Skip to content

Commit

Permalink
[UDP]: Clean up UDP-Lite receive checksum
Browse files Browse the repository at this point in the history 
This patch eliminates some duplicate code for the verification of
receive checksums between UDP-Lite and UDP.  It does this by
introducing __skb_checksum_complete_head which is identical to
__skb_checksum_complete_head apart from the fact that it takes
a length parameter rather than computing the first skb->len bytes.

As a result UDP-Lite will be able to use hardware checksum offload
for packets which do not use partial coverage checksums.  It also
means that UDP-Lite loopback no longer does unnecessary checksum
verification.

If any NICs start support UDP-Lite this would also start working
automatically.

This patch removes the assumption that msg_flags has MSG_TRUNC clear
upon entry in recvmsg.

Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
  • Loading branch information
@herbertx
herbertx authored and David S. Miller committed Apr 26, 2007
1 parent 1ab6eb6 commit 759e5d0
Show file tree
Hide file tree
Showing 8 changed files with 109 additions and 120 deletions.
1 change: 1 addition & 0 deletions include/linux/skbuff.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -1372,6 +1372,7 @@ static inline void __net_timestamp(struct sk_buff *skb)
}


extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
extern __sum16 __skb_checksum_complete(struct sk_buff *skb);

/**
Expand Down
5 changes: 1 addition & 4 deletions include/net/udp.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -72,10 +72,7 @@ struct sk_buff;
*/
static inline __sum16 __udp_lib_checksum_complete(struct sk_buff *skb)
{
if (! UDP_SKB_CB(skb)->partial_cov)
return __skb_checksum_complete(skb);
return csum_fold(skb_checksum(skb, 0, UDP_SKB_CB(skb)->cscov,
skb->csum));
return __skb_checksum_complete_head(skb, UDP_SKB_CB(skb)->cscov);
}

static inline int udp_lib_checksum_complete(struct sk_buff *skb)
Expand Down
39 changes: 6 additions & 33 deletions include/net/udplite.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -47,11 +47,10 @@ static inline int udplite_checksum_init(struct sk_buff *skb, struct udphdr *uh)
return 1;
}

UDP_SKB_CB(skb)->partial_cov = 0;
cscov = ntohs(uh->len);

if (cscov == 0) /* Indicates that full coverage is required. */
cscov = skb->len;
;
else if (cscov < 8 || cscov > skb->len) {
/*
* Coverage length violates RFC 3828: log and discard silently.
Expand All @@ -60,42 +59,16 @@ static inline int udplite_checksum_init(struct sk_buff *skb, struct udphdr *uh)
cscov, skb->len);
return 1;

} else if (cscov < skb->len)
} else if (cscov < skb->len) {
UDP_SKB_CB(skb)->partial_cov = 1;

UDP_SKB_CB(skb)->cscov = cscov;

/*
* There is no known NIC manufacturer supporting UDP-Lite yet,
* hence ip_summed is always (re-)set to CHECKSUM_NONE.
*/
skb->ip_summed = CHECKSUM_NONE;
UDP_SKB_CB(skb)->cscov = cscov;
if (skb->ip_summed == CHECKSUM_COMPLETE)
skb->ip_summed = CHECKSUM_NONE;
}

return 0;
}

static __inline__ int udplite4_csum_init(struct sk_buff *skb, struct udphdr *uh)
{
int rc = udplite_checksum_init(skb, uh);

if (!rc)
skb->csum = csum_tcpudp_nofold(skb->nh.iph->saddr,
skb->nh.iph->daddr,
skb->len, IPPROTO_UDPLITE, 0);
return rc;
}

static __inline__ int udplite6_csum_init(struct sk_buff *skb, struct udphdr *uh)
{
int rc = udplite_checksum_init(skb, uh);

if (!rc)
skb->csum = ~csum_unfold(csum_ipv6_magic(&skb->nh.ipv6h->saddr,
&skb->nh.ipv6h->daddr,
skb->len, IPPROTO_UDPLITE, 0));
return rc;
}

static inline int udplite_sender_cscov(struct udp_sock *up, struct udphdr *uh)
{
int cscov = up->len;
Expand Down
10 changes: 8 additions & 2 deletions net/core/datagram.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -411,18 +411,24 @@ static int skb_copy_and_csum_datagram(const struct sk_buff *skb, int offset,
return -EFAULT;
}

__sum16 __skb_checksum_complete(struct sk_buff *skb)
__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len)
{
__sum16 sum;

sum = csum_fold(skb_checksum(skb, 0, skb->len, skb->csum));
sum = csum_fold(skb_checksum(skb, 0, len, skb->csum));
if (likely(!sum)) {
if (unlikely(skb->ip_summed == CHECKSUM_COMPLETE))
netdev_rx_csum_fault(skb->dev);
skb->ip_summed = CHECKSUM_UNNECESSARY;
}
return sum;
}
EXPORT_SYMBOL(__skb_checksum_complete_head);

__sum16 __skb_checksum_complete(struct sk_buff *skb)
{
return __skb_checksum_complete_head(skb, skb->len);
}
EXPORT_SYMBOL(__skb_checksum_complete);

/**
Expand Down
96 changes: 49 additions & 47 deletions net/ipv4/udp.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -810,7 +810,9 @@ int udp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
struct inet_sock *inet = inet_sk(sk);
struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
struct sk_buff *skb;
int copied, err, copy_only, is_udplite = IS_UDPLITE(sk);
unsigned int ulen, copied;
int err;
int is_udplite = IS_UDPLITE(sk);

/*
* Check any passed addresses
Expand All @@ -826,28 +828,25 @@ int udp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
if (!skb)
goto out;

copied = skb->len - sizeof(struct udphdr);
if (copied > len) {
copied = len;
ulen = skb->len - sizeof(struct udphdr);
copied = len;
if (copied > ulen)
copied = ulen;
else if (copied < ulen)
msg->msg_flags |= MSG_TRUNC;
}

/*
* Decide whether to checksum and/or copy data.
*
* UDP: checksum may have been computed in HW,
* (re-)compute it if message is truncated.
* UDP-Lite: always needs to checksum, no HW support.
* If checksum is needed at all, try to do it while copying the
* data. If the data is truncated, or if we only want a partial
* coverage checksum (UDP-Lite), do it before the copy.
*/
copy_only = (skb->ip_summed==CHECKSUM_UNNECESSARY);

if (is_udplite || (!copy_only && msg->msg_flags&MSG_TRUNC)) {
if (__udp_lib_checksum_complete(skb))
if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) {
if (udp_lib_checksum_complete(skb))
goto csum_copy_err;
copy_only = 1;
}

if (copy_only)
if (skb->ip_summed == CHECKSUM_UNNECESSARY)
err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr),
msg->msg_iov, copied );
else {
Expand Down Expand Up @@ -875,7 +874,7 @@ int udp_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,

err = copied;
if (flags & MSG_TRUNC)
err = skb->len - sizeof(struct udphdr);
err = ulen;

out_free:
skb_free_datagram(sk, skb);
Expand Down Expand Up @@ -1095,10 +1094,9 @@ int udp_queue_rcv_skb(struct sock * sk, struct sk_buff *skb)
}
}

if (sk->sk_filter && skb->ip_summed != CHECKSUM_UNNECESSARY) {
if (__udp_lib_checksum_complete(skb))
if (sk->sk_filter) {
if (udp_lib_checksum_complete(skb))
goto drop;
skb->ip_summed = CHECKSUM_UNNECESSARY;
}

if ((rc = sock_queue_rcv_skb(sk,skb)) < 0) {
Expand Down Expand Up @@ -1166,33 +1164,44 @@ static int __udp4_lib_mcast_deliver(struct sk_buff *skb,
* Otherwise, csum completion requires chacksumming packet body,
* including udp header and folding it to skb->csum.
*/
static inline void udp4_csum_init(struct sk_buff *skb, struct udphdr *uh)
static inline int udp4_csum_init(struct sk_buff *skb, struct udphdr *uh,
int proto)
{
int err;

UDP_SKB_CB(skb)->partial_cov = 0;
UDP_SKB_CB(skb)->cscov = skb->len;

if (proto == IPPROTO_UDPLITE) {
err = udplite_checksum_init(skb, uh);
if (err)
return err;
}

if (uh->check == 0) {
skb->ip_summed = CHECKSUM_UNNECESSARY;
} else if (skb->ip_summed == CHECKSUM_COMPLETE) {
if (!csum_tcpudp_magic(skb->nh.iph->saddr, skb->nh.iph->daddr,
skb->len, IPPROTO_UDP, skb->csum ))
skb->len, proto, skb->csum))
skb->ip_summed = CHECKSUM_UNNECESSARY;
}
if (skb->ip_summed != CHECKSUM_UNNECESSARY)
skb->csum = csum_tcpudp_nofold(skb->nh.iph->saddr,
skb->nh.iph->daddr,
skb->len, IPPROTO_UDP, 0);
skb->len, proto, 0);
/* Probably, we should checksum udp header (it should be in cache
* in any case) and data in tiny packets (< rx copybreak).
*/

/* UDP = UDP-Lite with a non-partial checksum coverage */
UDP_SKB_CB(skb)->partial_cov = 0;
return 0;
}

/*
* All we need to do is get the socket, and then do a checksum.
*/

int __udp4_lib_rcv(struct sk_buff *skb, struct hlist_head udptable[],
int is_udplite)
int proto)
{
struct sock *sk;
struct udphdr *uh = skb->h.uh;
Expand All @@ -1211,19 +1220,16 @@ int __udp4_lib_rcv(struct sk_buff *skb, struct hlist_head udptable[],
if (ulen > skb->len)
goto short_packet;

if(! is_udplite ) { /* UDP validates ulen. */

if (proto == IPPROTO_UDP) {
/* UDP validates ulen. */
if (ulen < sizeof(*uh) || pskb_trim_rcsum(skb, ulen))
goto short_packet;
uh = skb->h.uh;

udp4_csum_init(skb, uh);

} else { /* UDP-Lite validates cscov. */
if (udplite4_csum_init(skb, uh))
goto csum_error;
}

if (udp4_csum_init(skb, uh, proto))
goto csum_error;

if(rt->rt_flags & (RTCF_BROADCAST|RTCF_MULTICAST))
return __udp4_lib_mcast_deliver(skb, uh, saddr, daddr, udptable);

Expand All @@ -1250,7 +1256,7 @@ int __udp4_lib_rcv(struct sk_buff *skb, struct hlist_head udptable[],
if (udp_lib_checksum_complete(skb))
goto csum_error;

UDP_INC_STATS_BH(UDP_MIB_NOPORTS, is_udplite);
UDP_INC_STATS_BH(UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);

/*
Expand All @@ -1262,7 +1268,7 @@ int __udp4_lib_rcv(struct sk_buff *skb, struct hlist_head udptable[],

short_packet:
LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: short packet: From %u.%u.%u.%u:%u %d/%d to %u.%u.%u.%u:%u\n",
is_udplite? "-Lite" : "",
proto == IPPROTO_UDPLITE ? "-Lite" : "",
NIPQUAD(saddr),
ntohs(uh->source),
ulen,
Expand All @@ -1277,21 +1283,21 @@ int __udp4_lib_rcv(struct sk_buff *skb, struct hlist_head udptable[],
* the network is concerned, anyway) as per 4.1.3.4 (MUST).
*/
LIMIT_NETDEBUG(KERN_DEBUG "UDP%s: bad checksum. From %d.%d.%d.%d:%d to %d.%d.%d.%d:%d ulen %d\n",
is_udplite? "-Lite" : "",
proto == IPPROTO_UDPLITE ? "-Lite" : "",
NIPQUAD(saddr),
ntohs(uh->source),
NIPQUAD(daddr),
ntohs(uh->dest),
ulen);
drop:
UDP_INC_STATS_BH(UDP_MIB_INERRORS, is_udplite);
UDP_INC_STATS_BH(UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
kfree_skb(skb);
return(0);
}

__inline__ int udp_rcv(struct sk_buff *skb)
{
return __udp4_lib_rcv(skb, udp_hash, 0);
return __udp4_lib_rcv(skb, udp_hash, IPPROTO_UDP);
}

int udp_destroy_sock(struct sock *sk)
Expand Down Expand Up @@ -1486,15 +1492,11 @@ unsigned int udp_poll(struct file *file, struct socket *sock, poll_table *wait)
struct sk_buff *skb;

spin_lock_bh(&rcvq->lock);
while ((skb = skb_peek(rcvq)) != NULL) {
if (udp_lib_checksum_complete(skb)) {
UDP_INC_STATS_BH(UDP_MIB_INERRORS, is_lite);
__skb_unlink(skb, rcvq);
kfree_skb(skb);
} else {
skb->ip_summed = CHECKSUM_UNNECESSARY;
break;
}
while ((skb = skb_peek(rcvq)) != NULL &&
udp_lib_checksum_complete(skb)) {
UDP_INC_STATS_BH(UDP_MIB_INERRORS, is_lite);
__skb_unlink(skb, rcvq);
kfree_skb(skb);
}
spin_unlock_bh(&rcvq->lock);

Expand Down
2 changes: 1 addition & 1 deletion net/ipv4/udplite.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -31,7 +31,7 @@ static int udplite_v4_get_port(struct sock *sk, unsigned short snum)

static int udplite_rcv(struct sk_buff *skb)
{
return __udp4_lib_rcv(skb, udplite_hash, 1);
return __udp4_lib_rcv(skb, udplite_hash, IPPROTO_UDPLITE);
}

static void udplite_err(struct sk_buff *skb, u32 info)
Expand Down
Loading

0 comments on commit 759e5d0

Please sign in to comment.