2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * The User Datagram Protocol (UDP).
9 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
10 * Arnt Gulbrandsen, <agulbra@nvg.unit.no>
11 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
12 * Hirokazu Takahashi, <taka@valinux.co.jp>
15 * Alan Cox : verify_area() calls
16 * Alan Cox : stopped close while in use off icmp
17 * messages. Not a fix but a botch that
18 * for udp at least is 'valid'.
19 * Alan Cox : Fixed icmp handling properly
20 * Alan Cox : Correct error for oversized datagrams
21 * Alan Cox : Tidied select() semantics.
22 * Alan Cox : udp_err() fixed properly, also now
23 * select and read wake correctly on errors
24 * Alan Cox : udp_send verify_area moved to avoid mem leak
25 * Alan Cox : UDP can count its memory
26 * Alan Cox : send to an unknown connection causes
27 * an ECONNREFUSED off the icmp, but
29 * Alan Cox : Switched to new sk_buff handlers. No more backlog!
30 * Alan Cox : Using generic datagram code. Even smaller and the PEEK
31 * bug no longer crashes it.
32 * Fred Van Kempen : Net2e support for sk->broadcast.
33 * Alan Cox : Uses skb_free_datagram
34 * Alan Cox : Added get/set sockopt support.
35 * Alan Cox : Broadcasting without option set returns EACCES.
36 * Alan Cox : No wakeup calls. Instead we now use the callbacks.
37 * Alan Cox : Use ip_tos and ip_ttl
38 * Alan Cox : SNMP Mibs
39 * Alan Cox : MSG_DONTROUTE, and 0.0.0.0 support.
40 * Matt Dillon : UDP length checks.
41 * Alan Cox : Smarter af_inet used properly.
42 * Alan Cox : Use new kernel side addressing.
43 * Alan Cox : Incorrect return on truncated datagram receive.
44 * Arnt Gulbrandsen : New udp_send and stuff
45 * Alan Cox : Cache last socket
46 * Alan Cox : Route cache
47 * Jon Peatfield : Minor efficiency fix to sendto().
48 * Mike Shaver : RFC1122 checks.
49 * Alan Cox : Nonblocking error fix.
50 * Willy Konynenberg : Transparent proxying support.
51 * Mike McLagan : Routing by source
52 * David S. Miller : New socket lookup architecture.
53 * Last socket cache retained as it
54 * does have a high hit rate.
55 * Olaf Kirch : Don't linearise iovec on sendmsg.
56 * Andi Kleen : Some cleanups, cache destination entry
58 * Vitaly E. Lavrov : Transparent proxy revived after year coma.
59 * Melvin Smith : Check msg_name not msg_namelen in sendto(),
60 * return ENOTCONN for unconnected sockets (POSIX)
61 * Janos Farkas : don't deliver multi/broadcasts to a different
62 * bound-to-device socket
63 * Hirokazu Takahashi : HW checksumming for outgoing UDP
65 * Hirokazu Takahashi : sendfile() on UDP works now.
66 * Arnaldo C. Melo : convert /proc/net/udp to seq_file
67 * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which
68 * Alexey Kuznetsov: allow both IPv4 and IPv6 sockets to bind
69 * a single port at the same time.
70 * Derek Atkins <derek@ihtfp.com>: Add Encapulation Support
71 * James Chapman : Add L2TP encapsulation type.
74 * This program is free software; you can redistribute it and/or
75 * modify it under the terms of the GNU General Public License
76 * as published by the Free Software Foundation; either version
77 * 2 of the License, or (at your option) any later version.
80 #include <asm/system.h>
81 #include <asm/uaccess.h>
82 #include <asm/ioctls.h>
83 #include <linux/bootmem.h>
84 #include <linux/highmem.h>
85 #include <linux/swap.h>
86 #include <linux/types.h>
87 #include <linux/fcntl.h>
88 #include <linux/module.h>
89 #include <linux/socket.h>
90 #include <linux/sockios.h>
91 #include <linux/igmp.h>
93 #include <linux/errno.h>
94 #include <linux/timer.h>
96 #include <linux/inet.h>
97 #include <linux/netdevice.h>
98 #include <net/tcp_states.h>
99 #include <linux/skbuff.h>
100 #include <linux/proc_fs.h>
101 #include <linux/seq_file.h>
102 #include <net/net_namespace.h>
103 #include <net/icmp.h>
104 #include <net/route.h>
105 #include <net/checksum.h>
106 #include <net/xfrm.h>
107 #include "udp_impl.h"
109 struct udp_table udp_table
;
110 EXPORT_SYMBOL(udp_table
);
112 int sysctl_udp_mem
[3] __read_mostly
;
113 int sysctl_udp_rmem_min __read_mostly
;
114 int sysctl_udp_wmem_min __read_mostly
;
116 EXPORT_SYMBOL(sysctl_udp_mem
);
117 EXPORT_SYMBOL(sysctl_udp_rmem_min
);
118 EXPORT_SYMBOL(sysctl_udp_wmem_min
);
120 atomic_t udp_memory_allocated
;
121 EXPORT_SYMBOL(udp_memory_allocated
);
123 static int udp_lib_lport_inuse(struct net
*net
, __u16 num
,
124 const struct udp_hslot
*hslot
,
126 int (*saddr_comp
)(const struct sock
*sk1
,
127 const struct sock
*sk2
))
130 struct hlist_nulls_node
*node
;
132 sk_nulls_for_each(sk2
, node
, &hslot
->head
)
133 if (net_eq(sock_net(sk2
), net
) &&
135 sk2
->sk_hash
== num
&&
136 (!sk2
->sk_reuse
|| !sk
->sk_reuse
) &&
137 (!sk2
->sk_bound_dev_if
|| !sk
->sk_bound_dev_if
138 || sk2
->sk_bound_dev_if
== sk
->sk_bound_dev_if
) &&
139 (*saddr_comp
)(sk
, sk2
))
145 * udp_lib_get_port - UDP/-Lite port lookup for IPv4 and IPv6
147 * @sk: socket struct in question
148 * @snum: port number to look up
149 * @saddr_comp: AF-dependent comparison of bound local IP addresses
151 int udp_lib_get_port(struct sock
*sk
, unsigned short snum
,
152 int (*saddr_comp
)(const struct sock
*sk1
,
153 const struct sock
*sk2
) )
155 struct udp_hslot
*hslot
;
156 struct udp_table
*udptable
= sk
->sk_prot
->h
.udp_table
;
158 struct net
*net
= sock_net(sk
);
161 int low
, high
, remaining
;
163 unsigned short first
;
165 inet_get_local_port_range(&low
, &high
);
166 remaining
= (high
- low
) + 1;
169 snum
= first
= rand
% remaining
+ low
;
172 hslot
= &udptable
->hash
[udp_hashfn(net
, snum
)];
173 spin_lock_bh(&hslot
->lock
);
174 if (!udp_lib_lport_inuse(net
, snum
, hslot
, sk
, saddr_comp
))
176 spin_unlock_bh(&hslot
->lock
);
179 } while (snum
< low
|| snum
> high
);
184 hslot
= &udptable
->hash
[udp_hashfn(net
, snum
)];
185 spin_lock_bh(&hslot
->lock
);
186 if (udp_lib_lport_inuse(net
, snum
, hslot
, sk
, saddr_comp
))
189 inet_sk(sk
)->num
= snum
;
191 if (sk_unhashed(sk
)) {
192 sk_nulls_add_node_rcu(sk
, &hslot
->head
);
193 sock_prot_inuse_add(sock_net(sk
), sk
->sk_prot
, 1);
197 spin_unlock_bh(&hslot
->lock
);
202 static int ipv4_rcv_saddr_equal(const struct sock
*sk1
, const struct sock
*sk2
)
204 struct inet_sock
*inet1
= inet_sk(sk1
), *inet2
= inet_sk(sk2
);
206 return ( !ipv6_only_sock(sk2
) &&
207 (!inet1
->rcv_saddr
|| !inet2
->rcv_saddr
||
208 inet1
->rcv_saddr
== inet2
->rcv_saddr
));
211 int udp_v4_get_port(struct sock
*sk
, unsigned short snum
)
213 return udp_lib_get_port(sk
, snum
, ipv4_rcv_saddr_equal
);
216 static inline int compute_score(struct sock
*sk
, struct net
*net
, __be32 saddr
,
218 __be16 sport
, __be32 daddr
, __be16 dport
, int dif
)
222 if (net_eq(sock_net(sk
), net
) && sk
->sk_hash
== hnum
&&
223 !ipv6_only_sock(sk
)) {
224 struct inet_sock
*inet
= inet_sk(sk
);
226 score
= (sk
->sk_family
== PF_INET
? 1 : 0);
227 if (inet
->rcv_saddr
) {
228 if (inet
->rcv_saddr
!= daddr
)
233 if (inet
->daddr
!= saddr
)
238 if (inet
->dport
!= sport
)
242 if (sk
->sk_bound_dev_if
) {
243 if (sk
->sk_bound_dev_if
!= dif
)
251 /* UDP is nearly always wildcards out the wazoo, it makes no sense to try
252 * harder than this. -DaveM
254 static struct sock
*__udp4_lib_lookup(struct net
*net
, __be32 saddr
,
255 __be16 sport
, __be32 daddr
, __be16 dport
,
256 int dif
, struct udp_table
*udptable
)
258 struct sock
*sk
, *result
;
259 struct hlist_nulls_node
*node
;
260 unsigned short hnum
= ntohs(dport
);
261 unsigned int hash
= udp_hashfn(net
, hnum
);
262 struct udp_hslot
*hslot
= &udptable
->hash
[hash
];
269 sk_nulls_for_each_rcu(sk
, node
, &hslot
->head
) {
270 score
= compute_score(sk
, net
, saddr
, hnum
, sport
,
272 if (score
> badness
) {
278 * if the nulls value we got at the end of this lookup is
279 * not the expected one, we must restart lookup.
280 * We probably met an item that was moved to another chain.
282 if (get_nulls_value(node
) != hash
)
286 if (unlikely(!atomic_inc_not_zero(&result
->sk_refcnt
)))
288 else if (unlikely(compute_score(result
, net
, saddr
, hnum
, sport
,
289 daddr
, dport
, dif
) < badness
)) {
298 static inline struct sock
*__udp4_lib_lookup_skb(struct sk_buff
*skb
,
299 __be16 sport
, __be16 dport
,
300 struct udp_table
*udptable
)
303 const struct iphdr
*iph
= ip_hdr(skb
);
305 if (unlikely(sk
= skb_steal_sock(skb
)))
308 return __udp4_lib_lookup(dev_net(skb
->dst
->dev
), iph
->saddr
, sport
,
309 iph
->daddr
, dport
, inet_iif(skb
),
313 struct sock
*udp4_lib_lookup(struct net
*net
, __be32 saddr
, __be16 sport
,
314 __be32 daddr
, __be16 dport
, int dif
)
316 return __udp4_lib_lookup(net
, saddr
, sport
, daddr
, dport
, dif
, &udp_table
);
318 EXPORT_SYMBOL_GPL(udp4_lib_lookup
);
320 static inline struct sock
*udp_v4_mcast_next(struct net
*net
, struct sock
*sk
,
321 __be16 loc_port
, __be32 loc_addr
,
322 __be16 rmt_port
, __be32 rmt_addr
,
325 struct hlist_nulls_node
*node
;
327 unsigned short hnum
= ntohs(loc_port
);
329 sk_nulls_for_each_from(s
, node
) {
330 struct inet_sock
*inet
= inet_sk(s
);
332 if (!net_eq(sock_net(s
), net
) ||
333 s
->sk_hash
!= hnum
||
334 (inet
->daddr
&& inet
->daddr
!= rmt_addr
) ||
335 (inet
->dport
!= rmt_port
&& inet
->dport
) ||
336 (inet
->rcv_saddr
&& inet
->rcv_saddr
!= loc_addr
) ||
338 (s
->sk_bound_dev_if
&& s
->sk_bound_dev_if
!= dif
))
340 if (!ip_mc_sf_allow(s
, loc_addr
, rmt_addr
, dif
))
350 * This routine is called by the ICMP module when it gets some
351 * sort of error condition. If err < 0 then the socket should
352 * be closed and the error returned to the user. If err > 0
353 * it's just the icmp type << 8 | icmp code.
354 * Header points to the ip header of the error packet. We move
355 * on past this. Then (as it used to claim before adjustment)
356 * header points to the first 8 bytes of the udp header. We need
357 * to find the appropriate port.
360 void __udp4_lib_err(struct sk_buff
*skb
, u32 info
, struct udp_table
*udptable
)
362 struct inet_sock
*inet
;
363 struct iphdr
*iph
= (struct iphdr
*)skb
->data
;
364 struct udphdr
*uh
= (struct udphdr
*)(skb
->data
+(iph
->ihl
<<2));
365 const int type
= icmp_hdr(skb
)->type
;
366 const int code
= icmp_hdr(skb
)->code
;
370 struct net
*net
= dev_net(skb
->dev
);
372 sk
= __udp4_lib_lookup(net
, iph
->daddr
, uh
->dest
,
373 iph
->saddr
, uh
->source
, skb
->dev
->ifindex
, udptable
);
375 ICMP_INC_STATS_BH(net
, ICMP_MIB_INERRORS
);
376 return; /* No socket for error */
385 case ICMP_TIME_EXCEEDED
:
388 case ICMP_SOURCE_QUENCH
:
390 case ICMP_PARAMETERPROB
:
394 case ICMP_DEST_UNREACH
:
395 if (code
== ICMP_FRAG_NEEDED
) { /* Path MTU discovery */
396 if (inet
->pmtudisc
!= IP_PMTUDISC_DONT
) {
404 if (code
<= NR_ICMP_UNREACH
) {
405 harderr
= icmp_err_convert
[code
].fatal
;
406 err
= icmp_err_convert
[code
].errno
;
412 * RFC1122: OK. Passes ICMP errors back to application, as per
415 if (!inet
->recverr
) {
416 if (!harderr
|| sk
->sk_state
!= TCP_ESTABLISHED
)
419 ip_icmp_error(sk
, skb
, err
, uh
->dest
, info
, (u8
*)(uh
+1));
422 sk
->sk_error_report(sk
);
427 void udp_err(struct sk_buff
*skb
, u32 info
)
429 __udp4_lib_err(skb
, info
, &udp_table
);
433 * Throw away all pending data and cancel the corking. Socket is locked.
435 void udp_flush_pending_frames(struct sock
*sk
)
437 struct udp_sock
*up
= udp_sk(sk
);
442 ip_flush_pending_frames(sk
);
445 EXPORT_SYMBOL(udp_flush_pending_frames
);
448 * udp4_hwcsum_outgoing - handle outgoing HW checksumming
449 * @sk: socket we are sending on
450 * @skb: sk_buff containing the filled-in UDP header
451 * (checksum field must be zeroed out)
453 static void udp4_hwcsum_outgoing(struct sock
*sk
, struct sk_buff
*skb
,
454 __be32 src
, __be32 dst
, int len
)
457 struct udphdr
*uh
= udp_hdr(skb
);
460 if (skb_queue_len(&sk
->sk_write_queue
) == 1) {
462 * Only one fragment on the socket.
464 skb
->csum_start
= skb_transport_header(skb
) - skb
->head
;
465 skb
->csum_offset
= offsetof(struct udphdr
, check
);
466 uh
->check
= ~csum_tcpudp_magic(src
, dst
, len
, IPPROTO_UDP
, 0);
469 * HW-checksum won't work as there are two or more
470 * fragments on the socket so that all csums of sk_buffs
473 offset
= skb_transport_offset(skb
);
474 skb
->csum
= skb_checksum(skb
, offset
, skb
->len
- offset
, 0);
476 skb
->ip_summed
= CHECKSUM_NONE
;
478 skb_queue_walk(&sk
->sk_write_queue
, skb
) {
479 csum
= csum_add(csum
, skb
->csum
);
482 uh
->check
= csum_tcpudp_magic(src
, dst
, len
, IPPROTO_UDP
, csum
);
484 uh
->check
= CSUM_MANGLED_0
;
489 * Push out all pending data as one UDP datagram. Socket is locked.
491 static int udp_push_pending_frames(struct sock
*sk
)
493 struct udp_sock
*up
= udp_sk(sk
);
494 struct inet_sock
*inet
= inet_sk(sk
);
495 struct flowi
*fl
= &inet
->cork
.fl
;
499 int is_udplite
= IS_UDPLITE(sk
);
502 /* Grab the skbuff where UDP header space exists. */
503 if ((skb
= skb_peek(&sk
->sk_write_queue
)) == NULL
)
507 * Create a UDP header
510 uh
->source
= fl
->fl_ip_sport
;
511 uh
->dest
= fl
->fl_ip_dport
;
512 uh
->len
= htons(up
->len
);
515 if (is_udplite
) /* UDP-Lite */
516 csum
= udplite_csum_outgoing(sk
, skb
);
518 else if (sk
->sk_no_check
== UDP_CSUM_NOXMIT
) { /* UDP csum disabled */
520 skb
->ip_summed
= CHECKSUM_NONE
;
523 } else if (skb
->ip_summed
== CHECKSUM_PARTIAL
) { /* UDP hardware csum */
525 udp4_hwcsum_outgoing(sk
, skb
, fl
->fl4_src
,fl
->fl4_dst
, up
->len
);
528 } else /* `normal' UDP */
529 csum
= udp_csum_outgoing(sk
, skb
);
531 /* add protocol-dependent pseudo-header */
532 uh
->check
= csum_tcpudp_magic(fl
->fl4_src
, fl
->fl4_dst
, up
->len
,
533 sk
->sk_protocol
, csum
);
535 uh
->check
= CSUM_MANGLED_0
;
538 err
= ip_push_pending_frames(sk
);
543 UDP_INC_STATS_USER(sock_net(sk
),
544 UDP_MIB_OUTDATAGRAMS
, is_udplite
);
548 int udp_sendmsg(struct kiocb
*iocb
, struct sock
*sk
, struct msghdr
*msg
,
551 struct inet_sock
*inet
= inet_sk(sk
);
552 struct udp_sock
*up
= udp_sk(sk
);
554 struct ipcm_cookie ipc
;
555 struct rtable
*rt
= NULL
;
558 __be32 daddr
, faddr
, saddr
;
561 int err
, is_udplite
= IS_UDPLITE(sk
);
562 int corkreq
= up
->corkflag
|| msg
->msg_flags
&MSG_MORE
;
563 int (*getfrag
)(void *, char *, int, int, int, struct sk_buff
*);
572 if (msg
->msg_flags
&MSG_OOB
) /* Mirror BSD error message compatibility */
579 * There are pending frames.
580 * The socket lock must be held while it's corked.
583 if (likely(up
->pending
)) {
584 if (unlikely(up
->pending
!= AF_INET
)) {
592 ulen
+= sizeof(struct udphdr
);
595 * Get and verify the address.
598 struct sockaddr_in
* usin
= (struct sockaddr_in
*)msg
->msg_name
;
599 if (msg
->msg_namelen
< sizeof(*usin
))
601 if (usin
->sin_family
!= AF_INET
) {
602 if (usin
->sin_family
!= AF_UNSPEC
)
603 return -EAFNOSUPPORT
;
606 daddr
= usin
->sin_addr
.s_addr
;
607 dport
= usin
->sin_port
;
611 if (sk
->sk_state
!= TCP_ESTABLISHED
)
612 return -EDESTADDRREQ
;
615 /* Open fast path for connected socket.
616 Route will not be used, if at least one option is set.
620 ipc
.addr
= inet
->saddr
;
622 ipc
.oif
= sk
->sk_bound_dev_if
;
623 if (msg
->msg_controllen
) {
624 err
= ip_cmsg_send(sock_net(sk
), msg
, &ipc
);
635 ipc
.addr
= faddr
= daddr
;
637 if (ipc
.opt
&& ipc
.opt
->srr
) {
640 faddr
= ipc
.opt
->faddr
;
643 tos
= RT_TOS(inet
->tos
);
644 if (sock_flag(sk
, SOCK_LOCALROUTE
) ||
645 (msg
->msg_flags
& MSG_DONTROUTE
) ||
646 (ipc
.opt
&& ipc
.opt
->is_strictroute
)) {
651 if (ipv4_is_multicast(daddr
)) {
653 ipc
.oif
= inet
->mc_index
;
655 saddr
= inet
->mc_addr
;
660 rt
= (struct rtable
*)sk_dst_check(sk
, 0);
663 struct flowi fl
= { .oif
= ipc
.oif
,
668 .proto
= sk
->sk_protocol
,
669 .flags
= inet_sk_flowi_flags(sk
),
671 { .sport
= inet
->sport
,
672 .dport
= dport
} } };
673 struct net
*net
= sock_net(sk
);
675 security_sk_classify_flow(sk
, &fl
);
676 err
= ip_route_output_flow(net
, &rt
, &fl
, sk
, 1);
678 if (err
== -ENETUNREACH
)
679 IP_INC_STATS_BH(net
, IPSTATS_MIB_OUTNOROUTES
);
684 if ((rt
->rt_flags
& RTCF_BROADCAST
) &&
685 !sock_flag(sk
, SOCK_BROADCAST
))
688 sk_dst_set(sk
, dst_clone(&rt
->u
.dst
));
691 if (msg
->msg_flags
&MSG_CONFIRM
)
697 daddr
= ipc
.addr
= rt
->rt_dst
;
700 if (unlikely(up
->pending
)) {
701 /* The socket is already corked while preparing it. */
702 /* ... which is an evident application bug. --ANK */
705 LIMIT_NETDEBUG(KERN_DEBUG
"udp cork app bug 2\n");
710 * Now cork the socket to pend data.
712 inet
->cork
.fl
.fl4_dst
= daddr
;
713 inet
->cork
.fl
.fl_ip_dport
= dport
;
714 inet
->cork
.fl
.fl4_src
= saddr
;
715 inet
->cork
.fl
.fl_ip_sport
= inet
->sport
;
716 up
->pending
= AF_INET
;
720 getfrag
= is_udplite
? udplite_getfrag
: ip_generic_getfrag
;
721 err
= ip_append_data(sk
, getfrag
, msg
->msg_iov
, ulen
,
722 sizeof(struct udphdr
), &ipc
, &rt
,
723 corkreq
? msg
->msg_flags
|MSG_MORE
: msg
->msg_flags
);
725 udp_flush_pending_frames(sk
);
727 err
= udp_push_pending_frames(sk
);
728 else if (unlikely(skb_queue_empty(&sk
->sk_write_queue
)))
739 * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space. Reporting
740 * ENOBUFS might not be good (it's not tunable per se), but otherwise
741 * we don't have a good statistic (IpOutDiscards but it can be too many
742 * things). We could add another new stat but at least for now that
743 * seems like overkill.
745 if (err
== -ENOBUFS
|| test_bit(SOCK_NOSPACE
, &sk
->sk_socket
->flags
)) {
746 UDP_INC_STATS_USER(sock_net(sk
),
747 UDP_MIB_SNDBUFERRORS
, is_udplite
);
752 dst_confirm(&rt
->u
.dst
);
753 if (!(msg
->msg_flags
&MSG_PROBE
) || len
)
754 goto back_from_confirm
;
759 int udp_sendpage(struct sock
*sk
, struct page
*page
, int offset
,
760 size_t size
, int flags
)
762 struct udp_sock
*up
= udp_sk(sk
);
766 struct msghdr msg
= { .msg_flags
= flags
|MSG_MORE
};
768 /* Call udp_sendmsg to specify destination address which
769 * sendpage interface can't pass.
770 * This will succeed only when the socket is connected.
772 ret
= udp_sendmsg(NULL
, sk
, &msg
, 0);
779 if (unlikely(!up
->pending
)) {
782 LIMIT_NETDEBUG(KERN_DEBUG
"udp cork app bug 3\n");
786 ret
= ip_append_page(sk
, page
, offset
, size
, flags
);
787 if (ret
== -EOPNOTSUPP
) {
789 return sock_no_sendpage(sk
->sk_socket
, page
, offset
,
793 udp_flush_pending_frames(sk
);
798 if (!(up
->corkflag
|| (flags
&MSG_MORE
)))
799 ret
= udp_push_pending_frames(sk
);
808 * IOCTL requests applicable to the UDP protocol
811 int udp_ioctl(struct sock
*sk
, int cmd
, unsigned long arg
)
816 int amount
= atomic_read(&sk
->sk_wmem_alloc
);
817 return put_user(amount
, (int __user
*)arg
);
823 unsigned long amount
;
826 spin_lock_bh(&sk
->sk_receive_queue
.lock
);
827 skb
= skb_peek(&sk
->sk_receive_queue
);
830 * We will only return the amount
831 * of this packet since that is all
834 amount
= skb
->len
- sizeof(struct udphdr
);
836 spin_unlock_bh(&sk
->sk_receive_queue
.lock
);
837 return put_user(amount
, (int __user
*)arg
);
848 * This should be easy, if there is something there we
849 * return it, otherwise we block.
852 int udp_recvmsg(struct kiocb
*iocb
, struct sock
*sk
, struct msghdr
*msg
,
853 size_t len
, int noblock
, int flags
, int *addr_len
)
855 struct inet_sock
*inet
= inet_sk(sk
);
856 struct sockaddr_in
*sin
= (struct sockaddr_in
*)msg
->msg_name
;
858 unsigned int ulen
, copied
;
861 int is_udplite
= IS_UDPLITE(sk
);
864 * Check any passed addresses
867 *addr_len
=sizeof(*sin
);
869 if (flags
& MSG_ERRQUEUE
)
870 return ip_recv_error(sk
, msg
, len
);
873 skb
= __skb_recv_datagram(sk
, flags
| (noblock
? MSG_DONTWAIT
: 0),
878 ulen
= skb
->len
- sizeof(struct udphdr
);
882 else if (copied
< ulen
)
883 msg
->msg_flags
|= MSG_TRUNC
;
886 * If checksum is needed at all, try to do it while copying the
887 * data. If the data is truncated, or if we only want a partial
888 * coverage checksum (UDP-Lite), do it before the copy.
891 if (copied
< ulen
|| UDP_SKB_CB(skb
)->partial_cov
) {
892 if (udp_lib_checksum_complete(skb
))
896 if (skb_csum_unnecessary(skb
))
897 err
= skb_copy_datagram_iovec(skb
, sizeof(struct udphdr
),
898 msg
->msg_iov
, copied
);
900 err
= skb_copy_and_csum_datagram_iovec(skb
, sizeof(struct udphdr
), msg
->msg_iov
);
910 UDP_INC_STATS_USER(sock_net(sk
),
911 UDP_MIB_INDATAGRAMS
, is_udplite
);
913 sock_recv_timestamp(msg
, sk
, skb
);
915 /* Copy the address. */
918 sin
->sin_family
= AF_INET
;
919 sin
->sin_port
= udp_hdr(skb
)->source
;
920 sin
->sin_addr
.s_addr
= ip_hdr(skb
)->saddr
;
921 memset(sin
->sin_zero
, 0, sizeof(sin
->sin_zero
));
923 if (inet
->cmsg_flags
)
924 ip_cmsg_recv(msg
, skb
);
927 if (flags
& MSG_TRUNC
)
932 skb_free_datagram(sk
, skb
);
939 if (!skb_kill_datagram(sk
, skb
, flags
))
940 UDP_INC_STATS_USER(sock_net(sk
), UDP_MIB_INERRORS
, is_udplite
);
949 int udp_disconnect(struct sock
*sk
, int flags
)
951 struct inet_sock
*inet
= inet_sk(sk
);
953 * 1003.1g - break association.
956 sk
->sk_state
= TCP_CLOSE
;
959 sk
->sk_bound_dev_if
= 0;
960 if (!(sk
->sk_userlocks
& SOCK_BINDADDR_LOCK
))
961 inet_reset_saddr(sk
);
963 if (!(sk
->sk_userlocks
& SOCK_BINDPORT_LOCK
)) {
964 sk
->sk_prot
->unhash(sk
);
971 void udp_lib_unhash(struct sock
*sk
)
974 struct udp_table
*udptable
= sk
->sk_prot
->h
.udp_table
;
975 unsigned int hash
= udp_hashfn(sock_net(sk
), sk
->sk_hash
);
976 struct udp_hslot
*hslot
= &udptable
->hash
[hash
];
978 spin_lock_bh(&hslot
->lock
);
979 if (sk_nulls_del_node_init_rcu(sk
)) {
980 inet_sk(sk
)->num
= 0;
981 sock_prot_inuse_add(sock_net(sk
), sk
->sk_prot
, -1);
983 spin_unlock_bh(&hslot
->lock
);
986 EXPORT_SYMBOL(udp_lib_unhash
);
988 static int __udp_queue_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
)
990 int is_udplite
= IS_UDPLITE(sk
);
993 if ((rc
= sock_queue_rcv_skb(sk
, skb
)) < 0) {
994 /* Note that an ENOMEM error is charged twice */
996 UDP_INC_STATS_BH(sock_net(sk
), UDP_MIB_RCVBUFERRORS
,
1004 UDP_INC_STATS_BH(sock_net(sk
), UDP_MIB_INERRORS
, is_udplite
);
1012 * >0: "udp encap" protocol resubmission
1014 * Note that in the success and error cases, the skb is assumed to
1015 * have either been requeued or freed.
1017 int udp_queue_rcv_skb(struct sock
* sk
, struct sk_buff
*skb
)
1019 struct udp_sock
*up
= udp_sk(sk
);
1021 int is_udplite
= IS_UDPLITE(sk
);
1024 * Charge it to the socket, dropping if the queue is full.
1026 if (!xfrm4_policy_check(sk
, XFRM_POLICY_IN
, skb
))
1030 if (up
->encap_type
) {
1032 * This is an encapsulation socket so pass the skb to
1033 * the socket's udp_encap_rcv() hook. Otherwise, just
1034 * fall through and pass this up the UDP socket.
1035 * up->encap_rcv() returns the following value:
1036 * =0 if skb was successfully passed to the encap
1037 * handler or was discarded by it.
1038 * >0 if skb should be passed on to UDP.
1039 * <0 if skb should be resubmitted as proto -N
1042 /* if we're overly short, let UDP handle it */
1043 if (skb
->len
> sizeof(struct udphdr
) &&
1044 up
->encap_rcv
!= NULL
) {
1047 ret
= (*up
->encap_rcv
)(sk
, skb
);
1049 UDP_INC_STATS_BH(sock_net(sk
),
1050 UDP_MIB_INDATAGRAMS
,
1056 /* FALLTHROUGH -- it's a UDP Packet */
1060 * UDP-Lite specific tests, ignored on UDP sockets
1062 if ((is_udplite
& UDPLITE_RECV_CC
) && UDP_SKB_CB(skb
)->partial_cov
) {
1065 * MIB statistics other than incrementing the error count are
1066 * disabled for the following two types of errors: these depend
1067 * on the application settings, not on the functioning of the
1068 * protocol stack as such.
1070 * RFC 3828 here recommends (sec 3.3): "There should also be a
1071 * way ... to ... at least let the receiving application block
1072 * delivery of packets with coverage values less than a value
1073 * provided by the application."
1075 if (up
->pcrlen
== 0) { /* full coverage was set */
1076 LIMIT_NETDEBUG(KERN_WARNING
"UDPLITE: partial coverage "
1077 "%d while full coverage %d requested\n",
1078 UDP_SKB_CB(skb
)->cscov
, skb
->len
);
1081 /* The next case involves violating the min. coverage requested
1082 * by the receiver. This is subtle: if receiver wants x and x is
1083 * greater than the buffersize/MTU then receiver will complain
1084 * that it wants x while sender emits packets of smaller size y.
1085 * Therefore the above ...()->partial_cov statement is essential.
1087 if (UDP_SKB_CB(skb
)->cscov
< up
->pcrlen
) {
1088 LIMIT_NETDEBUG(KERN_WARNING
1089 "UDPLITE: coverage %d too small, need min %d\n",
1090 UDP_SKB_CB(skb
)->cscov
, up
->pcrlen
);
1095 if (sk
->sk_filter
) {
1096 if (udp_lib_checksum_complete(skb
))
1103 if (!sock_owned_by_user(sk
))
1104 rc
= __udp_queue_rcv_skb(sk
, skb
);
1106 sk_add_backlog(sk
, skb
);
1112 UDP_INC_STATS_BH(sock_net(sk
), UDP_MIB_INERRORS
, is_udplite
);
1118 * Multicasts and broadcasts go to each listener.
1120 * Note: called only from the BH handler context,
1121 * so we don't need to lock the hashes.
1123 static int __udp4_lib_mcast_deliver(struct net
*net
, struct sk_buff
*skb
,
1125 __be32 saddr
, __be32 daddr
,
1126 struct udp_table
*udptable
)
1129 struct udp_hslot
*hslot
= &udptable
->hash
[udp_hashfn(net
, ntohs(uh
->dest
))];
1132 spin_lock(&hslot
->lock
);
1133 sk
= sk_nulls_head(&hslot
->head
);
1134 dif
= skb
->dev
->ifindex
;
1135 sk
= udp_v4_mcast_next(net
, sk
, uh
->dest
, daddr
, uh
->source
, saddr
, dif
);
1137 struct sock
*sknext
= NULL
;
1140 struct sk_buff
*skb1
= skb
;
1142 sknext
= udp_v4_mcast_next(net
, sk_nulls_next(sk
), uh
->dest
,
1143 daddr
, uh
->source
, saddr
,
1146 skb1
= skb_clone(skb
, GFP_ATOMIC
);
1149 int ret
= udp_queue_rcv_skb(sk
, skb1
);
1151 /* we should probably re-process instead
1152 * of dropping packets here. */
1159 spin_unlock(&hslot
->lock
);
1163 /* Initialize UDP checksum. If exited with zero value (success),
1164 * CHECKSUM_UNNECESSARY means, that no more checks are required.
1165 * Otherwise, csum completion requires chacksumming packet body,
1166 * including udp header and folding it to skb->csum.
1168 static inline int udp4_csum_init(struct sk_buff
*skb
, struct udphdr
*uh
,
1171 const struct iphdr
*iph
;
1174 UDP_SKB_CB(skb
)->partial_cov
= 0;
1175 UDP_SKB_CB(skb
)->cscov
= skb
->len
;
1177 if (proto
== IPPROTO_UDPLITE
) {
1178 err
= udplite_checksum_init(skb
, uh
);
1184 if (uh
->check
== 0) {
1185 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
1186 } else if (skb
->ip_summed
== CHECKSUM_COMPLETE
) {
1187 if (!csum_tcpudp_magic(iph
->saddr
, iph
->daddr
, skb
->len
,
1189 skb
->ip_summed
= CHECKSUM_UNNECESSARY
;
1191 if (!skb_csum_unnecessary(skb
))
1192 skb
->csum
= csum_tcpudp_nofold(iph
->saddr
, iph
->daddr
,
1193 skb
->len
, proto
, 0);
1194 /* Probably, we should checksum udp header (it should be in cache
1195 * in any case) and data in tiny packets (< rx copybreak).
1202 * All we need to do is get the socket, and then do a checksum.
1205 int __udp4_lib_rcv(struct sk_buff
*skb
, struct udp_table
*udptable
,
1209 struct udphdr
*uh
= udp_hdr(skb
);
1210 unsigned short ulen
;
1211 struct rtable
*rt
= (struct rtable
*)skb
->dst
;
1212 __be32 saddr
= ip_hdr(skb
)->saddr
;
1213 __be32 daddr
= ip_hdr(skb
)->daddr
;
1214 struct net
*net
= dev_net(skb
->dev
);
1217 * Validate the packet.
1219 if (!pskb_may_pull(skb
, sizeof(struct udphdr
)))
1220 goto drop
; /* No space for header. */
1222 ulen
= ntohs(uh
->len
);
1223 if (ulen
> skb
->len
)
1226 if (proto
== IPPROTO_UDP
) {
1227 /* UDP validates ulen. */
1228 if (ulen
< sizeof(*uh
) || pskb_trim_rcsum(skb
, ulen
))
1233 if (udp4_csum_init(skb
, uh
, proto
))
1236 if (rt
->rt_flags
& (RTCF_BROADCAST
|RTCF_MULTICAST
))
1237 return __udp4_lib_mcast_deliver(net
, skb
, uh
,
1238 saddr
, daddr
, udptable
);
1240 sk
= __udp4_lib_lookup_skb(skb
, uh
->source
, uh
->dest
, udptable
);
1243 int ret
= udp_queue_rcv_skb(sk
, skb
);
1246 /* a return value > 0 means to resubmit the input, but
1247 * it wants the return to be -protocol, or 0
1254 if (!xfrm4_policy_check(NULL
, XFRM_POLICY_IN
, skb
))
1258 /* No socket. Drop packet silently, if checksum is wrong */
1259 if (udp_lib_checksum_complete(skb
))
1262 UDP_INC_STATS_BH(net
, UDP_MIB_NOPORTS
, proto
== IPPROTO_UDPLITE
);
1263 icmp_send(skb
, ICMP_DEST_UNREACH
, ICMP_PORT_UNREACH
, 0);
1266 * Hmm. We got an UDP packet to a port to which we
1267 * don't wanna listen. Ignore it.
1273 LIMIT_NETDEBUG(KERN_DEBUG
"UDP%s: short packet: From %pI4:%u %d/%d to %pI4:%u\n",
1274 proto
== IPPROTO_UDPLITE
? "-Lite" : "",
1285 * RFC1122: OK. Discards the bad packet silently (as far as
1286 * the network is concerned, anyway) as per 4.1.3.4 (MUST).
1288 LIMIT_NETDEBUG(KERN_DEBUG
"UDP%s: bad checksum. From %pI4:%u to %pI4:%u ulen %d\n",
1289 proto
== IPPROTO_UDPLITE
? "-Lite" : "",
1296 UDP_INC_STATS_BH(net
, UDP_MIB_INERRORS
, proto
== IPPROTO_UDPLITE
);
1301 int udp_rcv(struct sk_buff
*skb
)
1303 return __udp4_lib_rcv(skb
, &udp_table
, IPPROTO_UDP
);
1306 void udp_destroy_sock(struct sock
*sk
)
1309 udp_flush_pending_frames(sk
);
1314 * Socket option code for UDP
1316 int udp_lib_setsockopt(struct sock
*sk
, int level
, int optname
,
1317 char __user
*optval
, int optlen
,
1318 int (*push_pending_frames
)(struct sock
*))
1320 struct udp_sock
*up
= udp_sk(sk
);
1323 int is_udplite
= IS_UDPLITE(sk
);
1325 if (optlen
<sizeof(int))
1328 if (get_user(val
, (int __user
*)optval
))
1338 (*push_pending_frames
)(sk
);
1346 case UDP_ENCAP_ESPINUDP
:
1347 case UDP_ENCAP_ESPINUDP_NON_IKE
:
1348 up
->encap_rcv
= xfrm4_udp_encap_rcv
;
1350 case UDP_ENCAP_L2TPINUDP
:
1351 up
->encap_type
= val
;
1360 * UDP-Lite's partial checksum coverage (RFC 3828).
1362 /* The sender sets actual checksum coverage length via this option.
1363 * The case coverage > packet length is handled by send module. */
1364 case UDPLITE_SEND_CSCOV
:
1365 if (!is_udplite
) /* Disable the option on UDP sockets */
1366 return -ENOPROTOOPT
;
1367 if (val
!= 0 && val
< 8) /* Illegal coverage: use default (8) */
1369 else if (val
> USHORT_MAX
)
1372 up
->pcflag
|= UDPLITE_SEND_CC
;
1375 /* The receiver specifies a minimum checksum coverage value. To make
1376 * sense, this should be set to at least 8 (as done below). If zero is
1377 * used, this again means full checksum coverage. */
1378 case UDPLITE_RECV_CSCOV
:
1379 if (!is_udplite
) /* Disable the option on UDP sockets */
1380 return -ENOPROTOOPT
;
1381 if (val
!= 0 && val
< 8) /* Avoid silly minimal values. */
1383 else if (val
> USHORT_MAX
)
1386 up
->pcflag
|= UDPLITE_RECV_CC
;
1397 int udp_setsockopt(struct sock
*sk
, int level
, int optname
,
1398 char __user
*optval
, int optlen
)
1400 if (level
== SOL_UDP
|| level
== SOL_UDPLITE
)
1401 return udp_lib_setsockopt(sk
, level
, optname
, optval
, optlen
,
1402 udp_push_pending_frames
);
1403 return ip_setsockopt(sk
, level
, optname
, optval
, optlen
);
1406 #ifdef CONFIG_COMPAT
1407 int compat_udp_setsockopt(struct sock
*sk
, int level
, int optname
,
1408 char __user
*optval
, int optlen
)
1410 if (level
== SOL_UDP
|| level
== SOL_UDPLITE
)
1411 return udp_lib_setsockopt(sk
, level
, optname
, optval
, optlen
,
1412 udp_push_pending_frames
);
1413 return compat_ip_setsockopt(sk
, level
, optname
, optval
, optlen
);
1417 int udp_lib_getsockopt(struct sock
*sk
, int level
, int optname
,
1418 char __user
*optval
, int __user
*optlen
)
1420 struct udp_sock
*up
= udp_sk(sk
);
1423 if (get_user(len
,optlen
))
1426 len
= min_t(unsigned int, len
, sizeof(int));
1437 val
= up
->encap_type
;
1440 /* The following two cannot be changed on UDP sockets, the return is
1441 * always 0 (which corresponds to the full checksum coverage of UDP). */
1442 case UDPLITE_SEND_CSCOV
:
1446 case UDPLITE_RECV_CSCOV
:
1451 return -ENOPROTOOPT
;
1454 if (put_user(len
, optlen
))
1456 if (copy_to_user(optval
, &val
,len
))
1461 int udp_getsockopt(struct sock
*sk
, int level
, int optname
,
1462 char __user
*optval
, int __user
*optlen
)
1464 if (level
== SOL_UDP
|| level
== SOL_UDPLITE
)
1465 return udp_lib_getsockopt(sk
, level
, optname
, optval
, optlen
);
1466 return ip_getsockopt(sk
, level
, optname
, optval
, optlen
);
1469 #ifdef CONFIG_COMPAT
1470 int compat_udp_getsockopt(struct sock
*sk
, int level
, int optname
,
1471 char __user
*optval
, int __user
*optlen
)
1473 if (level
== SOL_UDP
|| level
== SOL_UDPLITE
)
1474 return udp_lib_getsockopt(sk
, level
, optname
, optval
, optlen
);
1475 return compat_ip_getsockopt(sk
, level
, optname
, optval
, optlen
);
1479 * udp_poll - wait for a UDP event.
1480 * @file - file struct
1482 * @wait - poll table
1484 * This is same as datagram poll, except for the special case of
1485 * blocking sockets. If application is using a blocking fd
1486 * and a packet with checksum error is in the queue;
1487 * then it could get return from select indicating data available
1488 * but then block when reading it. Add special case code
1489 * to work around these arguably broken applications.
1491 unsigned int udp_poll(struct file
*file
, struct socket
*sock
, poll_table
*wait
)
1493 unsigned int mask
= datagram_poll(file
, sock
, wait
);
1494 struct sock
*sk
= sock
->sk
;
1495 int is_lite
= IS_UDPLITE(sk
);
1497 /* Check for false positives due to checksum errors */
1498 if ( (mask
& POLLRDNORM
) &&
1499 !(file
->f_flags
& O_NONBLOCK
) &&
1500 !(sk
->sk_shutdown
& RCV_SHUTDOWN
)){
1501 struct sk_buff_head
*rcvq
= &sk
->sk_receive_queue
;
1502 struct sk_buff
*skb
;
1504 spin_lock_bh(&rcvq
->lock
);
1505 while ((skb
= skb_peek(rcvq
)) != NULL
&&
1506 udp_lib_checksum_complete(skb
)) {
1507 UDP_INC_STATS_BH(sock_net(sk
),
1508 UDP_MIB_INERRORS
, is_lite
);
1509 __skb_unlink(skb
, rcvq
);
1512 spin_unlock_bh(&rcvq
->lock
);
1514 /* nothing to see, move along */
1516 mask
&= ~(POLLIN
| POLLRDNORM
);
1523 struct proto udp_prot
= {
1525 .owner
= THIS_MODULE
,
1526 .close
= udp_lib_close
,
1527 .connect
= ip4_datagram_connect
,
1528 .disconnect
= udp_disconnect
,
1530 .destroy
= udp_destroy_sock
,
1531 .setsockopt
= udp_setsockopt
,
1532 .getsockopt
= udp_getsockopt
,
1533 .sendmsg
= udp_sendmsg
,
1534 .recvmsg
= udp_recvmsg
,
1535 .sendpage
= udp_sendpage
,
1536 .backlog_rcv
= __udp_queue_rcv_skb
,
1537 .hash
= udp_lib_hash
,
1538 .unhash
= udp_lib_unhash
,
1539 .get_port
= udp_v4_get_port
,
1540 .memory_allocated
= &udp_memory_allocated
,
1541 .sysctl_mem
= sysctl_udp_mem
,
1542 .sysctl_wmem
= &sysctl_udp_wmem_min
,
1543 .sysctl_rmem
= &sysctl_udp_rmem_min
,
1544 .obj_size
= sizeof(struct udp_sock
),
1545 .slab_flags
= SLAB_DESTROY_BY_RCU
,
1546 .h
.udp_table
= &udp_table
,
1547 #ifdef CONFIG_COMPAT
1548 .compat_setsockopt
= compat_udp_setsockopt
,
1549 .compat_getsockopt
= compat_udp_getsockopt
,
1553 /* ------------------------------------------------------------------------ */
1554 #ifdef CONFIG_PROC_FS
1556 static struct sock
*udp_get_first(struct seq_file
*seq
, int start
)
1559 struct udp_iter_state
*state
= seq
->private;
1560 struct net
*net
= seq_file_net(seq
);
1562 for (state
->bucket
= start
; state
->bucket
< UDP_HTABLE_SIZE
; ++state
->bucket
) {
1563 struct hlist_nulls_node
*node
;
1564 struct udp_hslot
*hslot
= &state
->udp_table
->hash
[state
->bucket
];
1565 spin_lock_bh(&hslot
->lock
);
1566 sk_nulls_for_each(sk
, node
, &hslot
->head
) {
1567 if (!net_eq(sock_net(sk
), net
))
1569 if (sk
->sk_family
== state
->family
)
1572 spin_unlock_bh(&hslot
->lock
);
1579 static struct sock
*udp_get_next(struct seq_file
*seq
, struct sock
*sk
)
1581 struct udp_iter_state
*state
= seq
->private;
1582 struct net
*net
= seq_file_net(seq
);
1585 sk
= sk_nulls_next(sk
);
1586 } while (sk
&& (!net_eq(sock_net(sk
), net
) || sk
->sk_family
!= state
->family
));
1589 spin_unlock_bh(&state
->udp_table
->hash
[state
->bucket
].lock
);
1590 return udp_get_first(seq
, state
->bucket
+ 1);
1595 static struct sock
*udp_get_idx(struct seq_file
*seq
, loff_t pos
)
1597 struct sock
*sk
= udp_get_first(seq
, 0);
1600 while (pos
&& (sk
= udp_get_next(seq
, sk
)) != NULL
)
1602 return pos
? NULL
: sk
;
1605 static void *udp_seq_start(struct seq_file
*seq
, loff_t
*pos
)
1607 return *pos
? udp_get_idx(seq
, *pos
-1) : SEQ_START_TOKEN
;
1610 static void *udp_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
1614 if (v
== SEQ_START_TOKEN
)
1615 sk
= udp_get_idx(seq
, 0);
1617 sk
= udp_get_next(seq
, v
);
1623 static void udp_seq_stop(struct seq_file
*seq
, void *v
)
1625 struct udp_iter_state
*state
= seq
->private;
1627 if (state
->bucket
< UDP_HTABLE_SIZE
)
1628 spin_unlock_bh(&state
->udp_table
->hash
[state
->bucket
].lock
);
1631 static int udp_seq_open(struct inode
*inode
, struct file
*file
)
1633 struct udp_seq_afinfo
*afinfo
= PDE(inode
)->data
;
1634 struct udp_iter_state
*s
;
1637 err
= seq_open_net(inode
, file
, &afinfo
->seq_ops
,
1638 sizeof(struct udp_iter_state
));
1642 s
= ((struct seq_file
*)file
->private_data
)->private;
1643 s
->family
= afinfo
->family
;
1644 s
->udp_table
= afinfo
->udp_table
;
1648 /* ------------------------------------------------------------------------ */
1649 int udp_proc_register(struct net
*net
, struct udp_seq_afinfo
*afinfo
)
1651 struct proc_dir_entry
*p
;
1654 afinfo
->seq_fops
.open
= udp_seq_open
;
1655 afinfo
->seq_fops
.read
= seq_read
;
1656 afinfo
->seq_fops
.llseek
= seq_lseek
;
1657 afinfo
->seq_fops
.release
= seq_release_net
;
1659 afinfo
->seq_ops
.start
= udp_seq_start
;
1660 afinfo
->seq_ops
.next
= udp_seq_next
;
1661 afinfo
->seq_ops
.stop
= udp_seq_stop
;
1663 p
= proc_create_data(afinfo
->name
, S_IRUGO
, net
->proc_net
,
1664 &afinfo
->seq_fops
, afinfo
);
1670 void udp_proc_unregister(struct net
*net
, struct udp_seq_afinfo
*afinfo
)
1672 proc_net_remove(net
, afinfo
->name
);
1675 /* ------------------------------------------------------------------------ */
1676 static void udp4_format_sock(struct sock
*sp
, struct seq_file
*f
,
1677 int bucket
, int *len
)
1679 struct inet_sock
*inet
= inet_sk(sp
);
1680 __be32 dest
= inet
->daddr
;
1681 __be32 src
= inet
->rcv_saddr
;
1682 __u16 destp
= ntohs(inet
->dport
);
1683 __u16 srcp
= ntohs(inet
->sport
);
1685 seq_printf(f
, "%4d: %08X:%04X %08X:%04X"
1686 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d%n",
1687 bucket
, src
, srcp
, dest
, destp
, sp
->sk_state
,
1688 atomic_read(&sp
->sk_wmem_alloc
),
1689 atomic_read(&sp
->sk_rmem_alloc
),
1690 0, 0L, 0, sock_i_uid(sp
), 0, sock_i_ino(sp
),
1691 atomic_read(&sp
->sk_refcnt
), sp
,
1692 atomic_read(&sp
->sk_drops
), len
);
1695 int udp4_seq_show(struct seq_file
*seq
, void *v
)
1697 if (v
== SEQ_START_TOKEN
)
1698 seq_printf(seq
, "%-127s\n",
1699 " sl local_address rem_address st tx_queue "
1700 "rx_queue tr tm->when retrnsmt uid timeout "
1701 "inode ref pointer drops");
1703 struct udp_iter_state
*state
= seq
->private;
1706 udp4_format_sock(v
, seq
, state
->bucket
, &len
);
1707 seq_printf(seq
, "%*s\n", 127 - len
,"");
1712 /* ------------------------------------------------------------------------ */
1713 static struct udp_seq_afinfo udp4_seq_afinfo
= {
1716 .udp_table
= &udp_table
,
1718 .owner
= THIS_MODULE
,
1721 .show
= udp4_seq_show
,
1725 static int udp4_proc_init_net(struct net
*net
)
1727 return udp_proc_register(net
, &udp4_seq_afinfo
);
1730 static void udp4_proc_exit_net(struct net
*net
)
1732 udp_proc_unregister(net
, &udp4_seq_afinfo
);
1735 static struct pernet_operations udp4_net_ops
= {
1736 .init
= udp4_proc_init_net
,
1737 .exit
= udp4_proc_exit_net
,
1740 int __init
udp4_proc_init(void)
1742 return register_pernet_subsys(&udp4_net_ops
);
1745 void udp4_proc_exit(void)
1747 unregister_pernet_subsys(&udp4_net_ops
);
1749 #endif /* CONFIG_PROC_FS */
1751 void __init
udp_table_init(struct udp_table
*table
)
1755 for (i
= 0; i
< UDP_HTABLE_SIZE
; i
++) {
1756 INIT_HLIST_NULLS_HEAD(&table
->hash
[i
].head
, i
);
1757 spin_lock_init(&table
->hash
[i
].lock
);
1761 void __init
udp_init(void)
1763 unsigned long nr_pages
, limit
;
1765 udp_table_init(&udp_table
);
1766 /* Set the pressure threshold up by the same strategy of TCP. It is a
1767 * fraction of global memory that is up to 1/2 at 256 MB, decreasing
1768 * toward zero with the amount of memory, with a floor of 128 pages.
1770 nr_pages
= totalram_pages
- totalhigh_pages
;
1771 limit
= min(nr_pages
, 1UL<<(28-PAGE_SHIFT
)) >> (20-PAGE_SHIFT
);
1772 limit
= (limit
* (nr_pages
>> (20-PAGE_SHIFT
))) >> (PAGE_SHIFT
-11);
1773 limit
= max(limit
, 128UL);
1774 sysctl_udp_mem
[0] = limit
/ 4 * 3;
1775 sysctl_udp_mem
[1] = limit
;
1776 sysctl_udp_mem
[2] = sysctl_udp_mem
[0] * 2;
1778 sysctl_udp_rmem_min
= SK_MEM_QUANTUM
;
1779 sysctl_udp_wmem_min
= SK_MEM_QUANTUM
;
1782 EXPORT_SYMBOL(udp_disconnect
);
1783 EXPORT_SYMBOL(udp_ioctl
);
1784 EXPORT_SYMBOL(udp_prot
);
1785 EXPORT_SYMBOL(udp_sendmsg
);
1786 EXPORT_SYMBOL(udp_lib_getsockopt
);
1787 EXPORT_SYMBOL(udp_lib_setsockopt
);
1788 EXPORT_SYMBOL(udp_poll
);
1789 EXPORT_SYMBOL(udp_lib_get_port
);
1791 #ifdef CONFIG_PROC_FS
1792 EXPORT_SYMBOL(udp_proc_register
);
1793 EXPORT_SYMBOL(udp_proc_unregister
);