2 * IUCV protocol stack for Linux on zSeries
4 * Copyright IBM Corp. 2006, 2009
6 * Author(s): Jennifer Hunt <jenhunt@us.ibm.com>
7 * Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
9 * Ursula Braun <ursula.braun@de.ibm.com>
12 #define KMSG_COMPONENT "af_iucv"
13 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
15 #include <linux/module.h>
16 #include <linux/types.h>
17 #include <linux/list.h>
18 #include <linux/errno.h>
19 #include <linux/kernel.h>
20 #include <linux/sched.h>
21 #include <linux/slab.h>
22 #include <linux/skbuff.h>
23 #include <linux/init.h>
24 #include <linux/poll.h>
26 #include <asm/ebcdic.h>
27 #include <asm/cpcmd.h>
28 #include <linux/kmod.h>
30 #include <net/iucv/iucv.h>
31 #include <net/iucv/af_iucv.h>
35 static char iucv_userid[80];
37 static const struct proto_ops iucv_sock_ops;
39 static struct proto iucv_proto = {
42 .obj_size = sizeof(struct iucv_sock),
45 /* special AF_IUCV IPRM messages */
46 static const u8 iprm_shutdown[8] =
47 {0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01};
49 #define TRGCLS_SIZE (sizeof(((struct iucv_message *)0)->class))
51 /* macros to set/get socket control buffer at correct offset */
52 #define CB_TAG(skb) ((skb)->cb) /* iucv message tag */
53 #define CB_TAG_LEN (sizeof(((struct iucv_message *) 0)->tag))
54 #define CB_TRGCLS(skb) ((skb)->cb + CB_TAG_LEN) /* iucv msg target class */
55 #define CB_TRGCLS_LEN (TRGCLS_SIZE)
57 #define __iucv_sock_wait(sk, condition, timeo, ret) \
59 DEFINE_WAIT(__wait); \
60 long __timeo = timeo; \
62 prepare_to_wait(sk->sk_sleep, &__wait, TASK_INTERRUPTIBLE); \
63 while (!(condition)) { \
68 if (signal_pending(current)) { \
69 ret = sock_intr_errno(__timeo); \
73 __timeo = schedule_timeout(__timeo); \
75 ret = sock_error(sk); \
79 finish_wait(sk->sk_sleep, &__wait); \
82 #define iucv_sock_wait(sk, condition, timeo) \
86 __iucv_sock_wait(sk, condition, timeo, __ret); \
90 static void iucv_sock_kill(struct sock *sk);
91 static void iucv_sock_close(struct sock *sk);
93 /* Call Back functions */
94 static void iucv_callback_rx(struct iucv_path *, struct iucv_message *);
95 static void iucv_callback_txdone(struct iucv_path *, struct iucv_message *);
96 static void iucv_callback_connack(struct iucv_path *, u8 ipuser[16]);
97 static int iucv_callback_connreq(struct iucv_path *, u8 ipvmid[8],
99 static void iucv_callback_connrej(struct iucv_path *, u8 ipuser[16]);
100 static void iucv_callback_shutdown(struct iucv_path *, u8 ipuser[16]);
102 static struct iucv_sock_list iucv_sk_list = {
103 .lock = __RW_LOCK_UNLOCKED(iucv_sk_list.lock),
104 .autobind_name = ATOMIC_INIT(0)
107 static struct iucv_handler af_iucv_handler = {
108 .path_pending = iucv_callback_connreq,
109 .path_complete = iucv_callback_connack,
110 .path_severed = iucv_callback_connrej,
111 .message_pending = iucv_callback_rx,
112 .message_complete = iucv_callback_txdone,
113 .path_quiesced = iucv_callback_shutdown,
116 static inline void high_nmcpy(unsigned char *dst, char *src)
121 static inline void low_nmcpy(unsigned char *dst, char *src)
123 memcpy(&dst[8], src, 8);
126 static int afiucv_pm_prepare(struct device *dev)
128 #ifdef CONFIG_PM_DEBUG
129 printk(KERN_WARNING "afiucv_pm_prepare\n");
134 static void afiucv_pm_complete(struct device *dev)
136 #ifdef CONFIG_PM_DEBUG
137 printk(KERN_WARNING "afiucv_pm_complete\n");
143 * afiucv_pm_freeze() - Freeze PM callback
144 * @dev: AFIUCV dummy device
146 * Sever all established IUCV communication pathes
148 static int afiucv_pm_freeze(struct device *dev)
150 struct iucv_sock *iucv;
152 struct hlist_node *node;
155 #ifdef CONFIG_PM_DEBUG
156 printk(KERN_WARNING "afiucv_pm_freeze\n");
158 read_lock(&iucv_sk_list.lock);
159 sk_for_each(sk, node, &iucv_sk_list.head) {
161 skb_queue_purge(&iucv->send_skb_q);
162 skb_queue_purge(&iucv->backlog_skb_q);
163 switch (sk->sk_state) {
169 err = iucv_path_sever(iucv->path, NULL);
170 iucv_path_free(iucv->path);
182 read_unlock(&iucv_sk_list.lock);
187 * afiucv_pm_restore_thaw() - Thaw and restore PM callback
188 * @dev: AFIUCV dummy device
190 * socket clean up after freeze
192 static int afiucv_pm_restore_thaw(struct device *dev)
194 struct iucv_sock *iucv;
196 struct hlist_node *node;
198 #ifdef CONFIG_PM_DEBUG
199 printk(KERN_WARNING "afiucv_pm_restore_thaw\n");
201 read_lock(&iucv_sk_list.lock);
202 sk_for_each(sk, node, &iucv_sk_list.head) {
204 switch (sk->sk_state) {
207 sk->sk_state = IUCV_DISCONN;
208 sk->sk_state_change(sk);
220 read_unlock(&iucv_sk_list.lock);
224 static struct dev_pm_ops afiucv_pm_ops = {
225 .prepare = afiucv_pm_prepare,
226 .complete = afiucv_pm_complete,
227 .freeze = afiucv_pm_freeze,
228 .thaw = afiucv_pm_restore_thaw,
229 .restore = afiucv_pm_restore_thaw,
232 static struct device_driver af_iucv_driver = {
233 .owner = THIS_MODULE,
236 .pm = &afiucv_pm_ops,
239 /* dummy device used as trigger for PM functions */
240 static struct device *af_iucv_dev;
243 * iucv_msg_length() - Returns the length of an iucv message.
244 * @msg: Pointer to struct iucv_message, MUST NOT be NULL
246 * The function returns the length of the specified iucv message @msg of data
247 * stored in a buffer and of data stored in the parameter list (PRMDATA).
249 * For IUCV_IPRMDATA, AF_IUCV uses the following convention to transport socket
251 * PRMDATA[0..6] socket data (max 7 bytes);
252 * PRMDATA[7] socket data length value (len is 0xff - PRMDATA[7])
254 * The socket data length is computed by substracting the socket data length
256 * If the socket data len is greater 7, then PRMDATA can be used for special
257 * notifications (see iucv_sock_shutdown); and further,
258 * if the socket data len is > 7, the function returns 8.
260 * Use this function to allocate socket buffers to store iucv message data.
262 static inline size_t iucv_msg_length(struct iucv_message *msg)
266 if (msg->flags & IUCV_IPRMDATA) {
267 datalen = 0xff - msg->rmmsg[7];
268 return (datalen < 8) ? datalen : 8;
274 * iucv_sock_in_state() - check for specific states
275 * @sk: sock structure
276 * @state: first iucv sk state
277 * @state: second iucv sk state
279 * Returns true if the socket in either in the first or second state.
281 static int iucv_sock_in_state(struct sock *sk, int state, int state2)
283 return (sk->sk_state == state || sk->sk_state == state2);
287 * iucv_below_msglim() - function to check if messages can be sent
288 * @sk: sock structure
290 * Returns true if the send queue length is lower than the message limit.
291 * Always returns true if the socket is not connected (no iucv path for
292 * checking the message limit).
294 static inline int iucv_below_msglim(struct sock *sk)
296 struct iucv_sock *iucv = iucv_sk(sk);
298 if (sk->sk_state != IUCV_CONNECTED)
300 return (skb_queue_len(&iucv->send_skb_q) < iucv->path->msglim);
304 * iucv_sock_wake_msglim() - Wake up thread waiting on msg limit
306 static void iucv_sock_wake_msglim(struct sock *sk)
308 read_lock(&sk->sk_callback_lock);
309 if (sk_has_sleeper(sk))
310 wake_up_interruptible_all(sk->sk_sleep);
311 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
312 read_unlock(&sk->sk_callback_lock);
316 static void iucv_sock_timeout(unsigned long arg)
318 struct sock *sk = (struct sock *)arg;
321 sk->sk_err = ETIMEDOUT;
322 sk->sk_state_change(sk);
329 static void iucv_sock_clear_timer(struct sock *sk)
331 sk_stop_timer(sk, &sk->sk_timer);
334 static struct sock *__iucv_get_sock_by_name(char *nm)
337 struct hlist_node *node;
339 sk_for_each(sk, node, &iucv_sk_list.head)
340 if (!memcmp(&iucv_sk(sk)->src_name, nm, 8))
346 static void iucv_sock_destruct(struct sock *sk)
348 skb_queue_purge(&sk->sk_receive_queue);
349 skb_queue_purge(&sk->sk_write_queue);
353 static void iucv_sock_cleanup_listen(struct sock *parent)
357 /* Close non-accepted connections */
358 while ((sk = iucv_accept_dequeue(parent, NULL))) {
363 parent->sk_state = IUCV_CLOSED;
366 /* Kill socket (only if zapped and orphaned) */
367 static void iucv_sock_kill(struct sock *sk)
369 if (!sock_flag(sk, SOCK_ZAPPED) || sk->sk_socket)
372 iucv_sock_unlink(&iucv_sk_list, sk);
373 sock_set_flag(sk, SOCK_DEAD);
377 /* Close an IUCV socket */
378 static void iucv_sock_close(struct sock *sk)
380 unsigned char user_data[16];
381 struct iucv_sock *iucv = iucv_sk(sk);
385 iucv_sock_clear_timer(sk);
388 switch (sk->sk_state) {
390 iucv_sock_cleanup_listen(sk);
397 sk->sk_state = IUCV_CLOSING;
398 sk->sk_state_change(sk);
400 if (!skb_queue_empty(&iucv->send_skb_q)) {
401 if (sock_flag(sk, SOCK_LINGER) && sk->sk_lingertime)
402 timeo = sk->sk_lingertime;
404 timeo = IUCV_DISCONN_TIMEOUT;
405 err = iucv_sock_wait(sk,
406 iucv_sock_in_state(sk, IUCV_CLOSED, 0),
410 case IUCV_CLOSING: /* fall through */
411 sk->sk_state = IUCV_CLOSED;
412 sk->sk_state_change(sk);
415 low_nmcpy(user_data, iucv->src_name);
416 high_nmcpy(user_data, iucv->dst_name);
417 ASCEBC(user_data, sizeof(user_data));
418 err = iucv_path_sever(iucv->path, user_data);
419 iucv_path_free(iucv->path);
423 sk->sk_err = ECONNRESET;
424 sk->sk_state_change(sk);
426 skb_queue_purge(&iucv->send_skb_q);
427 skb_queue_purge(&iucv->backlog_skb_q);
431 sock_set_flag(sk, SOCK_ZAPPED);
432 /* nothing to do here */
436 /* mark socket for deletion by iucv_sock_kill() */
437 sock_set_flag(sk, SOCK_ZAPPED);
442 static void iucv_sock_init(struct sock *sk, struct sock *parent)
445 sk->sk_type = parent->sk_type;
448 static struct sock *iucv_sock_alloc(struct socket *sock, int proto, gfp_t prio)
452 sk = sk_alloc(&init_net, PF_IUCV, prio, &iucv_proto);
456 sock_init_data(sock, sk);
457 INIT_LIST_HEAD(&iucv_sk(sk)->accept_q);
458 spin_lock_init(&iucv_sk(sk)->accept_q_lock);
459 skb_queue_head_init(&iucv_sk(sk)->send_skb_q);
460 INIT_LIST_HEAD(&iucv_sk(sk)->message_q.list);
461 spin_lock_init(&iucv_sk(sk)->message_q.lock);
462 skb_queue_head_init(&iucv_sk(sk)->backlog_skb_q);
463 iucv_sk(sk)->send_tag = 0;
464 iucv_sk(sk)->flags = 0;
465 iucv_sk(sk)->msglimit = IUCV_QUEUELEN_DEFAULT;
466 iucv_sk(sk)->path = NULL;
467 memset(&iucv_sk(sk)->src_user_id , 0, 32);
469 sk->sk_destruct = iucv_sock_destruct;
470 sk->sk_sndtimeo = IUCV_CONN_TIMEOUT;
471 sk->sk_allocation = GFP_DMA;
473 sock_reset_flag(sk, SOCK_ZAPPED);
475 sk->sk_protocol = proto;
476 sk->sk_state = IUCV_OPEN;
478 setup_timer(&sk->sk_timer, iucv_sock_timeout, (unsigned long)sk);
480 iucv_sock_link(&iucv_sk_list, sk);
484 /* Create an IUCV socket */
485 static int iucv_sock_create(struct net *net, struct socket *sock, int protocol)
489 if (protocol && protocol != PF_IUCV)
490 return -EPROTONOSUPPORT;
492 sock->state = SS_UNCONNECTED;
494 switch (sock->type) {
496 sock->ops = &iucv_sock_ops;
499 /* currently, proto ops can handle both sk types */
500 sock->ops = &iucv_sock_ops;
503 return -ESOCKTNOSUPPORT;
506 sk = iucv_sock_alloc(sock, protocol, GFP_KERNEL);
510 iucv_sock_init(sk, NULL);
515 void iucv_sock_link(struct iucv_sock_list *l, struct sock *sk)
517 write_lock_bh(&l->lock);
518 sk_add_node(sk, &l->head);
519 write_unlock_bh(&l->lock);
522 void iucv_sock_unlink(struct iucv_sock_list *l, struct sock *sk)
524 write_lock_bh(&l->lock);
525 sk_del_node_init(sk);
526 write_unlock_bh(&l->lock);
529 void iucv_accept_enqueue(struct sock *parent, struct sock *sk)
532 struct iucv_sock *par = iucv_sk(parent);
535 spin_lock_irqsave(&par->accept_q_lock, flags);
536 list_add_tail(&iucv_sk(sk)->accept_q, &par->accept_q);
537 spin_unlock_irqrestore(&par->accept_q_lock, flags);
538 iucv_sk(sk)->parent = parent;
539 parent->sk_ack_backlog++;
542 void iucv_accept_unlink(struct sock *sk)
545 struct iucv_sock *par = iucv_sk(iucv_sk(sk)->parent);
547 spin_lock_irqsave(&par->accept_q_lock, flags);
548 list_del_init(&iucv_sk(sk)->accept_q);
549 spin_unlock_irqrestore(&par->accept_q_lock, flags);
550 iucv_sk(sk)->parent->sk_ack_backlog--;
551 iucv_sk(sk)->parent = NULL;
555 struct sock *iucv_accept_dequeue(struct sock *parent, struct socket *newsock)
557 struct iucv_sock *isk, *n;
560 list_for_each_entry_safe(isk, n, &iucv_sk(parent)->accept_q, accept_q) {
561 sk = (struct sock *) isk;
564 if (sk->sk_state == IUCV_CLOSED) {
565 iucv_accept_unlink(sk);
570 if (sk->sk_state == IUCV_CONNECTED ||
571 sk->sk_state == IUCV_SEVERED ||
572 sk->sk_state == IUCV_DISCONN || /* due to PM restore */
574 iucv_accept_unlink(sk);
576 sock_graft(sk, newsock);
578 if (sk->sk_state == IUCV_SEVERED)
579 sk->sk_state = IUCV_DISCONN;
590 /* Bind an unbound socket */
591 static int iucv_sock_bind(struct socket *sock, struct sockaddr *addr,
594 struct sockaddr_iucv *sa = (struct sockaddr_iucv *) addr;
595 struct sock *sk = sock->sk;
596 struct iucv_sock *iucv;
599 /* Verify the input sockaddr */
600 if (!addr || addr->sa_family != AF_IUCV)
604 if (sk->sk_state != IUCV_OPEN) {
609 write_lock_bh(&iucv_sk_list.lock);
612 if (__iucv_get_sock_by_name(sa->siucv_name)) {
621 /* Bind the socket */
622 memcpy(iucv->src_name, sa->siucv_name, 8);
624 /* Copy the user id */
625 memcpy(iucv->src_user_id, iucv_userid, 8);
626 sk->sk_state = IUCV_BOUND;
630 /* Release the socket list lock */
631 write_unlock_bh(&iucv_sk_list.lock);
637 /* Automatically bind an unbound socket */
638 static int iucv_sock_autobind(struct sock *sk)
640 struct iucv_sock *iucv = iucv_sk(sk);
641 char query_buffer[80];
645 /* Set the userid and name */
646 cpcmd("QUERY USERID", query_buffer, sizeof(query_buffer), &err);
650 memcpy(iucv->src_user_id, query_buffer, 8);
652 write_lock_bh(&iucv_sk_list.lock);
654 sprintf(name, "%08x", atomic_inc_return(&iucv_sk_list.autobind_name));
655 while (__iucv_get_sock_by_name(name)) {
656 sprintf(name, "%08x",
657 atomic_inc_return(&iucv_sk_list.autobind_name));
660 write_unlock_bh(&iucv_sk_list.lock);
662 memcpy(&iucv->src_name, name, 8);
667 /* Connect an unconnected socket */
668 static int iucv_sock_connect(struct socket *sock, struct sockaddr *addr,
671 struct sockaddr_iucv *sa = (struct sockaddr_iucv *) addr;
672 struct sock *sk = sock->sk;
673 struct iucv_sock *iucv;
674 unsigned char user_data[16];
677 if (addr->sa_family != AF_IUCV || alen < sizeof(struct sockaddr_iucv))
680 if (sk->sk_state != IUCV_OPEN && sk->sk_state != IUCV_BOUND)
683 if (sk->sk_type != SOCK_STREAM && sk->sk_type != SOCK_SEQPACKET)
686 if (sk->sk_state == IUCV_OPEN) {
687 err = iucv_sock_autobind(sk);
694 /* Set the destination information */
695 memcpy(iucv_sk(sk)->dst_user_id, sa->siucv_user_id, 8);
696 memcpy(iucv_sk(sk)->dst_name, sa->siucv_name, 8);
698 high_nmcpy(user_data, sa->siucv_name);
699 low_nmcpy(user_data, iucv_sk(sk)->src_name);
700 ASCEBC(user_data, sizeof(user_data));
704 iucv->path = iucv_path_alloc(iucv->msglimit,
705 IUCV_IPRMDATA, GFP_KERNEL);
710 err = iucv_path_connect(iucv->path, &af_iucv_handler,
711 sa->siucv_user_id, NULL, user_data, sk);
713 iucv_path_free(iucv->path);
716 case 0x0b: /* Target communicator is not logged on */
719 case 0x0d: /* Max connections for this guest exceeded */
720 case 0x0e: /* Max connections for target guest exceeded */
723 case 0x0f: /* Missing IUCV authorization */
733 if (sk->sk_state != IUCV_CONNECTED) {
734 err = iucv_sock_wait(sk, iucv_sock_in_state(sk, IUCV_CONNECTED,
736 sock_sndtimeo(sk, flags & O_NONBLOCK));
739 if (sk->sk_state == IUCV_DISCONN) {
744 iucv_path_sever(iucv->path, NULL);
745 iucv_path_free(iucv->path);
754 /* Move a socket into listening state. */
755 static int iucv_sock_listen(struct socket *sock, int backlog)
757 struct sock *sk = sock->sk;
763 if (sk->sk_state != IUCV_BOUND)
766 if (sock->type != SOCK_STREAM && sock->type != SOCK_SEQPACKET)
769 sk->sk_max_ack_backlog = backlog;
770 sk->sk_ack_backlog = 0;
771 sk->sk_state = IUCV_LISTEN;
779 /* Accept a pending connection */
780 static int iucv_sock_accept(struct socket *sock, struct socket *newsock,
783 DECLARE_WAITQUEUE(wait, current);
784 struct sock *sk = sock->sk, *nsk;
788 lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
790 if (sk->sk_state != IUCV_LISTEN) {
795 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
797 /* Wait for an incoming connection */
798 add_wait_queue_exclusive(sk->sk_sleep, &wait);
799 while (!(nsk = iucv_accept_dequeue(sk, newsock))) {
800 set_current_state(TASK_INTERRUPTIBLE);
807 timeo = schedule_timeout(timeo);
808 lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
810 if (sk->sk_state != IUCV_LISTEN) {
815 if (signal_pending(current)) {
816 err = sock_intr_errno(timeo);
821 set_current_state(TASK_RUNNING);
822 remove_wait_queue(sk->sk_sleep, &wait);
827 newsock->state = SS_CONNECTED;
834 static int iucv_sock_getname(struct socket *sock, struct sockaddr *addr,
837 struct sockaddr_iucv *siucv = (struct sockaddr_iucv *) addr;
838 struct sock *sk = sock->sk;
840 addr->sa_family = AF_IUCV;
841 *len = sizeof(struct sockaddr_iucv);
844 memcpy(siucv->siucv_user_id, iucv_sk(sk)->dst_user_id, 8);
845 memcpy(siucv->siucv_name, &iucv_sk(sk)->dst_name, 8);
847 memcpy(siucv->siucv_user_id, iucv_sk(sk)->src_user_id, 8);
848 memcpy(siucv->siucv_name, iucv_sk(sk)->src_name, 8);
850 memset(&siucv->siucv_port, 0, sizeof(siucv->siucv_port));
851 memset(&siucv->siucv_addr, 0, sizeof(siucv->siucv_addr));
852 memset(siucv->siucv_nodeid, 0, sizeof(siucv->siucv_nodeid));
858 * iucv_send_iprm() - Send socket data in parameter list of an iucv message.
860 * @msg: Pointer to a struct iucv_message
861 * @skb: The socket data to send, skb->len MUST BE <= 7
863 * Send the socket data in the parameter list in the iucv message
864 * (IUCV_IPRMDATA). The socket data is stored at index 0 to 6 in the parameter
865 * list and the socket data len at index 7 (last byte).
866 * See also iucv_msg_length().
868 * Returns the error code from the iucv_message_send() call.
870 static int iucv_send_iprm(struct iucv_path *path, struct iucv_message *msg,
875 memcpy(prmdata, (void *) skb->data, skb->len);
876 prmdata[7] = 0xff - (u8) skb->len;
877 return iucv_message_send(path, msg, IUCV_IPRMDATA, 0,
878 (void *) prmdata, 8);
881 static int iucv_sock_sendmsg(struct kiocb *iocb, struct socket *sock,
882 struct msghdr *msg, size_t len)
884 struct sock *sk = sock->sk;
885 struct iucv_sock *iucv = iucv_sk(sk);
887 struct iucv_message txmsg;
888 struct cmsghdr *cmsg;
894 int noblock = msg->msg_flags & MSG_DONTWAIT;
896 err = sock_error(sk);
900 if (msg->msg_flags & MSG_OOB)
903 /* SOCK_SEQPACKET: we do not support segmented records */
904 if (sk->sk_type == SOCK_SEQPACKET && !(msg->msg_flags & MSG_EOR))
909 if (sk->sk_shutdown & SEND_SHUTDOWN) {
914 /* Return if the socket is not in connected state */
915 if (sk->sk_state != IUCV_CONNECTED) {
920 /* initialize defaults */
921 cmsg_done = 0; /* check for duplicate headers */
924 /* iterate over control messages */
925 for (cmsg = CMSG_FIRSTHDR(msg); cmsg;
926 cmsg = CMSG_NXTHDR(msg, cmsg)) {
928 if (!CMSG_OK(msg, cmsg)) {
933 if (cmsg->cmsg_level != SOL_IUCV)
936 if (cmsg->cmsg_type & cmsg_done) {
940 cmsg_done |= cmsg->cmsg_type;
942 switch (cmsg->cmsg_type) {
943 case SCM_IUCV_TRGCLS:
944 if (cmsg->cmsg_len != CMSG_LEN(TRGCLS_SIZE)) {
949 /* set iucv message target class */
951 (void *) CMSG_DATA(cmsg), TRGCLS_SIZE);
962 /* allocate one skb for each iucv message:
963 * this is fine for SOCK_SEQPACKET (unless we want to support
964 * segmented records using the MSG_EOR flag), but
965 * for SOCK_STREAM we might want to improve it in future */
966 skb = sock_alloc_send_skb(sk, len, noblock, &err);
969 if (memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len)) {
974 /* wait if outstanding messages for iucv path has reached */
975 timeo = sock_sndtimeo(sk, noblock);
976 err = iucv_sock_wait(sk, iucv_below_msglim(sk), timeo);
980 /* return -ECONNRESET if the socket is no longer connected */
981 if (sk->sk_state != IUCV_CONNECTED) {
986 /* increment and save iucv message tag for msg_completion cbk */
987 txmsg.tag = iucv->send_tag++;
988 memcpy(CB_TAG(skb), &txmsg.tag, CB_TAG_LEN);
989 skb_queue_tail(&iucv->send_skb_q, skb);
991 if (((iucv->path->flags & IUCV_IPRMDATA) & iucv->flags)
993 err = iucv_send_iprm(iucv->path, &txmsg, skb);
995 /* on success: there is no message_complete callback
996 * for an IPRMDATA msg; remove skb from send queue */
998 skb_unlink(skb, &iucv->send_skb_q);
1002 /* this error should never happen since the
1003 * IUCV_IPRMDATA path flag is set... sever path */
1005 iucv_path_sever(iucv->path, NULL);
1006 skb_unlink(skb, &iucv->send_skb_q);
1011 err = iucv_message_send(iucv->path, &txmsg, 0, 0,
1012 (void *) skb->data, skb->len);
1016 memcpy(user_id, iucv->dst_user_id, 8);
1018 memcpy(appl_id, iucv->dst_name, 8);
1019 pr_err("Application %s on z/VM guest %s"
1020 " exceeds message limit\n",
1025 skb_unlink(skb, &iucv->send_skb_q);
1039 static int iucv_fragment_skb(struct sock *sk, struct sk_buff *skb, int len)
1041 int dataleft, size, copied = 0;
1042 struct sk_buff *nskb;
1046 if (dataleft >= sk->sk_rcvbuf / 4)
1047 size = sk->sk_rcvbuf / 4;
1051 nskb = alloc_skb(size, GFP_ATOMIC | GFP_DMA);
1055 /* copy target class to control buffer of new skb */
1056 memcpy(CB_TRGCLS(nskb), CB_TRGCLS(skb), CB_TRGCLS_LEN);
1058 /* copy data fragment */
1059 memcpy(nskb->data, skb->data + copied, size);
1063 skb_reset_transport_header(nskb);
1064 skb_reset_network_header(nskb);
1067 skb_queue_tail(&iucv_sk(sk)->backlog_skb_q, nskb);
1073 static void iucv_process_message(struct sock *sk, struct sk_buff *skb,
1074 struct iucv_path *path,
1075 struct iucv_message *msg)
1080 len = iucv_msg_length(msg);
1082 /* store msg target class in the second 4 bytes of skb ctrl buffer */
1083 /* Note: the first 4 bytes are reserved for msg tag */
1084 memcpy(CB_TRGCLS(skb), &msg->class, CB_TRGCLS_LEN);
1086 /* check for special IPRM messages (e.g. iucv_sock_shutdown) */
1087 if ((msg->flags & IUCV_IPRMDATA) && len > 7) {
1088 if (memcmp(msg->rmmsg, iprm_shutdown, 8) == 0) {
1093 rc = iucv_message_receive(path, msg, msg->flags & IUCV_IPRMDATA,
1094 skb->data, len, NULL);
1099 /* we need to fragment iucv messages for SOCK_STREAM only;
1100 * for SOCK_SEQPACKET, it is only relevant if we support
1101 * record segmentation using MSG_EOR (see also recvmsg()) */
1102 if (sk->sk_type == SOCK_STREAM &&
1103 skb->truesize >= sk->sk_rcvbuf / 4) {
1104 rc = iucv_fragment_skb(sk, skb, len);
1108 iucv_path_sever(path, NULL);
1111 skb = skb_dequeue(&iucv_sk(sk)->backlog_skb_q);
1113 skb_reset_transport_header(skb);
1114 skb_reset_network_header(skb);
1119 if (sock_queue_rcv_skb(sk, skb))
1120 skb_queue_head(&iucv_sk(sk)->backlog_skb_q, skb);
1123 static void iucv_process_message_q(struct sock *sk)
1125 struct iucv_sock *iucv = iucv_sk(sk);
1126 struct sk_buff *skb;
1127 struct sock_msg_q *p, *n;
1129 list_for_each_entry_safe(p, n, &iucv->message_q.list, list) {
1130 skb = alloc_skb(iucv_msg_length(&p->msg), GFP_ATOMIC | GFP_DMA);
1133 iucv_process_message(sk, skb, p->path, &p->msg);
1136 if (!skb_queue_empty(&iucv->backlog_skb_q))
1141 static int iucv_sock_recvmsg(struct kiocb *iocb, struct socket *sock,
1142 struct msghdr *msg, size_t len, int flags)
1144 int noblock = flags & MSG_DONTWAIT;
1145 struct sock *sk = sock->sk;
1146 struct iucv_sock *iucv = iucv_sk(sk);
1147 unsigned int copied, rlen;
1148 struct sk_buff *skb, *rskb, *cskb;
1151 if ((sk->sk_state == IUCV_DISCONN || sk->sk_state == IUCV_SEVERED) &&
1152 skb_queue_empty(&iucv->backlog_skb_q) &&
1153 skb_queue_empty(&sk->sk_receive_queue) &&
1154 list_empty(&iucv->message_q.list))
1157 if (flags & (MSG_OOB))
1160 /* receive/dequeue next skb:
1161 * the function understands MSG_PEEK and, thus, does not dequeue skb */
1162 skb = skb_recv_datagram(sk, flags, noblock, &err);
1164 if (sk->sk_shutdown & RCV_SHUTDOWN)
1169 rlen = skb->len; /* real length of skb */
1170 copied = min_t(unsigned int, rlen, len);
1173 if (memcpy_toiovec(msg->msg_iov, cskb->data, copied)) {
1174 if (!(flags & MSG_PEEK))
1175 skb_queue_head(&sk->sk_receive_queue, skb);
1179 /* SOCK_SEQPACKET: set MSG_TRUNC if recv buf size is too small */
1180 if (sk->sk_type == SOCK_SEQPACKET) {
1182 msg->msg_flags |= MSG_TRUNC;
1183 /* each iucv message contains a complete record */
1184 msg->msg_flags |= MSG_EOR;
1187 /* create control message to store iucv msg target class:
1188 * get the trgcls from the control buffer of the skb due to
1189 * fragmentation of original iucv message. */
1190 err = put_cmsg(msg, SOL_IUCV, SCM_IUCV_TRGCLS,
1191 CB_TRGCLS_LEN, CB_TRGCLS(skb));
1193 if (!(flags & MSG_PEEK))
1194 skb_queue_head(&sk->sk_receive_queue, skb);
1198 /* Mark read part of skb as used */
1199 if (!(flags & MSG_PEEK)) {
1201 /* SOCK_STREAM: re-queue skb if it contains unreceived data */
1202 if (sk->sk_type == SOCK_STREAM) {
1203 skb_pull(skb, copied);
1205 skb_queue_head(&sk->sk_receive_queue, skb);
1212 /* Queue backlog skbs */
1213 rskb = skb_dequeue(&iucv->backlog_skb_q);
1215 if (sock_queue_rcv_skb(sk, rskb)) {
1216 skb_queue_head(&iucv->backlog_skb_q,
1220 rskb = skb_dequeue(&iucv->backlog_skb_q);
1223 if (skb_queue_empty(&iucv->backlog_skb_q)) {
1224 spin_lock_bh(&iucv->message_q.lock);
1225 if (!list_empty(&iucv->message_q.list))
1226 iucv_process_message_q(sk);
1227 spin_unlock_bh(&iucv->message_q.lock);
1232 /* SOCK_SEQPACKET: return real length if MSG_TRUNC is set */
1233 if (sk->sk_type == SOCK_SEQPACKET && (flags & MSG_TRUNC))
1239 static inline unsigned int iucv_accept_poll(struct sock *parent)
1241 struct iucv_sock *isk, *n;
1244 list_for_each_entry_safe(isk, n, &iucv_sk(parent)->accept_q, accept_q) {
1245 sk = (struct sock *) isk;
1247 if (sk->sk_state == IUCV_CONNECTED)
1248 return POLLIN | POLLRDNORM;
1254 unsigned int iucv_sock_poll(struct file *file, struct socket *sock,
1257 struct sock *sk = sock->sk;
1258 unsigned int mask = 0;
1260 sock_poll_wait(file, sk->sk_sleep, wait);
1262 if (sk->sk_state == IUCV_LISTEN)
1263 return iucv_accept_poll(sk);
1265 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
1268 if (sk->sk_shutdown & RCV_SHUTDOWN)
1271 if (sk->sk_shutdown == SHUTDOWN_MASK)
1274 if (!skb_queue_empty(&sk->sk_receive_queue) ||
1275 (sk->sk_shutdown & RCV_SHUTDOWN))
1276 mask |= POLLIN | POLLRDNORM;
1278 if (sk->sk_state == IUCV_CLOSED)
1281 if (sk->sk_state == IUCV_DISCONN || sk->sk_state == IUCV_SEVERED)
1284 if (sock_writeable(sk))
1285 mask |= POLLOUT | POLLWRNORM | POLLWRBAND;
1287 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1292 static int iucv_sock_shutdown(struct socket *sock, int how)
1294 struct sock *sk = sock->sk;
1295 struct iucv_sock *iucv = iucv_sk(sk);
1296 struct iucv_message txmsg;
1301 if ((how & ~SHUTDOWN_MASK) || !how)
1305 switch (sk->sk_state) {
1314 sk->sk_shutdown |= how;
1318 if (how == SEND_SHUTDOWN || how == SHUTDOWN_MASK) {
1321 err = iucv_message_send(iucv->path, &txmsg, IUCV_IPRMDATA, 0,
1322 (void *) iprm_shutdown, 8);
1338 if (how == RCV_SHUTDOWN || how == SHUTDOWN_MASK) {
1339 err = iucv_path_quiesce(iucv_sk(sk)->path, NULL);
1343 skb_queue_purge(&sk->sk_receive_queue);
1346 /* Wake up anyone sleeping in poll */
1347 sk->sk_state_change(sk);
1354 static int iucv_sock_release(struct socket *sock)
1356 struct sock *sk = sock->sk;
1362 iucv_sock_close(sk);
1364 /* Unregister with IUCV base support */
1365 if (iucv_sk(sk)->path) {
1366 iucv_path_sever(iucv_sk(sk)->path, NULL);
1367 iucv_path_free(iucv_sk(sk)->path);
1368 iucv_sk(sk)->path = NULL;
1376 /* getsockopt and setsockopt */
1377 static int iucv_sock_setsockopt(struct socket *sock, int level, int optname,
1378 char __user *optval, int optlen)
1380 struct sock *sk = sock->sk;
1381 struct iucv_sock *iucv = iucv_sk(sk);
1385 if (level != SOL_IUCV)
1386 return -ENOPROTOOPT;
1388 if (optlen < sizeof(int))
1391 if (get_user(val, (int __user *) optval))
1398 case SO_IPRMDATA_MSG:
1400 iucv->flags |= IUCV_IPRMDATA;
1402 iucv->flags &= ~IUCV_IPRMDATA;
1405 switch (sk->sk_state) {
1408 if (val < 1 || val > (u16)(~0))
1411 iucv->msglimit = val;
1427 static int iucv_sock_getsockopt(struct socket *sock, int level, int optname,
1428 char __user *optval, int __user *optlen)
1430 struct sock *sk = sock->sk;
1431 struct iucv_sock *iucv = iucv_sk(sk);
1434 if (level != SOL_IUCV)
1435 return -ENOPROTOOPT;
1437 if (get_user(len, optlen))
1443 len = min_t(unsigned int, len, sizeof(int));
1446 case SO_IPRMDATA_MSG:
1447 val = (iucv->flags & IUCV_IPRMDATA) ? 1 : 0;
1451 val = (iucv->path != NULL) ? iucv->path->msglim /* connected */
1452 : iucv->msglimit; /* default */
1456 return -ENOPROTOOPT;
1459 if (put_user(len, optlen))
1461 if (copy_to_user(optval, &val, len))
1468 /* Callback wrappers - called from iucv base support */
1469 static int iucv_callback_connreq(struct iucv_path *path,
1470 u8 ipvmid[8], u8 ipuser[16])
1472 unsigned char user_data[16];
1473 unsigned char nuser_data[16];
1474 unsigned char src_name[8];
1475 struct hlist_node *node;
1476 struct sock *sk, *nsk;
1477 struct iucv_sock *iucv, *niucv;
1480 memcpy(src_name, ipuser, 8);
1481 EBCASC(src_name, 8);
1482 /* Find out if this path belongs to af_iucv. */
1483 read_lock(&iucv_sk_list.lock);
1486 sk_for_each(sk, node, &iucv_sk_list.head)
1487 if (sk->sk_state == IUCV_LISTEN &&
1488 !memcmp(&iucv_sk(sk)->src_name, src_name, 8)) {
1490 * Found a listening socket with
1491 * src_name == ipuser[0-7].
1496 read_unlock(&iucv_sk_list.lock);
1498 /* No socket found, not one of our paths. */
1503 /* Check if parent socket is listening */
1504 low_nmcpy(user_data, iucv->src_name);
1505 high_nmcpy(user_data, iucv->dst_name);
1506 ASCEBC(user_data, sizeof(user_data));
1507 if (sk->sk_state != IUCV_LISTEN) {
1508 err = iucv_path_sever(path, user_data);
1509 iucv_path_free(path);
1513 /* Check for backlog size */
1514 if (sk_acceptq_is_full(sk)) {
1515 err = iucv_path_sever(path, user_data);
1516 iucv_path_free(path);
1520 /* Create the new socket */
1521 nsk = iucv_sock_alloc(NULL, sk->sk_type, GFP_ATOMIC);
1523 err = iucv_path_sever(path, user_data);
1524 iucv_path_free(path);
1528 niucv = iucv_sk(nsk);
1529 iucv_sock_init(nsk, sk);
1531 /* Set the new iucv_sock */
1532 memcpy(niucv->dst_name, ipuser + 8, 8);
1533 EBCASC(niucv->dst_name, 8);
1534 memcpy(niucv->dst_user_id, ipvmid, 8);
1535 memcpy(niucv->src_name, iucv->src_name, 8);
1536 memcpy(niucv->src_user_id, iucv->src_user_id, 8);
1539 /* Call iucv_accept */
1540 high_nmcpy(nuser_data, ipuser + 8);
1541 memcpy(nuser_data + 8, niucv->src_name, 8);
1542 ASCEBC(nuser_data + 8, 8);
1544 /* set message limit for path based on msglimit of accepting socket */
1545 niucv->msglimit = iucv->msglimit;
1546 path->msglim = iucv->msglimit;
1547 err = iucv_path_accept(path, &af_iucv_handler, nuser_data, nsk);
1549 err = iucv_path_sever(path, user_data);
1550 iucv_path_free(path);
1551 iucv_sock_kill(nsk);
1555 iucv_accept_enqueue(sk, nsk);
1557 /* Wake up accept */
1558 nsk->sk_state = IUCV_CONNECTED;
1559 sk->sk_data_ready(sk, 1);
1566 static void iucv_callback_connack(struct iucv_path *path, u8 ipuser[16])
1568 struct sock *sk = path->private;
1570 sk->sk_state = IUCV_CONNECTED;
1571 sk->sk_state_change(sk);
1574 static void iucv_callback_rx(struct iucv_path *path, struct iucv_message *msg)
1576 struct sock *sk = path->private;
1577 struct iucv_sock *iucv = iucv_sk(sk);
1578 struct sk_buff *skb;
1579 struct sock_msg_q *save_msg;
1582 if (sk->sk_shutdown & RCV_SHUTDOWN) {
1583 iucv_message_reject(path, msg);
1587 spin_lock(&iucv->message_q.lock);
1589 if (!list_empty(&iucv->message_q.list) ||
1590 !skb_queue_empty(&iucv->backlog_skb_q))
1593 len = atomic_read(&sk->sk_rmem_alloc);
1594 len += iucv_msg_length(msg) + sizeof(struct sk_buff);
1595 if (len > sk->sk_rcvbuf)
1598 skb = alloc_skb(iucv_msg_length(msg), GFP_ATOMIC | GFP_DMA);
1602 iucv_process_message(sk, skb, path, msg);
1606 save_msg = kzalloc(sizeof(struct sock_msg_q), GFP_ATOMIC | GFP_DMA);
1609 save_msg->path = path;
1610 save_msg->msg = *msg;
1612 list_add_tail(&save_msg->list, &iucv->message_q.list);
1615 spin_unlock(&iucv->message_q.lock);
1618 static void iucv_callback_txdone(struct iucv_path *path,
1619 struct iucv_message *msg)
1621 struct sock *sk = path->private;
1622 struct sk_buff *this = NULL;
1623 struct sk_buff_head *list = &iucv_sk(sk)->send_skb_q;
1624 struct sk_buff *list_skb = list->next;
1625 unsigned long flags;
1627 if (!skb_queue_empty(list)) {
1628 spin_lock_irqsave(&list->lock, flags);
1630 while (list_skb != (struct sk_buff *)list) {
1631 if (!memcmp(&msg->tag, CB_TAG(list_skb), CB_TAG_LEN)) {
1635 list_skb = list_skb->next;
1638 __skb_unlink(this, list);
1640 spin_unlock_irqrestore(&list->lock, flags);
1644 /* wake up any process waiting for sending */
1645 iucv_sock_wake_msglim(sk);
1650 if (sk->sk_state == IUCV_CLOSING) {
1651 if (skb_queue_empty(&iucv_sk(sk)->send_skb_q)) {
1652 sk->sk_state = IUCV_CLOSED;
1653 sk->sk_state_change(sk);
1659 static void iucv_callback_connrej(struct iucv_path *path, u8 ipuser[16])
1661 struct sock *sk = path->private;
1663 if (!list_empty(&iucv_sk(sk)->accept_q))
1664 sk->sk_state = IUCV_SEVERED;
1666 sk->sk_state = IUCV_DISCONN;
1668 sk->sk_state_change(sk);
1671 /* called if the other communication side shuts down its RECV direction;
1672 * in turn, the callback sets SEND_SHUTDOWN to disable sending of data.
1674 static void iucv_callback_shutdown(struct iucv_path *path, u8 ipuser[16])
1676 struct sock *sk = path->private;
1679 if (sk->sk_state != IUCV_CLOSED) {
1680 sk->sk_shutdown |= SEND_SHUTDOWN;
1681 sk->sk_state_change(sk);
1686 static const struct proto_ops iucv_sock_ops = {
1688 .owner = THIS_MODULE,
1689 .release = iucv_sock_release,
1690 .bind = iucv_sock_bind,
1691 .connect = iucv_sock_connect,
1692 .listen = iucv_sock_listen,
1693 .accept = iucv_sock_accept,
1694 .getname = iucv_sock_getname,
1695 .sendmsg = iucv_sock_sendmsg,
1696 .recvmsg = iucv_sock_recvmsg,
1697 .poll = iucv_sock_poll,
1698 .ioctl = sock_no_ioctl,
1699 .mmap = sock_no_mmap,
1700 .socketpair = sock_no_socketpair,
1701 .shutdown = iucv_sock_shutdown,
1702 .setsockopt = iucv_sock_setsockopt,
1703 .getsockopt = iucv_sock_getsockopt,
1706 static struct net_proto_family iucv_sock_family_ops = {
1708 .owner = THIS_MODULE,
1709 .create = iucv_sock_create,
1712 static int __init afiucv_init(void)
1716 if (!MACHINE_IS_VM) {
1717 pr_err("The af_iucv module cannot be loaded"
1719 err = -EPROTONOSUPPORT;
1722 cpcmd("QUERY USERID", iucv_userid, sizeof(iucv_userid), &err);
1723 if (unlikely(err)) {
1725 err = -EPROTONOSUPPORT;
1729 err = iucv_register(&af_iucv_handler, 0);
1732 err = proto_register(&iucv_proto, 0);
1735 err = sock_register(&iucv_sock_family_ops);
1738 /* establish dummy device */
1739 err = driver_register(&af_iucv_driver);
1742 af_iucv_dev = kzalloc(sizeof(struct device), GFP_KERNEL);
1747 dev_set_name(af_iucv_dev, "af_iucv");
1748 af_iucv_dev->bus = &iucv_bus;
1749 af_iucv_dev->parent = iucv_root;
1750 af_iucv_dev->release = (void (*)(struct device *))kfree;
1751 af_iucv_dev->driver = &af_iucv_driver;
1752 err = device_register(af_iucv_dev);
1759 driver_unregister(&af_iucv_driver);
1761 sock_unregister(PF_IUCV);
1763 proto_unregister(&iucv_proto);
1765 iucv_unregister(&af_iucv_handler, 0);
1770 static void __exit afiucv_exit(void)
1772 device_unregister(af_iucv_dev);
1773 driver_unregister(&af_iucv_driver);
1774 sock_unregister(PF_IUCV);
1775 proto_unregister(&iucv_proto);
1776 iucv_unregister(&af_iucv_handler, 0);
1779 module_init(afiucv_init);
1780 module_exit(afiucv_exit);
1782 MODULE_AUTHOR("Jennifer Hunt <jenhunt@us.ibm.com>");
1783 MODULE_DESCRIPTION("IUCV Sockets ver " VERSION);
1784 MODULE_VERSION(VERSION);
1785 MODULE_LICENSE("GPL");
1786 MODULE_ALIAS_NETPROTO(PF_IUCV);