xref: /linux/net/iucv/af_iucv.c (revision bf80eef2212a1e8451df13b52533f4bc31bb4f8e)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  *  IUCV protocol stack for Linux on zSeries
4  *
5  *  Copyright IBM Corp. 2006, 2009
6  *
7  *  Author(s):	Jennifer Hunt <jenhunt@us.ibm.com>
8  *		Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
9  *  PM functions:
10  *		Ursula Braun <ursula.braun@de.ibm.com>
11  */
12 
13 #define KMSG_COMPONENT "af_iucv"
14 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
15 
16 #include <linux/filter.h>
17 #include <linux/module.h>
18 #include <linux/netdevice.h>
19 #include <linux/types.h>
20 #include <linux/limits.h>
21 #include <linux/list.h>
22 #include <linux/errno.h>
23 #include <linux/kernel.h>
24 #include <linux/sched/signal.h>
25 #include <linux/slab.h>
26 #include <linux/skbuff.h>
27 #include <linux/init.h>
28 #include <linux/poll.h>
29 #include <linux/security.h>
30 #include <net/sock.h>
31 #include <asm/ebcdic.h>
32 #include <asm/cpcmd.h>
33 #include <linux/kmod.h>
34 
35 #include <net/iucv/af_iucv.h>
36 
37 #define VERSION "1.2"
38 
39 static char iucv_userid[80];
40 
41 static struct proto iucv_proto = {
42 	.name		= "AF_IUCV",
43 	.owner		= THIS_MODULE,
44 	.obj_size	= sizeof(struct iucv_sock),
45 };
46 
47 static struct iucv_interface *pr_iucv;
48 static struct iucv_handler af_iucv_handler;
49 
50 /* special AF_IUCV IPRM messages */
51 static const u8 iprm_shutdown[8] =
52 	{0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01};
53 
54 #define TRGCLS_SIZE	sizeof_field(struct iucv_message, class)
55 
56 #define __iucv_sock_wait(sk, condition, timeo, ret)			\
57 do {									\
58 	DEFINE_WAIT(__wait);						\
59 	long __timeo = timeo;						\
60 	ret = 0;							\
61 	prepare_to_wait(sk_sleep(sk), &__wait, TASK_INTERRUPTIBLE);	\
62 	while (!(condition)) {						\
63 		if (!__timeo) {						\
64 			ret = -EAGAIN;					\
65 			break;						\
66 		}							\
67 		if (signal_pending(current)) {				\
68 			ret = sock_intr_errno(__timeo);			\
69 			break;						\
70 		}							\
71 		release_sock(sk);					\
72 		__timeo = schedule_timeout(__timeo);			\
73 		lock_sock(sk);						\
74 		ret = sock_error(sk);					\
75 		if (ret)						\
76 			break;						\
77 	}								\
78 	finish_wait(sk_sleep(sk), &__wait);				\
79 } while (0)
80 
81 #define iucv_sock_wait(sk, condition, timeo)				\
82 ({									\
83 	int __ret = 0;							\
84 	if (!(condition))						\
85 		__iucv_sock_wait(sk, condition, timeo, __ret);		\
86 	__ret;								\
87 })
88 
89 static struct sock *iucv_accept_dequeue(struct sock *parent,
90 					struct socket *newsock);
91 static void iucv_sock_kill(struct sock *sk);
92 static void iucv_sock_close(struct sock *sk);
93 
94 static void afiucv_hs_callback_txnotify(struct sock *sk, enum iucv_tx_notify);
95 
96 static struct iucv_sock_list iucv_sk_list = {
97 	.lock = __RW_LOCK_UNLOCKED(iucv_sk_list.lock),
98 	.autobind_name = ATOMIC_INIT(0)
99 };
100 
101 static inline void high_nmcpy(unsigned char *dst, char *src)
102 {
103        memcpy(dst, src, 8);
104 }
105 
106 static inline void low_nmcpy(unsigned char *dst, char *src)
107 {
108        memcpy(&dst[8], src, 8);
109 }
110 
111 /**
112  * iucv_msg_length() - Returns the length of an iucv message.
113  * @msg:	Pointer to struct iucv_message, MUST NOT be NULL
114  *
115  * The function returns the length of the specified iucv message @msg of data
116  * stored in a buffer and of data stored in the parameter list (PRMDATA).
117  *
118  * For IUCV_IPRMDATA, AF_IUCV uses the following convention to transport socket
119  * data:
120  *	PRMDATA[0..6]	socket data (max 7 bytes);
121  *	PRMDATA[7]	socket data length value (len is 0xff - PRMDATA[7])
122  *
123  * The socket data length is computed by subtracting the socket data length
124  * value from 0xFF.
125  * If the socket data len is greater 7, then PRMDATA can be used for special
126  * notifications (see iucv_sock_shutdown); and further,
127  * if the socket data len is > 7, the function returns 8.
128  *
129  * Use this function to allocate socket buffers to store iucv message data.
130  */
131 static inline size_t iucv_msg_length(struct iucv_message *msg)
132 {
133 	size_t datalen;
134 
135 	if (msg->flags & IUCV_IPRMDATA) {
136 		datalen = 0xff - msg->rmmsg[7];
137 		return (datalen < 8) ? datalen : 8;
138 	}
139 	return msg->length;
140 }
141 
142 /**
143  * iucv_sock_in_state() - check for specific states
144  * @sk:		sock structure
145  * @state:	first iucv sk state
146  * @state2:	second iucv sk state
147  *
148  * Returns true if the socket in either in the first or second state.
149  */
150 static int iucv_sock_in_state(struct sock *sk, int state, int state2)
151 {
152 	return (sk->sk_state == state || sk->sk_state == state2);
153 }
154 
155 /**
156  * iucv_below_msglim() - function to check if messages can be sent
157  * @sk:		sock structure
158  *
159  * Returns true if the send queue length is lower than the message limit.
160  * Always returns true if the socket is not connected (no iucv path for
161  * checking the message limit).
162  */
163 static inline int iucv_below_msglim(struct sock *sk)
164 {
165 	struct iucv_sock *iucv = iucv_sk(sk);
166 
167 	if (sk->sk_state != IUCV_CONNECTED)
168 		return 1;
169 	if (iucv->transport == AF_IUCV_TRANS_IUCV)
170 		return (atomic_read(&iucv->skbs_in_xmit) < iucv->path->msglim);
171 	else
172 		return ((atomic_read(&iucv->msg_sent) < iucv->msglimit_peer) &&
173 			(atomic_read(&iucv->pendings) <= 0));
174 }
175 
176 /*
177  * iucv_sock_wake_msglim() - Wake up thread waiting on msg limit
178  */
179 static void iucv_sock_wake_msglim(struct sock *sk)
180 {
181 	struct socket_wq *wq;
182 
183 	rcu_read_lock();
184 	wq = rcu_dereference(sk->sk_wq);
185 	if (skwq_has_sleeper(wq))
186 		wake_up_interruptible_all(&wq->wait);
187 	sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
188 	rcu_read_unlock();
189 }
190 
191 /*
192  * afiucv_hs_send() - send a message through HiperSockets transport
193  */
194 static int afiucv_hs_send(struct iucv_message *imsg, struct sock *sock,
195 		   struct sk_buff *skb, u8 flags)
196 {
197 	struct iucv_sock *iucv = iucv_sk(sock);
198 	struct af_iucv_trans_hdr *phs_hdr;
199 	int err, confirm_recv = 0;
200 
201 	phs_hdr = skb_push(skb, sizeof(*phs_hdr));
202 	memset(phs_hdr, 0, sizeof(*phs_hdr));
203 	skb_reset_network_header(skb);
204 
205 	phs_hdr->magic = ETH_P_AF_IUCV;
206 	phs_hdr->version = 1;
207 	phs_hdr->flags = flags;
208 	if (flags == AF_IUCV_FLAG_SYN)
209 		phs_hdr->window = iucv->msglimit;
210 	else if ((flags == AF_IUCV_FLAG_WIN) || !flags) {
211 		confirm_recv = atomic_read(&iucv->msg_recv);
212 		phs_hdr->window = confirm_recv;
213 		if (confirm_recv)
214 			phs_hdr->flags = phs_hdr->flags | AF_IUCV_FLAG_WIN;
215 	}
216 	memcpy(phs_hdr->destUserID, iucv->dst_user_id, 8);
217 	memcpy(phs_hdr->destAppName, iucv->dst_name, 8);
218 	memcpy(phs_hdr->srcUserID, iucv->src_user_id, 8);
219 	memcpy(phs_hdr->srcAppName, iucv->src_name, 8);
220 	ASCEBC(phs_hdr->destUserID, sizeof(phs_hdr->destUserID));
221 	ASCEBC(phs_hdr->destAppName, sizeof(phs_hdr->destAppName));
222 	ASCEBC(phs_hdr->srcUserID, sizeof(phs_hdr->srcUserID));
223 	ASCEBC(phs_hdr->srcAppName, sizeof(phs_hdr->srcAppName));
224 	if (imsg)
225 		memcpy(&phs_hdr->iucv_hdr, imsg, sizeof(struct iucv_message));
226 
227 	skb->dev = iucv->hs_dev;
228 	if (!skb->dev) {
229 		err = -ENODEV;
230 		goto err_free;
231 	}
232 
233 	dev_hard_header(skb, skb->dev, ETH_P_AF_IUCV, NULL, NULL, skb->len);
234 
235 	if (!(skb->dev->flags & IFF_UP) || !netif_carrier_ok(skb->dev)) {
236 		err = -ENETDOWN;
237 		goto err_free;
238 	}
239 	if (skb->len > skb->dev->mtu) {
240 		if (sock->sk_type == SOCK_SEQPACKET) {
241 			err = -EMSGSIZE;
242 			goto err_free;
243 		}
244 		err = pskb_trim(skb, skb->dev->mtu);
245 		if (err)
246 			goto err_free;
247 	}
248 	skb->protocol = cpu_to_be16(ETH_P_AF_IUCV);
249 
250 	atomic_inc(&iucv->skbs_in_xmit);
251 	err = dev_queue_xmit(skb);
252 	if (net_xmit_eval(err)) {
253 		atomic_dec(&iucv->skbs_in_xmit);
254 	} else {
255 		atomic_sub(confirm_recv, &iucv->msg_recv);
256 		WARN_ON(atomic_read(&iucv->msg_recv) < 0);
257 	}
258 	return net_xmit_eval(err);
259 
260 err_free:
261 	kfree_skb(skb);
262 	return err;
263 }
264 
265 static struct sock *__iucv_get_sock_by_name(char *nm)
266 {
267 	struct sock *sk;
268 
269 	sk_for_each(sk, &iucv_sk_list.head)
270 		if (!memcmp(&iucv_sk(sk)->src_name, nm, 8))
271 			return sk;
272 
273 	return NULL;
274 }
275 
276 static void iucv_sock_destruct(struct sock *sk)
277 {
278 	skb_queue_purge(&sk->sk_receive_queue);
279 	skb_queue_purge(&sk->sk_error_queue);
280 
281 	if (!sock_flag(sk, SOCK_DEAD)) {
282 		pr_err("Attempt to release alive iucv socket %p\n", sk);
283 		return;
284 	}
285 
286 	WARN_ON(atomic_read(&sk->sk_rmem_alloc));
287 	WARN_ON(refcount_read(&sk->sk_wmem_alloc));
288 	WARN_ON(sk->sk_wmem_queued);
289 	WARN_ON(sk->sk_forward_alloc);
290 }
291 
292 /* Cleanup Listen */
293 static void iucv_sock_cleanup_listen(struct sock *parent)
294 {
295 	struct sock *sk;
296 
297 	/* Close non-accepted connections */
298 	while ((sk = iucv_accept_dequeue(parent, NULL))) {
299 		iucv_sock_close(sk);
300 		iucv_sock_kill(sk);
301 	}
302 
303 	parent->sk_state = IUCV_CLOSED;
304 }
305 
306 static void iucv_sock_link(struct iucv_sock_list *l, struct sock *sk)
307 {
308 	write_lock_bh(&l->lock);
309 	sk_add_node(sk, &l->head);
310 	write_unlock_bh(&l->lock);
311 }
312 
313 static void iucv_sock_unlink(struct iucv_sock_list *l, struct sock *sk)
314 {
315 	write_lock_bh(&l->lock);
316 	sk_del_node_init(sk);
317 	write_unlock_bh(&l->lock);
318 }
319 
320 /* Kill socket (only if zapped and orphaned) */
321 static void iucv_sock_kill(struct sock *sk)
322 {
323 	if (!sock_flag(sk, SOCK_ZAPPED) || sk->sk_socket)
324 		return;
325 
326 	iucv_sock_unlink(&iucv_sk_list, sk);
327 	sock_set_flag(sk, SOCK_DEAD);
328 	sock_put(sk);
329 }
330 
331 /* Terminate an IUCV path */
332 static void iucv_sever_path(struct sock *sk, int with_user_data)
333 {
334 	unsigned char user_data[16];
335 	struct iucv_sock *iucv = iucv_sk(sk);
336 	struct iucv_path *path = iucv->path;
337 
338 	if (iucv->path) {
339 		iucv->path = NULL;
340 		if (with_user_data) {
341 			low_nmcpy(user_data, iucv->src_name);
342 			high_nmcpy(user_data, iucv->dst_name);
343 			ASCEBC(user_data, sizeof(user_data));
344 			pr_iucv->path_sever(path, user_data);
345 		} else
346 			pr_iucv->path_sever(path, NULL);
347 		iucv_path_free(path);
348 	}
349 }
350 
351 /* Send controlling flags through an IUCV socket for HIPER transport */
352 static int iucv_send_ctrl(struct sock *sk, u8 flags)
353 {
354 	struct iucv_sock *iucv = iucv_sk(sk);
355 	int err = 0;
356 	int blen;
357 	struct sk_buff *skb;
358 	u8 shutdown = 0;
359 
360 	blen = sizeof(struct af_iucv_trans_hdr) +
361 	       LL_RESERVED_SPACE(iucv->hs_dev);
362 	if (sk->sk_shutdown & SEND_SHUTDOWN) {
363 		/* controlling flags should be sent anyway */
364 		shutdown = sk->sk_shutdown;
365 		sk->sk_shutdown &= RCV_SHUTDOWN;
366 	}
367 	skb = sock_alloc_send_skb(sk, blen, 1, &err);
368 	if (skb) {
369 		skb_reserve(skb, blen);
370 		err = afiucv_hs_send(NULL, sk, skb, flags);
371 	}
372 	if (shutdown)
373 		sk->sk_shutdown = shutdown;
374 	return err;
375 }
376 
377 /* Close an IUCV socket */
378 static void iucv_sock_close(struct sock *sk)
379 {
380 	struct iucv_sock *iucv = iucv_sk(sk);
381 	unsigned long timeo;
382 	int err = 0;
383 
384 	lock_sock(sk);
385 
386 	switch (sk->sk_state) {
387 	case IUCV_LISTEN:
388 		iucv_sock_cleanup_listen(sk);
389 		break;
390 
391 	case IUCV_CONNECTED:
392 		if (iucv->transport == AF_IUCV_TRANS_HIPER) {
393 			err = iucv_send_ctrl(sk, AF_IUCV_FLAG_FIN);
394 			sk->sk_state = IUCV_DISCONN;
395 			sk->sk_state_change(sk);
396 		}
397 		fallthrough;
398 
399 	case IUCV_DISCONN:
400 		sk->sk_state = IUCV_CLOSING;
401 		sk->sk_state_change(sk);
402 
403 		if (!err && atomic_read(&iucv->skbs_in_xmit) > 0) {
404 			if (sock_flag(sk, SOCK_LINGER) && sk->sk_lingertime)
405 				timeo = sk->sk_lingertime;
406 			else
407 				timeo = IUCV_DISCONN_TIMEOUT;
408 			iucv_sock_wait(sk,
409 					iucv_sock_in_state(sk, IUCV_CLOSED, 0),
410 					timeo);
411 		}
412 		fallthrough;
413 
414 	case IUCV_CLOSING:
415 		sk->sk_state = IUCV_CLOSED;
416 		sk->sk_state_change(sk);
417 
418 		sk->sk_err = ECONNRESET;
419 		sk->sk_state_change(sk);
420 
421 		skb_queue_purge(&iucv->send_skb_q);
422 		skb_queue_purge(&iucv->backlog_skb_q);
423 		fallthrough;
424 
425 	default:
426 		iucv_sever_path(sk, 1);
427 	}
428 
429 	if (iucv->hs_dev) {
430 		dev_put(iucv->hs_dev);
431 		iucv->hs_dev = NULL;
432 		sk->sk_bound_dev_if = 0;
433 	}
434 
435 	/* mark socket for deletion by iucv_sock_kill() */
436 	sock_set_flag(sk, SOCK_ZAPPED);
437 
438 	release_sock(sk);
439 }
440 
441 static void iucv_sock_init(struct sock *sk, struct sock *parent)
442 {
443 	if (parent) {
444 		sk->sk_type = parent->sk_type;
445 		security_sk_clone(parent, sk);
446 	}
447 }
448 
449 static struct sock *iucv_sock_alloc(struct socket *sock, int proto, gfp_t prio, int kern)
450 {
451 	struct sock *sk;
452 	struct iucv_sock *iucv;
453 
454 	sk = sk_alloc(&init_net, PF_IUCV, prio, &iucv_proto, kern);
455 	if (!sk)
456 		return NULL;
457 	iucv = iucv_sk(sk);
458 
459 	sock_init_data(sock, sk);
460 	INIT_LIST_HEAD(&iucv->accept_q);
461 	spin_lock_init(&iucv->accept_q_lock);
462 	skb_queue_head_init(&iucv->send_skb_q);
463 	INIT_LIST_HEAD(&iucv->message_q.list);
464 	spin_lock_init(&iucv->message_q.lock);
465 	skb_queue_head_init(&iucv->backlog_skb_q);
466 	iucv->send_tag = 0;
467 	atomic_set(&iucv->pendings, 0);
468 	iucv->flags = 0;
469 	iucv->msglimit = 0;
470 	atomic_set(&iucv->skbs_in_xmit, 0);
471 	atomic_set(&iucv->msg_sent, 0);
472 	atomic_set(&iucv->msg_recv, 0);
473 	iucv->path = NULL;
474 	iucv->sk_txnotify = afiucv_hs_callback_txnotify;
475 	memset(&iucv->init, 0, sizeof(iucv->init));
476 	if (pr_iucv)
477 		iucv->transport = AF_IUCV_TRANS_IUCV;
478 	else
479 		iucv->transport = AF_IUCV_TRANS_HIPER;
480 
481 	sk->sk_destruct = iucv_sock_destruct;
482 	sk->sk_sndtimeo = IUCV_CONN_TIMEOUT;
483 
484 	sock_reset_flag(sk, SOCK_ZAPPED);
485 
486 	sk->sk_protocol = proto;
487 	sk->sk_state	= IUCV_OPEN;
488 
489 	iucv_sock_link(&iucv_sk_list, sk);
490 	return sk;
491 }
492 
493 static void iucv_accept_enqueue(struct sock *parent, struct sock *sk)
494 {
495 	unsigned long flags;
496 	struct iucv_sock *par = iucv_sk(parent);
497 
498 	sock_hold(sk);
499 	spin_lock_irqsave(&par->accept_q_lock, flags);
500 	list_add_tail(&iucv_sk(sk)->accept_q, &par->accept_q);
501 	spin_unlock_irqrestore(&par->accept_q_lock, flags);
502 	iucv_sk(sk)->parent = parent;
503 	sk_acceptq_added(parent);
504 }
505 
506 static void iucv_accept_unlink(struct sock *sk)
507 {
508 	unsigned long flags;
509 	struct iucv_sock *par = iucv_sk(iucv_sk(sk)->parent);
510 
511 	spin_lock_irqsave(&par->accept_q_lock, flags);
512 	list_del_init(&iucv_sk(sk)->accept_q);
513 	spin_unlock_irqrestore(&par->accept_q_lock, flags);
514 	sk_acceptq_removed(iucv_sk(sk)->parent);
515 	iucv_sk(sk)->parent = NULL;
516 	sock_put(sk);
517 }
518 
519 static struct sock *iucv_accept_dequeue(struct sock *parent,
520 					struct socket *newsock)
521 {
522 	struct iucv_sock *isk, *n;
523 	struct sock *sk;
524 
525 	list_for_each_entry_safe(isk, n, &iucv_sk(parent)->accept_q, accept_q) {
526 		sk = (struct sock *) isk;
527 		lock_sock(sk);
528 
529 		if (sk->sk_state == IUCV_CLOSED) {
530 			iucv_accept_unlink(sk);
531 			release_sock(sk);
532 			continue;
533 		}
534 
535 		if (sk->sk_state == IUCV_CONNECTED ||
536 		    sk->sk_state == IUCV_DISCONN ||
537 		    !newsock) {
538 			iucv_accept_unlink(sk);
539 			if (newsock)
540 				sock_graft(sk, newsock);
541 
542 			release_sock(sk);
543 			return sk;
544 		}
545 
546 		release_sock(sk);
547 	}
548 	return NULL;
549 }
550 
551 static void __iucv_auto_name(struct iucv_sock *iucv)
552 {
553 	char name[12];
554 
555 	sprintf(name, "%08x", atomic_inc_return(&iucv_sk_list.autobind_name));
556 	while (__iucv_get_sock_by_name(name)) {
557 		sprintf(name, "%08x",
558 			atomic_inc_return(&iucv_sk_list.autobind_name));
559 	}
560 	memcpy(iucv->src_name, name, 8);
561 }
562 
563 /* Bind an unbound socket */
564 static int iucv_sock_bind(struct socket *sock, struct sockaddr *addr,
565 			  int addr_len)
566 {
567 	DECLARE_SOCKADDR(struct sockaddr_iucv *, sa, addr);
568 	char uid[sizeof(sa->siucv_user_id)];
569 	struct sock *sk = sock->sk;
570 	struct iucv_sock *iucv;
571 	int err = 0;
572 	struct net_device *dev;
573 
574 	/* Verify the input sockaddr */
575 	if (addr_len < sizeof(struct sockaddr_iucv) ||
576 	    addr->sa_family != AF_IUCV)
577 		return -EINVAL;
578 
579 	lock_sock(sk);
580 	if (sk->sk_state != IUCV_OPEN) {
581 		err = -EBADFD;
582 		goto done;
583 	}
584 
585 	write_lock_bh(&iucv_sk_list.lock);
586 
587 	iucv = iucv_sk(sk);
588 	if (__iucv_get_sock_by_name(sa->siucv_name)) {
589 		err = -EADDRINUSE;
590 		goto done_unlock;
591 	}
592 	if (iucv->path)
593 		goto done_unlock;
594 
595 	/* Bind the socket */
596 	if (pr_iucv)
597 		if (!memcmp(sa->siucv_user_id, iucv_userid, 8))
598 			goto vm_bind; /* VM IUCV transport */
599 
600 	/* try hiper transport */
601 	memcpy(uid, sa->siucv_user_id, sizeof(uid));
602 	ASCEBC(uid, 8);
603 	rcu_read_lock();
604 	for_each_netdev_rcu(&init_net, dev) {
605 		if (!memcmp(dev->perm_addr, uid, 8)) {
606 			memcpy(iucv->src_user_id, sa->siucv_user_id, 8);
607 			/* Check for uninitialized siucv_name */
608 			if (strncmp(sa->siucv_name, "        ", 8) == 0)
609 				__iucv_auto_name(iucv);
610 			else
611 				memcpy(iucv->src_name, sa->siucv_name, 8);
612 			sk->sk_bound_dev_if = dev->ifindex;
613 			iucv->hs_dev = dev;
614 			dev_hold(dev);
615 			sk->sk_state = IUCV_BOUND;
616 			iucv->transport = AF_IUCV_TRANS_HIPER;
617 			if (!iucv->msglimit)
618 				iucv->msglimit = IUCV_HIPER_MSGLIM_DEFAULT;
619 			rcu_read_unlock();
620 			goto done_unlock;
621 		}
622 	}
623 	rcu_read_unlock();
624 vm_bind:
625 	if (pr_iucv) {
626 		/* use local userid for backward compat */
627 		memcpy(iucv->src_name, sa->siucv_name, 8);
628 		memcpy(iucv->src_user_id, iucv_userid, 8);
629 		sk->sk_state = IUCV_BOUND;
630 		iucv->transport = AF_IUCV_TRANS_IUCV;
631 		sk->sk_allocation |= GFP_DMA;
632 		if (!iucv->msglimit)
633 			iucv->msglimit = IUCV_QUEUELEN_DEFAULT;
634 		goto done_unlock;
635 	}
636 	/* found no dev to bind */
637 	err = -ENODEV;
638 done_unlock:
639 	/* Release the socket list lock */
640 	write_unlock_bh(&iucv_sk_list.lock);
641 done:
642 	release_sock(sk);
643 	return err;
644 }
645 
646 /* Automatically bind an unbound socket */
647 static int iucv_sock_autobind(struct sock *sk)
648 {
649 	struct iucv_sock *iucv = iucv_sk(sk);
650 	int err = 0;
651 
652 	if (unlikely(!pr_iucv))
653 		return -EPROTO;
654 
655 	memcpy(iucv->src_user_id, iucv_userid, 8);
656 	iucv->transport = AF_IUCV_TRANS_IUCV;
657 	sk->sk_allocation |= GFP_DMA;
658 
659 	write_lock_bh(&iucv_sk_list.lock);
660 	__iucv_auto_name(iucv);
661 	write_unlock_bh(&iucv_sk_list.lock);
662 
663 	if (!iucv->msglimit)
664 		iucv->msglimit = IUCV_QUEUELEN_DEFAULT;
665 
666 	return err;
667 }
668 
669 static int afiucv_path_connect(struct socket *sock, struct sockaddr *addr)
670 {
671 	DECLARE_SOCKADDR(struct sockaddr_iucv *, sa, addr);
672 	struct sock *sk = sock->sk;
673 	struct iucv_sock *iucv = iucv_sk(sk);
674 	unsigned char user_data[16];
675 	int err;
676 
677 	high_nmcpy(user_data, sa->siucv_name);
678 	low_nmcpy(user_data, iucv->src_name);
679 	ASCEBC(user_data, sizeof(user_data));
680 
681 	/* Create path. */
682 	iucv->path = iucv_path_alloc(iucv->msglimit,
683 				     IUCV_IPRMDATA, GFP_KERNEL);
684 	if (!iucv->path) {
685 		err = -ENOMEM;
686 		goto done;
687 	}
688 	err = pr_iucv->path_connect(iucv->path, &af_iucv_handler,
689 				    sa->siucv_user_id, NULL, user_data,
690 				    sk);
691 	if (err) {
692 		iucv_path_free(iucv->path);
693 		iucv->path = NULL;
694 		switch (err) {
695 		case 0x0b:	/* Target communicator is not logged on */
696 			err = -ENETUNREACH;
697 			break;
698 		case 0x0d:	/* Max connections for this guest exceeded */
699 		case 0x0e:	/* Max connections for target guest exceeded */
700 			err = -EAGAIN;
701 			break;
702 		case 0x0f:	/* Missing IUCV authorization */
703 			err = -EACCES;
704 			break;
705 		default:
706 			err = -ECONNREFUSED;
707 			break;
708 		}
709 	}
710 done:
711 	return err;
712 }
713 
714 /* Connect an unconnected socket */
715 static int iucv_sock_connect(struct socket *sock, struct sockaddr *addr,
716 			     int alen, int flags)
717 {
718 	DECLARE_SOCKADDR(struct sockaddr_iucv *, sa, addr);
719 	struct sock *sk = sock->sk;
720 	struct iucv_sock *iucv = iucv_sk(sk);
721 	int err;
722 
723 	if (alen < sizeof(struct sockaddr_iucv) || addr->sa_family != AF_IUCV)
724 		return -EINVAL;
725 
726 	if (sk->sk_state != IUCV_OPEN && sk->sk_state != IUCV_BOUND)
727 		return -EBADFD;
728 
729 	if (sk->sk_state == IUCV_OPEN &&
730 	    iucv->transport == AF_IUCV_TRANS_HIPER)
731 		return -EBADFD; /* explicit bind required */
732 
733 	if (sk->sk_type != SOCK_STREAM && sk->sk_type != SOCK_SEQPACKET)
734 		return -EINVAL;
735 
736 	if (sk->sk_state == IUCV_OPEN) {
737 		err = iucv_sock_autobind(sk);
738 		if (unlikely(err))
739 			return err;
740 	}
741 
742 	lock_sock(sk);
743 
744 	/* Set the destination information */
745 	memcpy(iucv->dst_user_id, sa->siucv_user_id, 8);
746 	memcpy(iucv->dst_name, sa->siucv_name, 8);
747 
748 	if (iucv->transport == AF_IUCV_TRANS_HIPER)
749 		err = iucv_send_ctrl(sock->sk, AF_IUCV_FLAG_SYN);
750 	else
751 		err = afiucv_path_connect(sock, addr);
752 	if (err)
753 		goto done;
754 
755 	if (sk->sk_state != IUCV_CONNECTED)
756 		err = iucv_sock_wait(sk, iucv_sock_in_state(sk, IUCV_CONNECTED,
757 							    IUCV_DISCONN),
758 				     sock_sndtimeo(sk, flags & O_NONBLOCK));
759 
760 	if (sk->sk_state == IUCV_DISCONN || sk->sk_state == IUCV_CLOSED)
761 		err = -ECONNREFUSED;
762 
763 	if (err && iucv->transport == AF_IUCV_TRANS_IUCV)
764 		iucv_sever_path(sk, 0);
765 
766 done:
767 	release_sock(sk);
768 	return err;
769 }
770 
771 /* Move a socket into listening state. */
772 static int iucv_sock_listen(struct socket *sock, int backlog)
773 {
774 	struct sock *sk = sock->sk;
775 	int err;
776 
777 	lock_sock(sk);
778 
779 	err = -EINVAL;
780 	if (sk->sk_state != IUCV_BOUND)
781 		goto done;
782 
783 	if (sock->type != SOCK_STREAM && sock->type != SOCK_SEQPACKET)
784 		goto done;
785 
786 	sk->sk_max_ack_backlog = backlog;
787 	sk->sk_ack_backlog = 0;
788 	sk->sk_state = IUCV_LISTEN;
789 	err = 0;
790 
791 done:
792 	release_sock(sk);
793 	return err;
794 }
795 
796 /* Accept a pending connection */
797 static int iucv_sock_accept(struct socket *sock, struct socket *newsock,
798 			    int flags, bool kern)
799 {
800 	DECLARE_WAITQUEUE(wait, current);
801 	struct sock *sk = sock->sk, *nsk;
802 	long timeo;
803 	int err = 0;
804 
805 	lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
806 
807 	if (sk->sk_state != IUCV_LISTEN) {
808 		err = -EBADFD;
809 		goto done;
810 	}
811 
812 	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
813 
814 	/* Wait for an incoming connection */
815 	add_wait_queue_exclusive(sk_sleep(sk), &wait);
816 	while (!(nsk = iucv_accept_dequeue(sk, newsock))) {
817 		set_current_state(TASK_INTERRUPTIBLE);
818 		if (!timeo) {
819 			err = -EAGAIN;
820 			break;
821 		}
822 
823 		release_sock(sk);
824 		timeo = schedule_timeout(timeo);
825 		lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
826 
827 		if (sk->sk_state != IUCV_LISTEN) {
828 			err = -EBADFD;
829 			break;
830 		}
831 
832 		if (signal_pending(current)) {
833 			err = sock_intr_errno(timeo);
834 			break;
835 		}
836 	}
837 
838 	set_current_state(TASK_RUNNING);
839 	remove_wait_queue(sk_sleep(sk), &wait);
840 
841 	if (err)
842 		goto done;
843 
844 	newsock->state = SS_CONNECTED;
845 
846 done:
847 	release_sock(sk);
848 	return err;
849 }
850 
851 static int iucv_sock_getname(struct socket *sock, struct sockaddr *addr,
852 			     int peer)
853 {
854 	DECLARE_SOCKADDR(struct sockaddr_iucv *, siucv, addr);
855 	struct sock *sk = sock->sk;
856 	struct iucv_sock *iucv = iucv_sk(sk);
857 
858 	addr->sa_family = AF_IUCV;
859 
860 	if (peer) {
861 		memcpy(siucv->siucv_user_id, iucv->dst_user_id, 8);
862 		memcpy(siucv->siucv_name, iucv->dst_name, 8);
863 	} else {
864 		memcpy(siucv->siucv_user_id, iucv->src_user_id, 8);
865 		memcpy(siucv->siucv_name, iucv->src_name, 8);
866 	}
867 	memset(&siucv->siucv_port, 0, sizeof(siucv->siucv_port));
868 	memset(&siucv->siucv_addr, 0, sizeof(siucv->siucv_addr));
869 	memset(&siucv->siucv_nodeid, 0, sizeof(siucv->siucv_nodeid));
870 
871 	return sizeof(struct sockaddr_iucv);
872 }
873 
874 /**
875  * iucv_send_iprm() - Send socket data in parameter list of an iucv message.
876  * @path:	IUCV path
877  * @msg:	Pointer to a struct iucv_message
878  * @skb:	The socket data to send, skb->len MUST BE <= 7
879  *
880  * Send the socket data in the parameter list in the iucv message
881  * (IUCV_IPRMDATA). The socket data is stored at index 0 to 6 in the parameter
882  * list and the socket data len at index 7 (last byte).
883  * See also iucv_msg_length().
884  *
885  * Returns the error code from the iucv_message_send() call.
886  */
887 static int iucv_send_iprm(struct iucv_path *path, struct iucv_message *msg,
888 			  struct sk_buff *skb)
889 {
890 	u8 prmdata[8];
891 
892 	memcpy(prmdata, (void *) skb->data, skb->len);
893 	prmdata[7] = 0xff - (u8) skb->len;
894 	return pr_iucv->message_send(path, msg, IUCV_IPRMDATA, 0,
895 				 (void *) prmdata, 8);
896 }
897 
898 static int iucv_sock_sendmsg(struct socket *sock, struct msghdr *msg,
899 			     size_t len)
900 {
901 	struct sock *sk = sock->sk;
902 	struct iucv_sock *iucv = iucv_sk(sk);
903 	size_t headroom = 0;
904 	size_t linear;
905 	struct sk_buff *skb;
906 	struct iucv_message txmsg = {0};
907 	struct cmsghdr *cmsg;
908 	int cmsg_done;
909 	long timeo;
910 	char user_id[9];
911 	char appl_id[9];
912 	int err;
913 	int noblock = msg->msg_flags & MSG_DONTWAIT;
914 
915 	err = sock_error(sk);
916 	if (err)
917 		return err;
918 
919 	if (msg->msg_flags & MSG_OOB)
920 		return -EOPNOTSUPP;
921 
922 	/* SOCK_SEQPACKET: we do not support segmented records */
923 	if (sk->sk_type == SOCK_SEQPACKET && !(msg->msg_flags & MSG_EOR))
924 		return -EOPNOTSUPP;
925 
926 	lock_sock(sk);
927 
928 	if (sk->sk_shutdown & SEND_SHUTDOWN) {
929 		err = -EPIPE;
930 		goto out;
931 	}
932 
933 	/* Return if the socket is not in connected state */
934 	if (sk->sk_state != IUCV_CONNECTED) {
935 		err = -ENOTCONN;
936 		goto out;
937 	}
938 
939 	/* initialize defaults */
940 	cmsg_done   = 0;	/* check for duplicate headers */
941 
942 	/* iterate over control messages */
943 	for_each_cmsghdr(cmsg, msg) {
944 		if (!CMSG_OK(msg, cmsg)) {
945 			err = -EINVAL;
946 			goto out;
947 		}
948 
949 		if (cmsg->cmsg_level != SOL_IUCV)
950 			continue;
951 
952 		if (cmsg->cmsg_type & cmsg_done) {
953 			err = -EINVAL;
954 			goto out;
955 		}
956 		cmsg_done |= cmsg->cmsg_type;
957 
958 		switch (cmsg->cmsg_type) {
959 		case SCM_IUCV_TRGCLS:
960 			if (cmsg->cmsg_len != CMSG_LEN(TRGCLS_SIZE)) {
961 				err = -EINVAL;
962 				goto out;
963 			}
964 
965 			/* set iucv message target class */
966 			memcpy(&txmsg.class,
967 				(void *) CMSG_DATA(cmsg), TRGCLS_SIZE);
968 
969 			break;
970 
971 		default:
972 			err = -EINVAL;
973 			goto out;
974 		}
975 	}
976 
977 	/* allocate one skb for each iucv message:
978 	 * this is fine for SOCK_SEQPACKET (unless we want to support
979 	 * segmented records using the MSG_EOR flag), but
980 	 * for SOCK_STREAM we might want to improve it in future */
981 	if (iucv->transport == AF_IUCV_TRANS_HIPER) {
982 		headroom = sizeof(struct af_iucv_trans_hdr) +
983 			   LL_RESERVED_SPACE(iucv->hs_dev);
984 		linear = min(len, PAGE_SIZE - headroom);
985 	} else {
986 		if (len < PAGE_SIZE) {
987 			linear = len;
988 		} else {
989 			/* In nonlinear "classic" iucv skb,
990 			 * reserve space for iucv_array
991 			 */
992 			headroom = sizeof(struct iucv_array) *
993 				   (MAX_SKB_FRAGS + 1);
994 			linear = PAGE_SIZE - headroom;
995 		}
996 	}
997 	skb = sock_alloc_send_pskb(sk, headroom + linear, len - linear,
998 				   noblock, &err, 0);
999 	if (!skb)
1000 		goto out;
1001 	if (headroom)
1002 		skb_reserve(skb, headroom);
1003 	skb_put(skb, linear);
1004 	skb->len = len;
1005 	skb->data_len = len - linear;
1006 	err = skb_copy_datagram_from_iter(skb, 0, &msg->msg_iter, len);
1007 	if (err)
1008 		goto fail;
1009 
1010 	/* wait if outstanding messages for iucv path has reached */
1011 	timeo = sock_sndtimeo(sk, noblock);
1012 	err = iucv_sock_wait(sk, iucv_below_msglim(sk), timeo);
1013 	if (err)
1014 		goto fail;
1015 
1016 	/* return -ECONNRESET if the socket is no longer connected */
1017 	if (sk->sk_state != IUCV_CONNECTED) {
1018 		err = -ECONNRESET;
1019 		goto fail;
1020 	}
1021 
1022 	/* increment and save iucv message tag for msg_completion cbk */
1023 	txmsg.tag = iucv->send_tag++;
1024 	IUCV_SKB_CB(skb)->tag = txmsg.tag;
1025 
1026 	if (iucv->transport == AF_IUCV_TRANS_HIPER) {
1027 		atomic_inc(&iucv->msg_sent);
1028 		err = afiucv_hs_send(&txmsg, sk, skb, 0);
1029 		if (err) {
1030 			atomic_dec(&iucv->msg_sent);
1031 			goto out;
1032 		}
1033 	} else { /* Classic VM IUCV transport */
1034 		skb_queue_tail(&iucv->send_skb_q, skb);
1035 		atomic_inc(&iucv->skbs_in_xmit);
1036 
1037 		if (((iucv->path->flags & IUCV_IPRMDATA) & iucv->flags) &&
1038 		    skb->len <= 7) {
1039 			err = iucv_send_iprm(iucv->path, &txmsg, skb);
1040 
1041 			/* on success: there is no message_complete callback */
1042 			/* for an IPRMDATA msg; remove skb from send queue   */
1043 			if (err == 0) {
1044 				atomic_dec(&iucv->skbs_in_xmit);
1045 				skb_unlink(skb, &iucv->send_skb_q);
1046 				consume_skb(skb);
1047 			}
1048 
1049 			/* this error should never happen since the	*/
1050 			/* IUCV_IPRMDATA path flag is set... sever path */
1051 			if (err == 0x15) {
1052 				pr_iucv->path_sever(iucv->path, NULL);
1053 				atomic_dec(&iucv->skbs_in_xmit);
1054 				skb_unlink(skb, &iucv->send_skb_q);
1055 				err = -EPIPE;
1056 				goto fail;
1057 			}
1058 		} else if (skb_is_nonlinear(skb)) {
1059 			struct iucv_array *iba = (struct iucv_array *)skb->head;
1060 			int i;
1061 
1062 			/* skip iucv_array lying in the headroom */
1063 			iba[0].address = (u32)(addr_t)skb->data;
1064 			iba[0].length = (u32)skb_headlen(skb);
1065 			for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1066 				skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1067 
1068 				iba[i + 1].address =
1069 					(u32)(addr_t)skb_frag_address(frag);
1070 				iba[i + 1].length = (u32)skb_frag_size(frag);
1071 			}
1072 			err = pr_iucv->message_send(iucv->path, &txmsg,
1073 						    IUCV_IPBUFLST, 0,
1074 						    (void *)iba, skb->len);
1075 		} else { /* non-IPRM Linear skb */
1076 			err = pr_iucv->message_send(iucv->path, &txmsg,
1077 					0, 0, (void *)skb->data, skb->len);
1078 		}
1079 		if (err) {
1080 			if (err == 3) {
1081 				user_id[8] = 0;
1082 				memcpy(user_id, iucv->dst_user_id, 8);
1083 				appl_id[8] = 0;
1084 				memcpy(appl_id, iucv->dst_name, 8);
1085 				pr_err(
1086 		"Application %s on z/VM guest %s exceeds message limit\n",
1087 					appl_id, user_id);
1088 				err = -EAGAIN;
1089 			} else {
1090 				err = -EPIPE;
1091 			}
1092 
1093 			atomic_dec(&iucv->skbs_in_xmit);
1094 			skb_unlink(skb, &iucv->send_skb_q);
1095 			goto fail;
1096 		}
1097 	}
1098 
1099 	release_sock(sk);
1100 	return len;
1101 
1102 fail:
1103 	kfree_skb(skb);
1104 out:
1105 	release_sock(sk);
1106 	return err;
1107 }
1108 
1109 static struct sk_buff *alloc_iucv_recv_skb(unsigned long len)
1110 {
1111 	size_t headroom, linear;
1112 	struct sk_buff *skb;
1113 	int err;
1114 
1115 	if (len < PAGE_SIZE) {
1116 		headroom = 0;
1117 		linear = len;
1118 	} else {
1119 		headroom = sizeof(struct iucv_array) * (MAX_SKB_FRAGS + 1);
1120 		linear = PAGE_SIZE - headroom;
1121 	}
1122 	skb = alloc_skb_with_frags(headroom + linear, len - linear,
1123 				   0, &err, GFP_ATOMIC | GFP_DMA);
1124 	WARN_ONCE(!skb,
1125 		  "alloc of recv iucv skb len=%lu failed with errcode=%d\n",
1126 		  len, err);
1127 	if (skb) {
1128 		if (headroom)
1129 			skb_reserve(skb, headroom);
1130 		skb_put(skb, linear);
1131 		skb->len = len;
1132 		skb->data_len = len - linear;
1133 	}
1134 	return skb;
1135 }
1136 
1137 /* iucv_process_message() - Receive a single outstanding IUCV message
1138  *
1139  * Locking: must be called with message_q.lock held
1140  */
1141 static void iucv_process_message(struct sock *sk, struct sk_buff *skb,
1142 				 struct iucv_path *path,
1143 				 struct iucv_message *msg)
1144 {
1145 	int rc;
1146 	unsigned int len;
1147 
1148 	len = iucv_msg_length(msg);
1149 
1150 	/* store msg target class in the second 4 bytes of skb ctrl buffer */
1151 	/* Note: the first 4 bytes are reserved for msg tag */
1152 	IUCV_SKB_CB(skb)->class = msg->class;
1153 
1154 	/* check for special IPRM messages (e.g. iucv_sock_shutdown) */
1155 	if ((msg->flags & IUCV_IPRMDATA) && len > 7) {
1156 		if (memcmp(msg->rmmsg, iprm_shutdown, 8) == 0) {
1157 			skb->data = NULL;
1158 			skb->len = 0;
1159 		}
1160 	} else {
1161 		if (skb_is_nonlinear(skb)) {
1162 			struct iucv_array *iba = (struct iucv_array *)skb->head;
1163 			int i;
1164 
1165 			iba[0].address = (u32)(addr_t)skb->data;
1166 			iba[0].length = (u32)skb_headlen(skb);
1167 			for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1168 				skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1169 
1170 				iba[i + 1].address =
1171 					(u32)(addr_t)skb_frag_address(frag);
1172 				iba[i + 1].length = (u32)skb_frag_size(frag);
1173 			}
1174 			rc = pr_iucv->message_receive(path, msg,
1175 					      IUCV_IPBUFLST,
1176 					      (void *)iba, len, NULL);
1177 		} else {
1178 			rc = pr_iucv->message_receive(path, msg,
1179 					      msg->flags & IUCV_IPRMDATA,
1180 					      skb->data, len, NULL);
1181 		}
1182 		if (rc) {
1183 			kfree_skb(skb);
1184 			return;
1185 		}
1186 		WARN_ON_ONCE(skb->len != len);
1187 	}
1188 
1189 	IUCV_SKB_CB(skb)->offset = 0;
1190 	if (sk_filter(sk, skb)) {
1191 		atomic_inc(&sk->sk_drops);	/* skb rejected by filter */
1192 		kfree_skb(skb);
1193 		return;
1194 	}
1195 	if (__sock_queue_rcv_skb(sk, skb))	/* handle rcv queue full */
1196 		skb_queue_tail(&iucv_sk(sk)->backlog_skb_q, skb);
1197 }
1198 
1199 /* iucv_process_message_q() - Process outstanding IUCV messages
1200  *
1201  * Locking: must be called with message_q.lock held
1202  */
1203 static void iucv_process_message_q(struct sock *sk)
1204 {
1205 	struct iucv_sock *iucv = iucv_sk(sk);
1206 	struct sk_buff *skb;
1207 	struct sock_msg_q *p, *n;
1208 
1209 	list_for_each_entry_safe(p, n, &iucv->message_q.list, list) {
1210 		skb = alloc_iucv_recv_skb(iucv_msg_length(&p->msg));
1211 		if (!skb)
1212 			break;
1213 		iucv_process_message(sk, skb, p->path, &p->msg);
1214 		list_del(&p->list);
1215 		kfree(p);
1216 		if (!skb_queue_empty(&iucv->backlog_skb_q))
1217 			break;
1218 	}
1219 }
1220 
1221 static int iucv_sock_recvmsg(struct socket *sock, struct msghdr *msg,
1222 			     size_t len, int flags)
1223 {
1224 	struct sock *sk = sock->sk;
1225 	struct iucv_sock *iucv = iucv_sk(sk);
1226 	unsigned int copied, rlen;
1227 	struct sk_buff *skb, *rskb, *cskb;
1228 	int err = 0;
1229 	u32 offset;
1230 
1231 	if ((sk->sk_state == IUCV_DISCONN) &&
1232 	    skb_queue_empty(&iucv->backlog_skb_q) &&
1233 	    skb_queue_empty(&sk->sk_receive_queue) &&
1234 	    list_empty(&iucv->message_q.list))
1235 		return 0;
1236 
1237 	if (flags & (MSG_OOB))
1238 		return -EOPNOTSUPP;
1239 
1240 	/* receive/dequeue next skb:
1241 	 * the function understands MSG_PEEK and, thus, does not dequeue skb */
1242 	skb = skb_recv_datagram(sk, flags, &err);
1243 	if (!skb) {
1244 		if (sk->sk_shutdown & RCV_SHUTDOWN)
1245 			return 0;
1246 		return err;
1247 	}
1248 
1249 	offset = IUCV_SKB_CB(skb)->offset;
1250 	rlen   = skb->len - offset;		/* real length of skb */
1251 	copied = min_t(unsigned int, rlen, len);
1252 	if (!rlen)
1253 		sk->sk_shutdown = sk->sk_shutdown | RCV_SHUTDOWN;
1254 
1255 	cskb = skb;
1256 	if (skb_copy_datagram_msg(cskb, offset, msg, copied)) {
1257 		if (!(flags & MSG_PEEK))
1258 			skb_queue_head(&sk->sk_receive_queue, skb);
1259 		return -EFAULT;
1260 	}
1261 
1262 	/* SOCK_SEQPACKET: set MSG_TRUNC if recv buf size is too small */
1263 	if (sk->sk_type == SOCK_SEQPACKET) {
1264 		if (copied < rlen)
1265 			msg->msg_flags |= MSG_TRUNC;
1266 		/* each iucv message contains a complete record */
1267 		msg->msg_flags |= MSG_EOR;
1268 	}
1269 
1270 	/* create control message to store iucv msg target class:
1271 	 * get the trgcls from the control buffer of the skb due to
1272 	 * fragmentation of original iucv message. */
1273 	err = put_cmsg(msg, SOL_IUCV, SCM_IUCV_TRGCLS,
1274 		       sizeof(IUCV_SKB_CB(skb)->class),
1275 		       (void *)&IUCV_SKB_CB(skb)->class);
1276 	if (err) {
1277 		if (!(flags & MSG_PEEK))
1278 			skb_queue_head(&sk->sk_receive_queue, skb);
1279 		return err;
1280 	}
1281 
1282 	/* Mark read part of skb as used */
1283 	if (!(flags & MSG_PEEK)) {
1284 
1285 		/* SOCK_STREAM: re-queue skb if it contains unreceived data */
1286 		if (sk->sk_type == SOCK_STREAM) {
1287 			if (copied < rlen) {
1288 				IUCV_SKB_CB(skb)->offset = offset + copied;
1289 				skb_queue_head(&sk->sk_receive_queue, skb);
1290 				goto done;
1291 			}
1292 		}
1293 
1294 		consume_skb(skb);
1295 		if (iucv->transport == AF_IUCV_TRANS_HIPER) {
1296 			atomic_inc(&iucv->msg_recv);
1297 			if (atomic_read(&iucv->msg_recv) > iucv->msglimit) {
1298 				WARN_ON(1);
1299 				iucv_sock_close(sk);
1300 				return -EFAULT;
1301 			}
1302 		}
1303 
1304 		/* Queue backlog skbs */
1305 		spin_lock_bh(&iucv->message_q.lock);
1306 		rskb = skb_dequeue(&iucv->backlog_skb_q);
1307 		while (rskb) {
1308 			IUCV_SKB_CB(rskb)->offset = 0;
1309 			if (__sock_queue_rcv_skb(sk, rskb)) {
1310 				/* handle rcv queue full */
1311 				skb_queue_head(&iucv->backlog_skb_q,
1312 						rskb);
1313 				break;
1314 			}
1315 			rskb = skb_dequeue(&iucv->backlog_skb_q);
1316 		}
1317 		if (skb_queue_empty(&iucv->backlog_skb_q)) {
1318 			if (!list_empty(&iucv->message_q.list))
1319 				iucv_process_message_q(sk);
1320 			if (atomic_read(&iucv->msg_recv) >=
1321 							iucv->msglimit / 2) {
1322 				err = iucv_send_ctrl(sk, AF_IUCV_FLAG_WIN);
1323 				if (err) {
1324 					sk->sk_state = IUCV_DISCONN;
1325 					sk->sk_state_change(sk);
1326 				}
1327 			}
1328 		}
1329 		spin_unlock_bh(&iucv->message_q.lock);
1330 	}
1331 
1332 done:
1333 	/* SOCK_SEQPACKET: return real length if MSG_TRUNC is set */
1334 	if (sk->sk_type == SOCK_SEQPACKET && (flags & MSG_TRUNC))
1335 		copied = rlen;
1336 
1337 	return copied;
1338 }
1339 
1340 static inline __poll_t iucv_accept_poll(struct sock *parent)
1341 {
1342 	struct iucv_sock *isk, *n;
1343 	struct sock *sk;
1344 
1345 	list_for_each_entry_safe(isk, n, &iucv_sk(parent)->accept_q, accept_q) {
1346 		sk = (struct sock *) isk;
1347 
1348 		if (sk->sk_state == IUCV_CONNECTED)
1349 			return EPOLLIN | EPOLLRDNORM;
1350 	}
1351 
1352 	return 0;
1353 }
1354 
1355 static __poll_t iucv_sock_poll(struct file *file, struct socket *sock,
1356 			       poll_table *wait)
1357 {
1358 	struct sock *sk = sock->sk;
1359 	__poll_t mask = 0;
1360 
1361 	sock_poll_wait(file, sock, wait);
1362 
1363 	if (sk->sk_state == IUCV_LISTEN)
1364 		return iucv_accept_poll(sk);
1365 
1366 	if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
1367 		mask |= EPOLLERR |
1368 			(sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? EPOLLPRI : 0);
1369 
1370 	if (sk->sk_shutdown & RCV_SHUTDOWN)
1371 		mask |= EPOLLRDHUP;
1372 
1373 	if (sk->sk_shutdown == SHUTDOWN_MASK)
1374 		mask |= EPOLLHUP;
1375 
1376 	if (!skb_queue_empty(&sk->sk_receive_queue) ||
1377 	    (sk->sk_shutdown & RCV_SHUTDOWN))
1378 		mask |= EPOLLIN | EPOLLRDNORM;
1379 
1380 	if (sk->sk_state == IUCV_CLOSED)
1381 		mask |= EPOLLHUP;
1382 
1383 	if (sk->sk_state == IUCV_DISCONN)
1384 		mask |= EPOLLIN;
1385 
1386 	if (sock_writeable(sk) && iucv_below_msglim(sk))
1387 		mask |= EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND;
1388 	else
1389 		sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
1390 
1391 	return mask;
1392 }
1393 
1394 static int iucv_sock_shutdown(struct socket *sock, int how)
1395 {
1396 	struct sock *sk = sock->sk;
1397 	struct iucv_sock *iucv = iucv_sk(sk);
1398 	struct iucv_message txmsg;
1399 	int err = 0;
1400 
1401 	how++;
1402 
1403 	if ((how & ~SHUTDOWN_MASK) || !how)
1404 		return -EINVAL;
1405 
1406 	lock_sock(sk);
1407 	switch (sk->sk_state) {
1408 	case IUCV_LISTEN:
1409 	case IUCV_DISCONN:
1410 	case IUCV_CLOSING:
1411 	case IUCV_CLOSED:
1412 		err = -ENOTCONN;
1413 		goto fail;
1414 	default:
1415 		break;
1416 	}
1417 
1418 	if ((how == SEND_SHUTDOWN || how == SHUTDOWN_MASK) &&
1419 	    sk->sk_state == IUCV_CONNECTED) {
1420 		if (iucv->transport == AF_IUCV_TRANS_IUCV) {
1421 			txmsg.class = 0;
1422 			txmsg.tag = 0;
1423 			err = pr_iucv->message_send(iucv->path, &txmsg,
1424 				IUCV_IPRMDATA, 0, (void *) iprm_shutdown, 8);
1425 			if (err) {
1426 				switch (err) {
1427 				case 1:
1428 					err = -ENOTCONN;
1429 					break;
1430 				case 2:
1431 					err = -ECONNRESET;
1432 					break;
1433 				default:
1434 					err = -ENOTCONN;
1435 					break;
1436 				}
1437 			}
1438 		} else
1439 			iucv_send_ctrl(sk, AF_IUCV_FLAG_SHT);
1440 	}
1441 
1442 	sk->sk_shutdown |= how;
1443 	if (how == RCV_SHUTDOWN || how == SHUTDOWN_MASK) {
1444 		if ((iucv->transport == AF_IUCV_TRANS_IUCV) &&
1445 		    iucv->path) {
1446 			err = pr_iucv->path_quiesce(iucv->path, NULL);
1447 			if (err)
1448 				err = -ENOTCONN;
1449 /*			skb_queue_purge(&sk->sk_receive_queue); */
1450 		}
1451 		skb_queue_purge(&sk->sk_receive_queue);
1452 	}
1453 
1454 	/* Wake up anyone sleeping in poll */
1455 	sk->sk_state_change(sk);
1456 
1457 fail:
1458 	release_sock(sk);
1459 	return err;
1460 }
1461 
1462 static int iucv_sock_release(struct socket *sock)
1463 {
1464 	struct sock *sk = sock->sk;
1465 	int err = 0;
1466 
1467 	if (!sk)
1468 		return 0;
1469 
1470 	iucv_sock_close(sk);
1471 
1472 	sock_orphan(sk);
1473 	iucv_sock_kill(sk);
1474 	return err;
1475 }
1476 
1477 /* getsockopt and setsockopt */
1478 static int iucv_sock_setsockopt(struct socket *sock, int level, int optname,
1479 				sockptr_t optval, unsigned int optlen)
1480 {
1481 	struct sock *sk = sock->sk;
1482 	struct iucv_sock *iucv = iucv_sk(sk);
1483 	int val;
1484 	int rc;
1485 
1486 	if (level != SOL_IUCV)
1487 		return -ENOPROTOOPT;
1488 
1489 	if (optlen < sizeof(int))
1490 		return -EINVAL;
1491 
1492 	if (copy_from_sockptr(&val, optval, sizeof(int)))
1493 		return -EFAULT;
1494 
1495 	rc = 0;
1496 
1497 	lock_sock(sk);
1498 	switch (optname) {
1499 	case SO_IPRMDATA_MSG:
1500 		if (val)
1501 			iucv->flags |= IUCV_IPRMDATA;
1502 		else
1503 			iucv->flags &= ~IUCV_IPRMDATA;
1504 		break;
1505 	case SO_MSGLIMIT:
1506 		switch (sk->sk_state) {
1507 		case IUCV_OPEN:
1508 		case IUCV_BOUND:
1509 			if (val < 1 || val > U16_MAX)
1510 				rc = -EINVAL;
1511 			else
1512 				iucv->msglimit = val;
1513 			break;
1514 		default:
1515 			rc = -EINVAL;
1516 			break;
1517 		}
1518 		break;
1519 	default:
1520 		rc = -ENOPROTOOPT;
1521 		break;
1522 	}
1523 	release_sock(sk);
1524 
1525 	return rc;
1526 }
1527 
1528 static int iucv_sock_getsockopt(struct socket *sock, int level, int optname,
1529 				char __user *optval, int __user *optlen)
1530 {
1531 	struct sock *sk = sock->sk;
1532 	struct iucv_sock *iucv = iucv_sk(sk);
1533 	unsigned int val;
1534 	int len;
1535 
1536 	if (level != SOL_IUCV)
1537 		return -ENOPROTOOPT;
1538 
1539 	if (get_user(len, optlen))
1540 		return -EFAULT;
1541 
1542 	if (len < 0)
1543 		return -EINVAL;
1544 
1545 	len = min_t(unsigned int, len, sizeof(int));
1546 
1547 	switch (optname) {
1548 	case SO_IPRMDATA_MSG:
1549 		val = (iucv->flags & IUCV_IPRMDATA) ? 1 : 0;
1550 		break;
1551 	case SO_MSGLIMIT:
1552 		lock_sock(sk);
1553 		val = (iucv->path != NULL) ? iucv->path->msglim	/* connected */
1554 					   : iucv->msglimit;	/* default */
1555 		release_sock(sk);
1556 		break;
1557 	case SO_MSGSIZE:
1558 		if (sk->sk_state == IUCV_OPEN)
1559 			return -EBADFD;
1560 		val = (iucv->hs_dev) ? iucv->hs_dev->mtu -
1561 				sizeof(struct af_iucv_trans_hdr) - ETH_HLEN :
1562 				0x7fffffff;
1563 		break;
1564 	default:
1565 		return -ENOPROTOOPT;
1566 	}
1567 
1568 	if (put_user(len, optlen))
1569 		return -EFAULT;
1570 	if (copy_to_user(optval, &val, len))
1571 		return -EFAULT;
1572 
1573 	return 0;
1574 }
1575 
1576 
1577 /* Callback wrappers - called from iucv base support */
1578 static int iucv_callback_connreq(struct iucv_path *path,
1579 				 u8 ipvmid[8], u8 ipuser[16])
1580 {
1581 	unsigned char user_data[16];
1582 	unsigned char nuser_data[16];
1583 	unsigned char src_name[8];
1584 	struct sock *sk, *nsk;
1585 	struct iucv_sock *iucv, *niucv;
1586 	int err;
1587 
1588 	memcpy(src_name, ipuser, 8);
1589 	EBCASC(src_name, 8);
1590 	/* Find out if this path belongs to af_iucv. */
1591 	read_lock(&iucv_sk_list.lock);
1592 	iucv = NULL;
1593 	sk = NULL;
1594 	sk_for_each(sk, &iucv_sk_list.head)
1595 		if (sk->sk_state == IUCV_LISTEN &&
1596 		    !memcmp(&iucv_sk(sk)->src_name, src_name, 8)) {
1597 			/*
1598 			 * Found a listening socket with
1599 			 * src_name == ipuser[0-7].
1600 			 */
1601 			iucv = iucv_sk(sk);
1602 			break;
1603 		}
1604 	read_unlock(&iucv_sk_list.lock);
1605 	if (!iucv)
1606 		/* No socket found, not one of our paths. */
1607 		return -EINVAL;
1608 
1609 	bh_lock_sock(sk);
1610 
1611 	/* Check if parent socket is listening */
1612 	low_nmcpy(user_data, iucv->src_name);
1613 	high_nmcpy(user_data, iucv->dst_name);
1614 	ASCEBC(user_data, sizeof(user_data));
1615 	if (sk->sk_state != IUCV_LISTEN) {
1616 		err = pr_iucv->path_sever(path, user_data);
1617 		iucv_path_free(path);
1618 		goto fail;
1619 	}
1620 
1621 	/* Check for backlog size */
1622 	if (sk_acceptq_is_full(sk)) {
1623 		err = pr_iucv->path_sever(path, user_data);
1624 		iucv_path_free(path);
1625 		goto fail;
1626 	}
1627 
1628 	/* Create the new socket */
1629 	nsk = iucv_sock_alloc(NULL, sk->sk_protocol, GFP_ATOMIC, 0);
1630 	if (!nsk) {
1631 		err = pr_iucv->path_sever(path, user_data);
1632 		iucv_path_free(path);
1633 		goto fail;
1634 	}
1635 
1636 	niucv = iucv_sk(nsk);
1637 	iucv_sock_init(nsk, sk);
1638 	niucv->transport = AF_IUCV_TRANS_IUCV;
1639 	nsk->sk_allocation |= GFP_DMA;
1640 
1641 	/* Set the new iucv_sock */
1642 	memcpy(niucv->dst_name, ipuser + 8, 8);
1643 	EBCASC(niucv->dst_name, 8);
1644 	memcpy(niucv->dst_user_id, ipvmid, 8);
1645 	memcpy(niucv->src_name, iucv->src_name, 8);
1646 	memcpy(niucv->src_user_id, iucv->src_user_id, 8);
1647 	niucv->path = path;
1648 
1649 	/* Call iucv_accept */
1650 	high_nmcpy(nuser_data, ipuser + 8);
1651 	memcpy(nuser_data + 8, niucv->src_name, 8);
1652 	ASCEBC(nuser_data + 8, 8);
1653 
1654 	/* set message limit for path based on msglimit of accepting socket */
1655 	niucv->msglimit = iucv->msglimit;
1656 	path->msglim = iucv->msglimit;
1657 	err = pr_iucv->path_accept(path, &af_iucv_handler, nuser_data, nsk);
1658 	if (err) {
1659 		iucv_sever_path(nsk, 1);
1660 		iucv_sock_kill(nsk);
1661 		goto fail;
1662 	}
1663 
1664 	iucv_accept_enqueue(sk, nsk);
1665 
1666 	/* Wake up accept */
1667 	nsk->sk_state = IUCV_CONNECTED;
1668 	sk->sk_data_ready(sk);
1669 	err = 0;
1670 fail:
1671 	bh_unlock_sock(sk);
1672 	return 0;
1673 }
1674 
1675 static void iucv_callback_connack(struct iucv_path *path, u8 ipuser[16])
1676 {
1677 	struct sock *sk = path->private;
1678 
1679 	sk->sk_state = IUCV_CONNECTED;
1680 	sk->sk_state_change(sk);
1681 }
1682 
1683 static void iucv_callback_rx(struct iucv_path *path, struct iucv_message *msg)
1684 {
1685 	struct sock *sk = path->private;
1686 	struct iucv_sock *iucv = iucv_sk(sk);
1687 	struct sk_buff *skb;
1688 	struct sock_msg_q *save_msg;
1689 	int len;
1690 
1691 	if (sk->sk_shutdown & RCV_SHUTDOWN) {
1692 		pr_iucv->message_reject(path, msg);
1693 		return;
1694 	}
1695 
1696 	spin_lock(&iucv->message_q.lock);
1697 
1698 	if (!list_empty(&iucv->message_q.list) ||
1699 	    !skb_queue_empty(&iucv->backlog_skb_q))
1700 		goto save_message;
1701 
1702 	len = atomic_read(&sk->sk_rmem_alloc);
1703 	len += SKB_TRUESIZE(iucv_msg_length(msg));
1704 	if (len > sk->sk_rcvbuf)
1705 		goto save_message;
1706 
1707 	skb = alloc_iucv_recv_skb(iucv_msg_length(msg));
1708 	if (!skb)
1709 		goto save_message;
1710 
1711 	iucv_process_message(sk, skb, path, msg);
1712 	goto out_unlock;
1713 
1714 save_message:
1715 	save_msg = kzalloc(sizeof(struct sock_msg_q), GFP_ATOMIC | GFP_DMA);
1716 	if (!save_msg)
1717 		goto out_unlock;
1718 	save_msg->path = path;
1719 	save_msg->msg = *msg;
1720 
1721 	list_add_tail(&save_msg->list, &iucv->message_q.list);
1722 
1723 out_unlock:
1724 	spin_unlock(&iucv->message_q.lock);
1725 }
1726 
1727 static void iucv_callback_txdone(struct iucv_path *path,
1728 				 struct iucv_message *msg)
1729 {
1730 	struct sock *sk = path->private;
1731 	struct sk_buff *this = NULL;
1732 	struct sk_buff_head *list;
1733 	struct sk_buff *list_skb;
1734 	struct iucv_sock *iucv;
1735 	unsigned long flags;
1736 
1737 	iucv = iucv_sk(sk);
1738 	list = &iucv->send_skb_q;
1739 
1740 	bh_lock_sock(sk);
1741 
1742 	spin_lock_irqsave(&list->lock, flags);
1743 	skb_queue_walk(list, list_skb) {
1744 		if (msg->tag == IUCV_SKB_CB(list_skb)->tag) {
1745 			this = list_skb;
1746 			break;
1747 		}
1748 	}
1749 	if (this) {
1750 		atomic_dec(&iucv->skbs_in_xmit);
1751 		__skb_unlink(this, list);
1752 	}
1753 
1754 	spin_unlock_irqrestore(&list->lock, flags);
1755 
1756 	if (this) {
1757 		consume_skb(this);
1758 		/* wake up any process waiting for sending */
1759 		iucv_sock_wake_msglim(sk);
1760 	}
1761 
1762 	if (sk->sk_state == IUCV_CLOSING) {
1763 		if (atomic_read(&iucv->skbs_in_xmit) == 0) {
1764 			sk->sk_state = IUCV_CLOSED;
1765 			sk->sk_state_change(sk);
1766 		}
1767 	}
1768 	bh_unlock_sock(sk);
1769 
1770 }
1771 
1772 static void iucv_callback_connrej(struct iucv_path *path, u8 ipuser[16])
1773 {
1774 	struct sock *sk = path->private;
1775 
1776 	if (sk->sk_state == IUCV_CLOSED)
1777 		return;
1778 
1779 	bh_lock_sock(sk);
1780 	iucv_sever_path(sk, 1);
1781 	sk->sk_state = IUCV_DISCONN;
1782 
1783 	sk->sk_state_change(sk);
1784 	bh_unlock_sock(sk);
1785 }
1786 
1787 /* called if the other communication side shuts down its RECV direction;
1788  * in turn, the callback sets SEND_SHUTDOWN to disable sending of data.
1789  */
1790 static void iucv_callback_shutdown(struct iucv_path *path, u8 ipuser[16])
1791 {
1792 	struct sock *sk = path->private;
1793 
1794 	bh_lock_sock(sk);
1795 	if (sk->sk_state != IUCV_CLOSED) {
1796 		sk->sk_shutdown |= SEND_SHUTDOWN;
1797 		sk->sk_state_change(sk);
1798 	}
1799 	bh_unlock_sock(sk);
1800 }
1801 
1802 static struct iucv_handler af_iucv_handler = {
1803 	.path_pending		= iucv_callback_connreq,
1804 	.path_complete		= iucv_callback_connack,
1805 	.path_severed		= iucv_callback_connrej,
1806 	.message_pending	= iucv_callback_rx,
1807 	.message_complete	= iucv_callback_txdone,
1808 	.path_quiesced		= iucv_callback_shutdown,
1809 };
1810 
1811 /***************** HiperSockets transport callbacks ********************/
1812 static void afiucv_swap_src_dest(struct sk_buff *skb)
1813 {
1814 	struct af_iucv_trans_hdr *trans_hdr = iucv_trans_hdr(skb);
1815 	char tmpID[8];
1816 	char tmpName[8];
1817 
1818 	ASCEBC(trans_hdr->destUserID, sizeof(trans_hdr->destUserID));
1819 	ASCEBC(trans_hdr->destAppName, sizeof(trans_hdr->destAppName));
1820 	ASCEBC(trans_hdr->srcUserID, sizeof(trans_hdr->srcUserID));
1821 	ASCEBC(trans_hdr->srcAppName, sizeof(trans_hdr->srcAppName));
1822 	memcpy(tmpID, trans_hdr->srcUserID, 8);
1823 	memcpy(tmpName, trans_hdr->srcAppName, 8);
1824 	memcpy(trans_hdr->srcUserID, trans_hdr->destUserID, 8);
1825 	memcpy(trans_hdr->srcAppName, trans_hdr->destAppName, 8);
1826 	memcpy(trans_hdr->destUserID, tmpID, 8);
1827 	memcpy(trans_hdr->destAppName, tmpName, 8);
1828 	skb_push(skb, ETH_HLEN);
1829 	memset(skb->data, 0, ETH_HLEN);
1830 }
1831 
1832 /*
1833  * afiucv_hs_callback_syn - react on received SYN
1834  */
1835 static int afiucv_hs_callback_syn(struct sock *sk, struct sk_buff *skb)
1836 {
1837 	struct af_iucv_trans_hdr *trans_hdr = iucv_trans_hdr(skb);
1838 	struct sock *nsk;
1839 	struct iucv_sock *iucv, *niucv;
1840 	int err;
1841 
1842 	iucv = iucv_sk(sk);
1843 	if (!iucv) {
1844 		/* no sock - connection refused */
1845 		afiucv_swap_src_dest(skb);
1846 		trans_hdr->flags = AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_FIN;
1847 		err = dev_queue_xmit(skb);
1848 		goto out;
1849 	}
1850 
1851 	nsk = iucv_sock_alloc(NULL, sk->sk_protocol, GFP_ATOMIC, 0);
1852 	bh_lock_sock(sk);
1853 	if ((sk->sk_state != IUCV_LISTEN) ||
1854 	    sk_acceptq_is_full(sk) ||
1855 	    !nsk) {
1856 		/* error on server socket - connection refused */
1857 		afiucv_swap_src_dest(skb);
1858 		trans_hdr->flags = AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_FIN;
1859 		err = dev_queue_xmit(skb);
1860 		iucv_sock_kill(nsk);
1861 		bh_unlock_sock(sk);
1862 		goto out;
1863 	}
1864 
1865 	niucv = iucv_sk(nsk);
1866 	iucv_sock_init(nsk, sk);
1867 	niucv->transport = AF_IUCV_TRANS_HIPER;
1868 	niucv->msglimit = iucv->msglimit;
1869 	if (!trans_hdr->window)
1870 		niucv->msglimit_peer = IUCV_HIPER_MSGLIM_DEFAULT;
1871 	else
1872 		niucv->msglimit_peer = trans_hdr->window;
1873 	memcpy(niucv->dst_name, trans_hdr->srcAppName, 8);
1874 	memcpy(niucv->dst_user_id, trans_hdr->srcUserID, 8);
1875 	memcpy(niucv->src_name, iucv->src_name, 8);
1876 	memcpy(niucv->src_user_id, iucv->src_user_id, 8);
1877 	nsk->sk_bound_dev_if = sk->sk_bound_dev_if;
1878 	niucv->hs_dev = iucv->hs_dev;
1879 	dev_hold(niucv->hs_dev);
1880 	afiucv_swap_src_dest(skb);
1881 	trans_hdr->flags = AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_ACK;
1882 	trans_hdr->window = niucv->msglimit;
1883 	/* if receiver acks the xmit connection is established */
1884 	err = dev_queue_xmit(skb);
1885 	if (!err) {
1886 		iucv_accept_enqueue(sk, nsk);
1887 		nsk->sk_state = IUCV_CONNECTED;
1888 		sk->sk_data_ready(sk);
1889 	} else
1890 		iucv_sock_kill(nsk);
1891 	bh_unlock_sock(sk);
1892 
1893 out:
1894 	return NET_RX_SUCCESS;
1895 }
1896 
1897 /*
1898  * afiucv_hs_callback_synack() - react on received SYN-ACK
1899  */
1900 static int afiucv_hs_callback_synack(struct sock *sk, struct sk_buff *skb)
1901 {
1902 	struct iucv_sock *iucv = iucv_sk(sk);
1903 
1904 	if (!iucv || sk->sk_state != IUCV_BOUND) {
1905 		kfree_skb(skb);
1906 		return NET_RX_SUCCESS;
1907 	}
1908 
1909 	bh_lock_sock(sk);
1910 	iucv->msglimit_peer = iucv_trans_hdr(skb)->window;
1911 	sk->sk_state = IUCV_CONNECTED;
1912 	sk->sk_state_change(sk);
1913 	bh_unlock_sock(sk);
1914 	consume_skb(skb);
1915 	return NET_RX_SUCCESS;
1916 }
1917 
1918 /*
1919  * afiucv_hs_callback_synfin() - react on received SYN_FIN
1920  */
1921 static int afiucv_hs_callback_synfin(struct sock *sk, struct sk_buff *skb)
1922 {
1923 	struct iucv_sock *iucv = iucv_sk(sk);
1924 
1925 	if (!iucv || sk->sk_state != IUCV_BOUND) {
1926 		kfree_skb(skb);
1927 		return NET_RX_SUCCESS;
1928 	}
1929 
1930 	bh_lock_sock(sk);
1931 	sk->sk_state = IUCV_DISCONN;
1932 	sk->sk_state_change(sk);
1933 	bh_unlock_sock(sk);
1934 	consume_skb(skb);
1935 	return NET_RX_SUCCESS;
1936 }
1937 
1938 /*
1939  * afiucv_hs_callback_fin() - react on received FIN
1940  */
1941 static int afiucv_hs_callback_fin(struct sock *sk, struct sk_buff *skb)
1942 {
1943 	struct iucv_sock *iucv = iucv_sk(sk);
1944 
1945 	/* other end of connection closed */
1946 	if (!iucv) {
1947 		kfree_skb(skb);
1948 		return NET_RX_SUCCESS;
1949 	}
1950 
1951 	bh_lock_sock(sk);
1952 	if (sk->sk_state == IUCV_CONNECTED) {
1953 		sk->sk_state = IUCV_DISCONN;
1954 		sk->sk_state_change(sk);
1955 	}
1956 	bh_unlock_sock(sk);
1957 	consume_skb(skb);
1958 	return NET_RX_SUCCESS;
1959 }
1960 
1961 /*
1962  * afiucv_hs_callback_win() - react on received WIN
1963  */
1964 static int afiucv_hs_callback_win(struct sock *sk, struct sk_buff *skb)
1965 {
1966 	struct iucv_sock *iucv = iucv_sk(sk);
1967 
1968 	if (!iucv)
1969 		return NET_RX_SUCCESS;
1970 
1971 	if (sk->sk_state != IUCV_CONNECTED)
1972 		return NET_RX_SUCCESS;
1973 
1974 	atomic_sub(iucv_trans_hdr(skb)->window, &iucv->msg_sent);
1975 	iucv_sock_wake_msglim(sk);
1976 	return NET_RX_SUCCESS;
1977 }
1978 
1979 /*
1980  * afiucv_hs_callback_rx() - react on received data
1981  */
1982 static int afiucv_hs_callback_rx(struct sock *sk, struct sk_buff *skb)
1983 {
1984 	struct iucv_sock *iucv = iucv_sk(sk);
1985 
1986 	if (!iucv) {
1987 		kfree_skb(skb);
1988 		return NET_RX_SUCCESS;
1989 	}
1990 
1991 	if (sk->sk_state != IUCV_CONNECTED) {
1992 		kfree_skb(skb);
1993 		return NET_RX_SUCCESS;
1994 	}
1995 
1996 	if (sk->sk_shutdown & RCV_SHUTDOWN) {
1997 		kfree_skb(skb);
1998 		return NET_RX_SUCCESS;
1999 	}
2000 
2001 	/* write stuff from iucv_msg to skb cb */
2002 	skb_pull(skb, sizeof(struct af_iucv_trans_hdr));
2003 	skb_reset_transport_header(skb);
2004 	skb_reset_network_header(skb);
2005 	IUCV_SKB_CB(skb)->offset = 0;
2006 	if (sk_filter(sk, skb)) {
2007 		atomic_inc(&sk->sk_drops);	/* skb rejected by filter */
2008 		kfree_skb(skb);
2009 		return NET_RX_SUCCESS;
2010 	}
2011 
2012 	spin_lock(&iucv->message_q.lock);
2013 	if (skb_queue_empty(&iucv->backlog_skb_q)) {
2014 		if (__sock_queue_rcv_skb(sk, skb))
2015 			/* handle rcv queue full */
2016 			skb_queue_tail(&iucv->backlog_skb_q, skb);
2017 	} else
2018 		skb_queue_tail(&iucv_sk(sk)->backlog_skb_q, skb);
2019 	spin_unlock(&iucv->message_q.lock);
2020 	return NET_RX_SUCCESS;
2021 }
2022 
2023 /*
2024  * afiucv_hs_rcv() - base function for arriving data through HiperSockets
2025  *                   transport
2026  *                   called from netif RX softirq
2027  */
2028 static int afiucv_hs_rcv(struct sk_buff *skb, struct net_device *dev,
2029 	struct packet_type *pt, struct net_device *orig_dev)
2030 {
2031 	struct sock *sk;
2032 	struct iucv_sock *iucv;
2033 	struct af_iucv_trans_hdr *trans_hdr;
2034 	int err = NET_RX_SUCCESS;
2035 	char nullstring[8];
2036 
2037 	if (!pskb_may_pull(skb, sizeof(*trans_hdr))) {
2038 		kfree_skb(skb);
2039 		return NET_RX_SUCCESS;
2040 	}
2041 
2042 	trans_hdr = iucv_trans_hdr(skb);
2043 	EBCASC(trans_hdr->destAppName, sizeof(trans_hdr->destAppName));
2044 	EBCASC(trans_hdr->destUserID, sizeof(trans_hdr->destUserID));
2045 	EBCASC(trans_hdr->srcAppName, sizeof(trans_hdr->srcAppName));
2046 	EBCASC(trans_hdr->srcUserID, sizeof(trans_hdr->srcUserID));
2047 	memset(nullstring, 0, sizeof(nullstring));
2048 	iucv = NULL;
2049 	sk = NULL;
2050 	read_lock(&iucv_sk_list.lock);
2051 	sk_for_each(sk, &iucv_sk_list.head) {
2052 		if (trans_hdr->flags == AF_IUCV_FLAG_SYN) {
2053 			if ((!memcmp(&iucv_sk(sk)->src_name,
2054 				     trans_hdr->destAppName, 8)) &&
2055 			    (!memcmp(&iucv_sk(sk)->src_user_id,
2056 				     trans_hdr->destUserID, 8)) &&
2057 			    (!memcmp(&iucv_sk(sk)->dst_name, nullstring, 8)) &&
2058 			    (!memcmp(&iucv_sk(sk)->dst_user_id,
2059 				     nullstring, 8))) {
2060 				iucv = iucv_sk(sk);
2061 				break;
2062 			}
2063 		} else {
2064 			if ((!memcmp(&iucv_sk(sk)->src_name,
2065 				     trans_hdr->destAppName, 8)) &&
2066 			    (!memcmp(&iucv_sk(sk)->src_user_id,
2067 				     trans_hdr->destUserID, 8)) &&
2068 			    (!memcmp(&iucv_sk(sk)->dst_name,
2069 				     trans_hdr->srcAppName, 8)) &&
2070 			    (!memcmp(&iucv_sk(sk)->dst_user_id,
2071 				     trans_hdr->srcUserID, 8))) {
2072 				iucv = iucv_sk(sk);
2073 				break;
2074 			}
2075 		}
2076 	}
2077 	read_unlock(&iucv_sk_list.lock);
2078 	if (!iucv)
2079 		sk = NULL;
2080 
2081 	/* no sock
2082 	how should we send with no sock
2083 	1) send without sock no send rc checking?
2084 	2) introduce default sock to handle this cases
2085 
2086 	 SYN -> send SYN|ACK in good case, send SYN|FIN in bad case
2087 	 data -> send FIN
2088 	 SYN|ACK, SYN|FIN, FIN -> no action? */
2089 
2090 	switch (trans_hdr->flags) {
2091 	case AF_IUCV_FLAG_SYN:
2092 		/* connect request */
2093 		err = afiucv_hs_callback_syn(sk, skb);
2094 		break;
2095 	case (AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_ACK):
2096 		/* connect request confirmed */
2097 		err = afiucv_hs_callback_synack(sk, skb);
2098 		break;
2099 	case (AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_FIN):
2100 		/* connect request refused */
2101 		err = afiucv_hs_callback_synfin(sk, skb);
2102 		break;
2103 	case (AF_IUCV_FLAG_FIN):
2104 		/* close request */
2105 		err = afiucv_hs_callback_fin(sk, skb);
2106 		break;
2107 	case (AF_IUCV_FLAG_WIN):
2108 		err = afiucv_hs_callback_win(sk, skb);
2109 		if (skb->len == sizeof(struct af_iucv_trans_hdr)) {
2110 			consume_skb(skb);
2111 			break;
2112 		}
2113 		fallthrough;	/* and receive non-zero length data */
2114 	case (AF_IUCV_FLAG_SHT):
2115 		/* shutdown request */
2116 		fallthrough;	/* and receive zero length data */
2117 	case 0:
2118 		/* plain data frame */
2119 		IUCV_SKB_CB(skb)->class = trans_hdr->iucv_hdr.class;
2120 		err = afiucv_hs_callback_rx(sk, skb);
2121 		break;
2122 	default:
2123 		kfree_skb(skb);
2124 	}
2125 
2126 	return err;
2127 }
2128 
2129 /*
2130  * afiucv_hs_callback_txnotify() - handle send notifications from HiperSockets
2131  *                                 transport
2132  */
2133 static void afiucv_hs_callback_txnotify(struct sock *sk, enum iucv_tx_notify n)
2134 {
2135 	struct iucv_sock *iucv = iucv_sk(sk);
2136 
2137 	if (sock_flag(sk, SOCK_ZAPPED))
2138 		return;
2139 
2140 	switch (n) {
2141 	case TX_NOTIFY_OK:
2142 		atomic_dec(&iucv->skbs_in_xmit);
2143 		iucv_sock_wake_msglim(sk);
2144 		break;
2145 	case TX_NOTIFY_PENDING:
2146 		atomic_inc(&iucv->pendings);
2147 		break;
2148 	case TX_NOTIFY_DELAYED_OK:
2149 		atomic_dec(&iucv->skbs_in_xmit);
2150 		if (atomic_dec_return(&iucv->pendings) <= 0)
2151 			iucv_sock_wake_msglim(sk);
2152 		break;
2153 	default:
2154 		atomic_dec(&iucv->skbs_in_xmit);
2155 		if (sk->sk_state == IUCV_CONNECTED) {
2156 			sk->sk_state = IUCV_DISCONN;
2157 			sk->sk_state_change(sk);
2158 		}
2159 	}
2160 
2161 	if (sk->sk_state == IUCV_CLOSING) {
2162 		if (atomic_read(&iucv->skbs_in_xmit) == 0) {
2163 			sk->sk_state = IUCV_CLOSED;
2164 			sk->sk_state_change(sk);
2165 		}
2166 	}
2167 }
2168 
2169 /*
2170  * afiucv_netdev_event: handle netdev notifier chain events
2171  */
2172 static int afiucv_netdev_event(struct notifier_block *this,
2173 			       unsigned long event, void *ptr)
2174 {
2175 	struct net_device *event_dev = netdev_notifier_info_to_dev(ptr);
2176 	struct sock *sk;
2177 	struct iucv_sock *iucv;
2178 
2179 	switch (event) {
2180 	case NETDEV_REBOOT:
2181 	case NETDEV_GOING_DOWN:
2182 		sk_for_each(sk, &iucv_sk_list.head) {
2183 			iucv = iucv_sk(sk);
2184 			if ((iucv->hs_dev == event_dev) &&
2185 			    (sk->sk_state == IUCV_CONNECTED)) {
2186 				if (event == NETDEV_GOING_DOWN)
2187 					iucv_send_ctrl(sk, AF_IUCV_FLAG_FIN);
2188 				sk->sk_state = IUCV_DISCONN;
2189 				sk->sk_state_change(sk);
2190 			}
2191 		}
2192 		break;
2193 	case NETDEV_DOWN:
2194 	case NETDEV_UNREGISTER:
2195 	default:
2196 		break;
2197 	}
2198 	return NOTIFY_DONE;
2199 }
2200 
2201 static struct notifier_block afiucv_netdev_notifier = {
2202 	.notifier_call = afiucv_netdev_event,
2203 };
2204 
2205 static const struct proto_ops iucv_sock_ops = {
2206 	.family		= PF_IUCV,
2207 	.owner		= THIS_MODULE,
2208 	.release	= iucv_sock_release,
2209 	.bind		= iucv_sock_bind,
2210 	.connect	= iucv_sock_connect,
2211 	.listen		= iucv_sock_listen,
2212 	.accept		= iucv_sock_accept,
2213 	.getname	= iucv_sock_getname,
2214 	.sendmsg	= iucv_sock_sendmsg,
2215 	.recvmsg	= iucv_sock_recvmsg,
2216 	.poll		= iucv_sock_poll,
2217 	.ioctl		= sock_no_ioctl,
2218 	.mmap		= sock_no_mmap,
2219 	.socketpair	= sock_no_socketpair,
2220 	.shutdown	= iucv_sock_shutdown,
2221 	.setsockopt	= iucv_sock_setsockopt,
2222 	.getsockopt	= iucv_sock_getsockopt,
2223 };
2224 
2225 static int iucv_sock_create(struct net *net, struct socket *sock, int protocol,
2226 			    int kern)
2227 {
2228 	struct sock *sk;
2229 
2230 	if (protocol && protocol != PF_IUCV)
2231 		return -EPROTONOSUPPORT;
2232 
2233 	sock->state = SS_UNCONNECTED;
2234 
2235 	switch (sock->type) {
2236 	case SOCK_STREAM:
2237 	case SOCK_SEQPACKET:
2238 		/* currently, proto ops can handle both sk types */
2239 		sock->ops = &iucv_sock_ops;
2240 		break;
2241 	default:
2242 		return -ESOCKTNOSUPPORT;
2243 	}
2244 
2245 	sk = iucv_sock_alloc(sock, protocol, GFP_KERNEL, kern);
2246 	if (!sk)
2247 		return -ENOMEM;
2248 
2249 	iucv_sock_init(sk, NULL);
2250 
2251 	return 0;
2252 }
2253 
2254 static const struct net_proto_family iucv_sock_family_ops = {
2255 	.family	= AF_IUCV,
2256 	.owner	= THIS_MODULE,
2257 	.create	= iucv_sock_create,
2258 };
2259 
2260 static struct packet_type iucv_packet_type = {
2261 	.type = cpu_to_be16(ETH_P_AF_IUCV),
2262 	.func = afiucv_hs_rcv,
2263 };
2264 
2265 static int __init afiucv_init(void)
2266 {
2267 	int err;
2268 
2269 	if (MACHINE_IS_VM && IS_ENABLED(CONFIG_IUCV)) {
2270 		cpcmd("QUERY USERID", iucv_userid, sizeof(iucv_userid), &err);
2271 		if (unlikely(err)) {
2272 			WARN_ON(err);
2273 			err = -EPROTONOSUPPORT;
2274 			goto out;
2275 		}
2276 
2277 		pr_iucv = &iucv_if;
2278 	} else {
2279 		memset(&iucv_userid, 0, sizeof(iucv_userid));
2280 		pr_iucv = NULL;
2281 	}
2282 
2283 	err = proto_register(&iucv_proto, 0);
2284 	if (err)
2285 		goto out;
2286 	err = sock_register(&iucv_sock_family_ops);
2287 	if (err)
2288 		goto out_proto;
2289 
2290 	if (pr_iucv) {
2291 		err = pr_iucv->iucv_register(&af_iucv_handler, 0);
2292 		if (err)
2293 			goto out_sock;
2294 	}
2295 
2296 	err = register_netdevice_notifier(&afiucv_netdev_notifier);
2297 	if (err)
2298 		goto out_notifier;
2299 
2300 	dev_add_pack(&iucv_packet_type);
2301 	return 0;
2302 
2303 out_notifier:
2304 	if (pr_iucv)
2305 		pr_iucv->iucv_unregister(&af_iucv_handler, 0);
2306 out_sock:
2307 	sock_unregister(PF_IUCV);
2308 out_proto:
2309 	proto_unregister(&iucv_proto);
2310 out:
2311 	return err;
2312 }
2313 
2314 static void __exit afiucv_exit(void)
2315 {
2316 	if (pr_iucv)
2317 		pr_iucv->iucv_unregister(&af_iucv_handler, 0);
2318 
2319 	unregister_netdevice_notifier(&afiucv_netdev_notifier);
2320 	dev_remove_pack(&iucv_packet_type);
2321 	sock_unregister(PF_IUCV);
2322 	proto_unregister(&iucv_proto);
2323 }
2324 
2325 module_init(afiucv_init);
2326 module_exit(afiucv_exit);
2327 
2328 MODULE_AUTHOR("Jennifer Hunt <jenhunt@us.ibm.com>");
2329 MODULE_DESCRIPTION("IUCV Sockets ver " VERSION);
2330 MODULE_VERSION(VERSION);
2331 MODULE_LICENSE("GPL");
2332 MODULE_ALIAS_NETPROTO(PF_IUCV);
2333