_type != IS_MULTICAST)
{
icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0, dev);
}
/*
* Hmm. We got an UDP broadcast to a port to which we
* don't wanna listen. Ignore it.
*/
skb->sk = NULL;
kfree_skb(skb, FREE_WRITE);
return(0);
}
udp_deliver(sk, skb);
return 0;
}
/* Shared by v4/v6 tcp. */
unsigned short udp_good_socknum(void)
{
int result;
static int start = 0;
int i, best, best_size_so_far;
SOCKHASH_LOCK();
/* Select initial not-so-random "best" */
best = PROT_SOCK + 1 + (start & 1023);
best_size_so_far = 32767; /* "big" num */
result = best;
for (i = 0; i < UDP_HTABLE_SIZE; i++, result++) {
struct sock *sk;
int size;
sk = udp_hash[result & (UDP_HTABLE_SIZE - 1)];
/* No clashes - take it */
if (!sk)
goto out;
/* Is this one better than our best so far? */
size = 0;
do {
if(++size >= best_size_so_far)
goto next;
} while((sk = sk->next) != NULL);
best_size_so_far = size;
best = result;
next:
}
while (udp_lport_inuse(best))
best += UDP_HTABLE_SIZE;
result = best;
out:
start = result;
SOCKHASH_UNLOCK();
return result;
}
static void udp_v4_hash(struct sock *sk)
{
struct sock **skp;
int num = sk->num;
num &= (UDP_HTABLE_SIZE - 1);
skp = &udp_hash[num];
SOCKHASH_LOCK();
sk->next = *skp;
*skp = sk;
sk->hashent = num;
SOCKHASH_UNLOCK();
}
static void udp_v4_unhash(struct sock *sk)
{
struct sock **skp;
int num = sk->num;
num &= (UDP_HTABLE_SIZE - 1);
skp = &udp_hash[num];
SOCKHASH_LOCK();
while(*skp != NULL) {
if(*skp == sk) {
*skp = sk->next;
break;
}
skp = &((*skp)->next);
}
SOCKHASH_UNLOCK();
}
static void udp_v4_rehash(struct sock *sk)
{
struct sock **skp;
int num = sk->num;
int oldnum = sk->hashent;
num &= (UDP_HTABLE_SIZE - 1);
skp = &udp_hash[oldnum];
SOCKHASH_LOCK();
while(*skp != NULL) {
if(*skp == sk) {
*skp = sk->next;
break;
}
skp = &((*skp)->next);
}
sk->next = udp_hash[num];
udp_hash[num] = sk;
sk->hashent = num;
SOCKHASH_UNLOCK();
}
/* UDP is nearly always wildcards out the wazoo, it makes no sense to try
* harder than this. -DaveM
*/
__inline__ struct sock *udp_v4_lookup(u32 saddr, u16 sport, u32 daddr, u16 dport,
struct device *dev)
{
struct sock *sk, *result = NULL;
unsigned short hnum = ntohs(dport);
int badness = -1;
for(sk = udp_hash[hnum & (UDP_HTABLE_SIZE - 1)]; sk != NULL; sk = sk->next) {
if((sk->num == hnum) && !(sk->dead && (sk->state == TCP_CLOSE))) {
int score = 0;
if(sk->rcv_saddr) {
if(sk->rcv_saddr != daddr)
continue;
score++;
}
if(sk->daddr) {
if(sk->daddr != saddr)
continue;
score++;
}
if(sk->dummy_th.dest) {
if(sk->dummy_th.dest != sport)
continue;
score++;
}
/* If this socket is bound to a particular interface,
* did the packet come in on it? */
if (sk->bound_device) {
if (dev == sk->bound_device)
score++;
else
continue; /* mismatch--not this sock */
}
if(score == 4) {
result = sk;
break;
} else if(score > badness) {
result = sk;
badness = score;
}
}
}
return result;
}
#ifdef CONFIG_IP_TRANSPARENT_PROXY
struct sock *udp_v4_proxy_lookup(u32 saddr, u16 sport, u32 daddr, u16 dport, u32 paddr, u16 rport,
struct device *dev)
{
struct sock *hh[3], *sk, *result = NULL;
int i;
int badness = -1;
unsigned short hnum = ntohs(dport);
unsigned short hpnum = ntohs(rport);
SOCKHASH_LOCK();
hh[0] = udp_hash[hnum & (UDP_HTABLE_SIZE - 1)];
hh[1] = udp_hash[hpnum & (UDP_HTABLE_SIZE - 1)];
for (i = 0; i < 2; i++) {
for(sk = hh[i]; sk != NULL; sk = sk->next) {
if(sk->num == hnum || sk->num == hpnum) {
int score = 0;
if(sk->dead && (sk->state == TCP_CLOSE))
continue;
if(sk->rcv_saddr) {
if((sk->num != hpnum || sk->rcv_saddr != paddr) &&
(sk->num != hnum || sk->rcv_saddr != daddr))
continue;
score++;
}
if(sk->daddr) {
if(sk->daddr != saddr)
continue;
score++;
}
if(sk->dummy_th.dest) {
if(sk->dummy_th.dest != sport)
continue;
score++;
}
/* If this socket is bound to a particular interface,
* did the packet come in on it? */
if(sk->bound_device) {
if (sk->bound_device != dev)
continue;
score++;
}
if(score == 4 && sk->num == hnum) {
result = sk;
break;
} else if(score > badness && (sk->num == hpnum || sk->rcv_saddr)) {
result = sk;
badness = score;
}
}
}
}
SOCKHASH_UNLOCK();
return result;
}
#endif
static inline struct sock *udp_v4_mcast_next(struct sock *sk,
unsigned short num,
unsigned long raddr,
unsigned short rnum,
unsigned long laddr,
struct device *dev)
{
struct sock *s = sk;
unsigned short hnum = ntohs(num);
for(; s; s = s->next) {
if ((s->num != hnum) ||
(s->dead && (s->state == TCP_CLOSE)) ||
(s->daddr && s->daddr!=raddr) ||
(s->dummy_th.dest != rnum && s->dummy_th.dest != 0) ||
((s->bound_device) && (s->bound_device!=dev)) ||
(s->rcv_saddr && s->rcv_saddr != laddr))
continue;
break;
}
return s;
}
#define min(a,b) ((a)<(b)?(a):(b))
/*
* This routine is called by the ICMP module when it gets some
* sort of error condition. If err < 0 then the socket should
* be closed and the error returned to the user. If err > 0
* it's just the icmp type << 8 | icmp code.
* Header points to the ip header of the error packet. We move
* on past this. Then (as it used to claim before adjustment)
* header points to the first 8 bytes of the udp header. We need
* to find the appropriate port.
*/
void udp_err(int type, int code, unsigned char *header, __u32 daddr,
__u32 saddr, struct inet_protocol *protocol, int len)
{
struct udphdr *uh;
struct sock *sk;
/*
* Find the 8 bytes of post IP header ICMP included for us
*/
if(len<sizeof(struct udphdr))
return;
uh = (struct udphdr *)header;
sk = udp_v4_lookup(daddr, uh->dest, saddr, uh->source, NULL);
if (sk == NULL)
return; /* No socket for error */
if (type == ICMP_SOURCE_QUENCH)
{ /* Slow down! */
if (sk->cong_window > 1)
sk->cong_window = sk->cong_window/2;
return;
}
if (type == ICMP_PARAMETERPROB)
{
sk->err = EPROTO;
sk->error_report(sk);
return;
}
/*
* Various people wanted BSD UDP semantics. Well they've come
* back out because they slow down response to stuff like dead
* or unreachable name servers and they screw term users something
* chronic. Oh and it violates RFC1122. So basically fix your
* client code people.
*/
/* RFC1122: OK. Passes ICMP errors back to application, as per */
/* 4.1.3.3. */
/* After the comment above, that should be no surprise. */
if(code<=NR_ICMP_UNREACH && icmp_err_convert[code].fatal)
{
/*
* 4.x BSD compatibility item. Break RFC1122 to
* get BSD socket semantics.
*/
if(sk->bsdism && sk->state!=TCP_ESTABLISHED)
return;
sk->err = icmp_err_convert[code].errno;
sk->error_report(sk);
}
}
static unsigned short udp_check(struct udphdr *uh, int len, unsigned long saddr, unsigned long daddr, unsigned long base)
{
return(csum_tcpudp_magic(saddr, daddr, len, IPPROTO_UDP, base));
}
struct udpfakehdr
{
struct udphdr uh;
__u32 daddr;
__u32 other;
const char *from;
__u32 wcheck;
};
/*
* Copy and checksum a UDP packet from user space into a buffer. We still have to do the planning to
* get ip_build_xmit to spot direct transfer to network card and provide an additional callback mode
* for direct user->board I/O transfers. That one will be fun.
*/
static void udp_getfrag(const void *p, __u32 saddr, char * to, unsigned int offset, unsigned int fraglen)
{
struct udpfakehdr *ufh = (struct udpfakehdr *)p;
const char *src;
char *dst;
unsigned int len;
if (offset)
{
len = fraglen;
src = ufh->from+(offset-sizeof(struct udphdr));
dst = to;
}
else
{
len = fraglen-sizeof(struct udphdr);
src = ufh->from;
dst = to+sizeof(struct udphdr);
}
ufh->wcheck = csum_partial_copy_fromuser(src, dst, len, ufh->wcheck);
if (offset == 0)
{
ufh->wcheck = csum_partial((char *)ufh, sizeof(struct udphdr),
ufh->wcheck);
ufh->uh.check = csum_tcpudp_magic(saddr, ufh->daddr,
ntohs(ufh->uh.len),
IPPROTO_UDP, ufh->wcheck);
if (ufh->uh.check == 0)
ufh->uh.check = -1;
memcpy(to, ufh, sizeof(struct udphdr));
}
}
/*
* Unchecksummed UDP is sufficiently critical to stuff like ATM video conferencing
* that we use two routines for this for speed. Probably we ought to have a CONFIG_FAST_NET
* set for >10Mb/second boards to activate this sort of coding. Timing needed to verify if
* this is a valid decision.
*/
static void udp_getfrag_nosum(const void *p, __u32 saddr, char * to, unsigned int offset, unsigned int fraglen)
{
struct udpfakehdr *ufh = (struct udpfakehdr *)p;
const char *src;
char *dst;
unsigned int len;
if (offset)
{
len = fraglen;
src = ufh->from+(offset-sizeof(struct udphdr));
dst = to;
}
else
{
len = fraglen-sizeof(struct udphdr);
src = ufh->from;
dst = to+sizeof(struct udphdr);
}
memcpy_fromfs(dst,src,len);
if (offset == 0)
memcpy(to, ufh, sizeof(struct udphdr));
}
/*
* Send UDP frames.
*/
static int udp_send(struct sock *sk, struct sockaddr_in *sin,
const unsigned char *from, int len, int rt,
__u32 saddr, int noblock)
{
int ulen = len + sizeof(struct udphdr);
int a;
struct udpfakehdr ufh;
if(ulen>65535-sizeof(struct iphdr))
return -EMSGSIZE;
ufh.uh.source = sk->dummy_th.source;
ufh.uh.dest = sin->sin_port;
ufh.uh.len = htons(ulen);
ufh.uh.check = 0;
ufh.daddr = sin->sin_addr.s_addr;
ufh.other = (htons(ulen) << 16) + IPPROTO_UDP*256;
ufh.from = from;
ufh.wcheck = 0;
#ifdef CONFIG_IP_TRANSPARENT_PROXY
if (rt&MSG_PROXY)
{
/*
* We map the first 8 bytes of a second sockaddr_in
* into the last 8 (unused) bytes of a sockaddr_in.
* This _is_ ugly, but it's the only way to do it
* easily, without adding system calls.
*/
struct sockaddr_in *sinfrom =
(struct sockaddr_in *) sin->sin_zero;
if (!suser())
return(-EPERM);
if (sinfrom->sin_family && sinfrom->sin_family != AF_INET)
return(-EINVAL);
if (sinfrom->sin_port == 0)
return(-EINVAL);
saddr = sinfrom->sin_addr.s_addr;
ufh.uh.source = sinfrom->sin_port;
}
#endif
/* RFC1122: OK. Provides the checksumming facility (MUST) as per */
/* 4.1.3.4. It's configurable by the application via setsockopt() */
/* (MAY) and it defaults to on (MUST). Almost makes up for the */
/* violation above. -- MS */
if(sk->no_check)
a = ip_build_xmit(sk, udp_getfrag_nosum, &ufh, ulen,
sin->sin_addr.s_addr, saddr, sk->opt, rt, IPPROTO_UDP, noblock);
else
a = ip_build_xmit(sk, udp_getfrag, &ufh, ulen,
sin->sin_addr.s_addr, saddr, sk->opt, rt, IPPROTO_UDP, noblock);
if(a<0)
return a;
udp_statistics.UdpOutDatagrams++;
return len;
}
void udp_err(struct sk_buff *skb, u32 info)
{
struct iphdr *iph = (struct iphdr*)skb->data;
struct udphdr *uh = (struct udphdr*)(skb->data+(iph->ihl<<2));
int type = skb->h.icmph->type;
int code = skb->h.icmph->code;
struct sock *sk;
int harderr;
int err;
sk = udp_v4_lookup(iph->daddr, uh->dest, iph->saddr, uh->source, skb->dev->ifindex);
if (sk == NULL) {
ICMP_INC_STATS_BH(IcmpInErrors);
return; /* No socket for error */
}
err = 0;
harderr = 0;
switch (type) {
default:
case ICMP_TIME_EXCEEDED:
err = EHOSTUNREACH;
break;
case ICMP_SOURCE_QUENCH:
goto out;
case ICMP_PARAMETERPROB:
err = EPROTO;
harderr = 1;
break;
case ICMP_DEST_UNREACH:
if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
if (sk->protinfo.af_inet.pmtudisc != IP_PMTUDISC_DONT) {
err = EMSGSIZE;
harderr = 1;
break;
}
goto out;
}
err = EHOSTUNREACH;
if (code <= NR_ICMP_UNREACH) {
harderr = icmp_err_convert[code].fatal;
err = icmp_err_convert[code].errno;
}
break;
}
/*
* RFC1122: OK. Passes ICMP errors back to application, as per
* 4.1.3.3.
*/
if (!sk->protinfo.af_inet.recverr) {
if (!harderr || sk->state != TCP_ESTABLISHED)
goto out;
} else {
ip_icmp_error(sk, skb, err, uh->dest, info, (u8*)(uh+1));
}
sk->err = err;
sk->error_report(sk);
out:
sock_put(sk);
}
