xref: /openbmc/linux/net/iucv/af_iucv.c (revision caa80275)
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/module.h>
17 #include <linux/netdevice.h>
18 #include <linux/types.h>
19 #include <linux/limits.h>
20 #include <linux/list.h>
21 #include <linux/errno.h>
22 #include <linux/kernel.h>
23 #include <linux/sched/signal.h>
24 #include <linux/slab.h>
25 #include <linux/skbuff.h>
26 #include <linux/init.h>
27 #include <linux/poll.h>
28 #include <linux/security.h>
29 #include <net/sock.h>
30 #include <asm/ebcdic.h>
31 #include <asm/cpcmd.h>
32 #include <linux/kmod.h>
33 
34 #include <net/iucv/af_iucv.h>
35 
36 #define VERSION "1.2"
37 
38 static char iucv_userid[80];
39 
40 static struct proto iucv_proto = {
41 	.name		= "AF_IUCV",
42 	.owner		= THIS_MODULE,
43 	.obj_size	= sizeof(struct iucv_sock),
44 };
45 
46 static struct iucv_interface *pr_iucv;
47 static struct iucv_handler af_iucv_handler;
48 
49 /* special AF_IUCV IPRM messages */
50 static const u8 iprm_shutdown[8] =
51 	{0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01};
52 
53 #define TRGCLS_SIZE	sizeof_field(struct iucv_message, class)
54 
55 #define __iucv_sock_wait(sk, condition, timeo, ret)			\
56 do {									\
57 	DEFINE_WAIT(__wait);						\
58 	long __timeo = timeo;						\
59 	ret = 0;							\
60 	prepare_to_wait(sk_sleep(sk), &__wait, TASK_INTERRUPTIBLE);	\
61 	while (!(condition)) {						\
62 		if (!__timeo) {						\
63 			ret = -EAGAIN;					\
64 			break;						\
65 		}							\
66 		if (signal_pending(current)) {				\
67 			ret = sock_intr_errno(__timeo);			\
68 			break;						\
69 		}							\
70 		release_sock(sk);					\
71 		__timeo = schedule_timeout(__timeo);			\
72 		lock_sock(sk);						\
73 		ret = sock_error(sk);					\
74 		if (ret)						\
75 			break;						\
76 	}								\
77 	finish_wait(sk_sleep(sk), &__wait);				\
78 } while (0)
79 
80 #define iucv_sock_wait(sk, condition, timeo)				\
81 ({									\
82 	int __ret = 0;							\
83 	if (!(condition))						\
84 		__iucv_sock_wait(sk, condition, timeo, __ret);		\
85 	__ret;								\
86 })
87 
88 static struct sock *iucv_accept_dequeue(struct sock *parent,
89 					struct socket *newsock);
90 static void iucv_sock_kill(struct sock *sk);
91 static void iucv_sock_close(struct sock *sk);
92 
93 static void afiucv_hs_callback_txnotify(struct sock *sk, enum iucv_tx_notify);
94 
95 static struct iucv_sock_list iucv_sk_list = {
96 	.lock = __RW_LOCK_UNLOCKED(iucv_sk_list.lock),
97 	.autobind_name = ATOMIC_INIT(0)
98 };
99 
100 static inline void high_nmcpy(unsigned char *dst, char *src)
101 {
102        memcpy(dst, src, 8);
103 }
104 
105 static inline void low_nmcpy(unsigned char *dst, char *src)
106 {
107        memcpy(&dst[8], src, 8);
108 }
109 
110 /**
111  * iucv_msg_length() - Returns the length of an iucv message.
112  * @msg:	Pointer to struct iucv_message, MUST NOT be NULL
113  *
114  * The function returns the length of the specified iucv message @msg of data
115  * stored in a buffer and of data stored in the parameter list (PRMDATA).
116  *
117  * For IUCV_IPRMDATA, AF_IUCV uses the following convention to transport socket
118  * data:
119  *	PRMDATA[0..6]	socket data (max 7 bytes);
120  *	PRMDATA[7]	socket data length value (len is 0xff - PRMDATA[7])
121  *
122  * The socket data length is computed by subtracting the socket data length
123  * value from 0xFF.
124  * If the socket data len is greater 7, then PRMDATA can be used for special
125  * notifications (see iucv_sock_shutdown); and further,
126  * if the socket data len is > 7, the function returns 8.
127  *
128  * Use this function to allocate socket buffers to store iucv message data.
129  */
130 static inline size_t iucv_msg_length(struct iucv_message *msg)
131 {
132 	size_t datalen;
133 
134 	if (msg->flags & IUCV_IPRMDATA) {
135 		datalen = 0xff - msg->rmmsg[7];
136 		return (datalen < 8) ? datalen : 8;
137 	}
138 	return msg->length;
139 }
140 
141 /**
142  * iucv_sock_in_state() - check for specific states
143  * @sk:		sock structure
144  * @state:	first iucv sk state
145  * @state:	second iucv sk state
146  *
147  * Returns true if the socket in either in the first or second state.
148  */
149 static int iucv_sock_in_state(struct sock *sk, int state, int state2)
150 {
151 	return (sk->sk_state == state || sk->sk_state == state2);
152 }
153 
154 /**
155  * iucv_below_msglim() - function to check if messages can be sent
156  * @sk:		sock structure
157  *
158  * Returns true if the send queue length is lower than the message limit.
159  * Always returns true if the socket is not connected (no iucv path for
160  * checking the message limit).
161  */
162 static inline int iucv_below_msglim(struct sock *sk)
163 {
164 	struct iucv_sock *iucv = iucv_sk(sk);
165 
166 	if (sk->sk_state != IUCV_CONNECTED)
167 		return 1;
168 	if (iucv->transport == AF_IUCV_TRANS_IUCV)
169 		return (atomic_read(&iucv->skbs_in_xmit) < iucv->path->msglim);
170 	else
171 		return ((atomic_read(&iucv->msg_sent) < iucv->msglimit_peer) &&
172 			(atomic_read(&iucv->pendings) <= 0));
173 }
174 
175 /**
176  * iucv_sock_wake_msglim() - Wake up thread waiting on msg limit
177  */
178 static void iucv_sock_wake_msglim(struct sock *sk)
179 {
180 	struct socket_wq *wq;
181 
182 	rcu_read_lock();
183 	wq = rcu_dereference(sk->sk_wq);
184 	if (skwq_has_sleeper(wq))
185 		wake_up_interruptible_all(&wq->wait);
186 	sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
187 	rcu_read_unlock();
188 }
189 
190 /**
191  * afiucv_hs_send() - send a message through HiperSockets transport
192  */
193 static int afiucv_hs_send(struct iucv_message *imsg, struct sock *sock,
194 		   struct sk_buff *skb, u8 flags)
195 {
196 	struct iucv_sock *iucv = iucv_sk(sock);
197 	struct af_iucv_trans_hdr *phs_hdr;
198 	int err, confirm_recv = 0;
199 
200 	phs_hdr = skb_push(skb, sizeof(*phs_hdr));
201 	memset(phs_hdr, 0, sizeof(*phs_hdr));
202 	skb_reset_network_header(skb);
203 
204 	phs_hdr->magic = ETH_P_AF_IUCV;
205 	phs_hdr->version = 1;
206 	phs_hdr->flags = flags;
207 	if (flags == AF_IUCV_FLAG_SYN)
208 		phs_hdr->window = iucv->msglimit;
209 	else if ((flags == AF_IUCV_FLAG_WIN) || !flags) {
210 		confirm_recv = atomic_read(&iucv->msg_recv);
211 		phs_hdr->window = confirm_recv;
212 		if (confirm_recv)
213 			phs_hdr->flags = phs_hdr->flags | AF_IUCV_FLAG_WIN;
214 	}
215 	memcpy(phs_hdr->destUserID, iucv->dst_user_id, 8);
216 	memcpy(phs_hdr->destAppName, iucv->dst_name, 8);
217 	memcpy(phs_hdr->srcUserID, iucv->src_user_id, 8);
218 	memcpy(phs_hdr->srcAppName, iucv->src_name, 8);
219 	ASCEBC(phs_hdr->destUserID, sizeof(phs_hdr->destUserID));
220 	ASCEBC(phs_hdr->destAppName, sizeof(phs_hdr->destAppName));
221 	ASCEBC(phs_hdr->srcUserID, sizeof(phs_hdr->srcUserID));
222 	ASCEBC(phs_hdr->srcAppName, sizeof(phs_hdr->srcAppName));
223 	if (imsg)
224 		memcpy(&phs_hdr->iucv_hdr, imsg, sizeof(struct iucv_message));
225 
226 	skb->dev = iucv->hs_dev;
227 	if (!skb->dev) {
228 		err = -ENODEV;
229 		goto err_free;
230 	}
231 
232 	dev_hard_header(skb, skb->dev, ETH_P_AF_IUCV, NULL, NULL, skb->len);
233 
234 	if (!(skb->dev->flags & IFF_UP) || !netif_carrier_ok(skb->dev)) {
235 		err = -ENETDOWN;
236 		goto err_free;
237 	}
238 	if (skb->len > skb->dev->mtu) {
239 		if (sock->sk_type == SOCK_SEQPACKET) {
240 			err = -EMSGSIZE;
241 			goto err_free;
242 		}
243 		err = pskb_trim(skb, skb->dev->mtu);
244 		if (err)
245 			goto err_free;
246 	}
247 	skb->protocol = cpu_to_be16(ETH_P_AF_IUCV);
248 
249 	atomic_inc(&iucv->skbs_in_xmit);
250 	err = dev_queue_xmit(skb);
251 	if (net_xmit_eval(err)) {
252 		atomic_dec(&iucv->skbs_in_xmit);
253 	} else {
254 		atomic_sub(confirm_recv, &iucv->msg_recv);
255 		WARN_ON(atomic_read(&iucv->msg_recv) < 0);
256 	}
257 	return net_xmit_eval(err);
258 
259 err_free:
260 	kfree_skb(skb);
261 	return err;
262 }
263 
264 static struct sock *__iucv_get_sock_by_name(char *nm)
265 {
266 	struct sock *sk;
267 
268 	sk_for_each(sk, &iucv_sk_list.head)
269 		if (!memcmp(&iucv_sk(sk)->src_name, nm, 8))
270 			return sk;
271 
272 	return NULL;
273 }
274 
275 static void iucv_sock_destruct(struct sock *sk)
276 {
277 	skb_queue_purge(&sk->sk_receive_queue);
278 	skb_queue_purge(&sk->sk_error_queue);
279 
280 	sk_mem_reclaim(sk);
281 
282 	if (!sock_flag(sk, SOCK_DEAD)) {
283 		pr_err("Attempt to release alive iucv socket %p\n", sk);
284 		return;
285 	}
286 
287 	WARN_ON(atomic_read(&sk->sk_rmem_alloc));
288 	WARN_ON(refcount_read(&sk->sk_wmem_alloc));
289 	WARN_ON(sk->sk_wmem_queued);
290 	WARN_ON(sk->sk_forward_alloc);
291 }
292 
293 /* Cleanup Listen */
294 static void iucv_sock_cleanup_listen(struct sock *parent)
295 {
296 	struct sock *sk;
297 
298 	/* Close non-accepted connections */
299 	while ((sk = iucv_accept_dequeue(parent, NULL))) {
300 		iucv_sock_close(sk);
301 		iucv_sock_kill(sk);
302 	}
303 
304 	parent->sk_state = IUCV_CLOSED;
305 }
306 
307 static void iucv_sock_link(struct iucv_sock_list *l, struct sock *sk)
308 {
309 	write_lock_bh(&l->lock);
310 	sk_add_node(sk, &l->head);
311 	write_unlock_bh(&l->lock);
312 }
313 
314 static void iucv_sock_unlink(struct iucv_sock_list *l, struct sock *sk)
315 {
316 	write_lock_bh(&l->lock);
317 	sk_del_node_init(sk);
318 	write_unlock_bh(&l->lock);
319 }
320 
321 /* Kill socket (only if zapped and orphaned) */
322 static void iucv_sock_kill(struct sock *sk)
323 {
324 	if (!sock_flag(sk, SOCK_ZAPPED) || sk->sk_socket)
325 		return;
326 
327 	iucv_sock_unlink(&iucv_sk_list, sk);
328 	sock_set_flag(sk, SOCK_DEAD);
329 	sock_put(sk);
330 }
331 
332 /* Terminate an IUCV path */
333 static void iucv_sever_path(struct sock *sk, int with_user_data)
334 {
335 	unsigned char user_data[16];
336 	struct iucv_sock *iucv = iucv_sk(sk);
337 	struct iucv_path *path = iucv->path;
338 
339 	if (iucv->path) {
340 		iucv->path = NULL;
341 		if (with_user_data) {
342 			low_nmcpy(user_data, iucv->src_name);
343 			high_nmcpy(user_data, iucv->dst_name);
344 			ASCEBC(user_data, sizeof(user_data));
345 			pr_iucv->path_sever(path, user_data);
346 		} else
347 			pr_iucv->path_sever(path, NULL);
348 		iucv_path_free(path);
349 	}
350 }
351 
352 /* Send controlling flags through an IUCV socket for HIPER transport */
353 static int iucv_send_ctrl(struct sock *sk, u8 flags)
354 {
355 	struct iucv_sock *iucv = iucv_sk(sk);
356 	int err = 0;
357 	int blen;
358 	struct sk_buff *skb;
359 	u8 shutdown = 0;
360 
361 	blen = sizeof(struct af_iucv_trans_hdr) +
362 	       LL_RESERVED_SPACE(iucv->hs_dev);
363 	if (sk->sk_shutdown & SEND_SHUTDOWN) {
364 		/* controlling flags should be sent anyway */
365 		shutdown = sk->sk_shutdown;
366 		sk->sk_shutdown &= RCV_SHUTDOWN;
367 	}
368 	skb = sock_alloc_send_skb(sk, blen, 1, &err);
369 	if (skb) {
370 		skb_reserve(skb, blen);
371 		err = afiucv_hs_send(NULL, sk, skb, flags);
372 	}
373 	if (shutdown)
374 		sk->sk_shutdown = shutdown;
375 	return err;
376 }
377 
378 /* Close an IUCV socket */
379 static void iucv_sock_close(struct sock *sk)
380 {
381 	struct iucv_sock *iucv = iucv_sk(sk);
382 	unsigned long timeo;
383 	int err = 0;
384 
385 	lock_sock(sk);
386 
387 	switch (sk->sk_state) {
388 	case IUCV_LISTEN:
389 		iucv_sock_cleanup_listen(sk);
390 		break;
391 
392 	case IUCV_CONNECTED:
393 		if (iucv->transport == AF_IUCV_TRANS_HIPER) {
394 			err = iucv_send_ctrl(sk, AF_IUCV_FLAG_FIN);
395 			sk->sk_state = IUCV_DISCONN;
396 			sk->sk_state_change(sk);
397 		}
398 		fallthrough;
399 
400 	case IUCV_DISCONN:
401 		sk->sk_state = IUCV_CLOSING;
402 		sk->sk_state_change(sk);
403 
404 		if (!err && atomic_read(&iucv->skbs_in_xmit) > 0) {
405 			if (sock_flag(sk, SOCK_LINGER) && sk->sk_lingertime)
406 				timeo = sk->sk_lingertime;
407 			else
408 				timeo = IUCV_DISCONN_TIMEOUT;
409 			iucv_sock_wait(sk,
410 					iucv_sock_in_state(sk, IUCV_CLOSED, 0),
411 					timeo);
412 		}
413 		fallthrough;
414 
415 	case IUCV_CLOSING:
416 		sk->sk_state = IUCV_CLOSED;
417 		sk->sk_state_change(sk);
418 
419 		sk->sk_err = ECONNRESET;
420 		sk->sk_state_change(sk);
421 
422 		skb_queue_purge(&iucv->send_skb_q);
423 		skb_queue_purge(&iucv->backlog_skb_q);
424 		fallthrough;
425 
426 	default:
427 		iucv_sever_path(sk, 1);
428 	}
429 
430 	if (iucv->hs_dev) {
431 		dev_put(iucv->hs_dev);
432 		iucv->hs_dev = NULL;
433 		sk->sk_bound_dev_if = 0;
434 	}
435 
436 	/* mark socket for deletion by iucv_sock_kill() */
437 	sock_set_flag(sk, SOCK_ZAPPED);
438 
439 	release_sock(sk);
440 }
441 
442 static void iucv_sock_init(struct sock *sk, struct sock *parent)
443 {
444 	if (parent) {
445 		sk->sk_type = parent->sk_type;
446 		security_sk_clone(parent, sk);
447 	}
448 }
449 
450 static struct sock *iucv_sock_alloc(struct socket *sock, int proto, gfp_t prio, int kern)
451 {
452 	struct sock *sk;
453 	struct iucv_sock *iucv;
454 
455 	sk = sk_alloc(&init_net, PF_IUCV, prio, &iucv_proto, kern);
456 	if (!sk)
457 		return NULL;
458 	iucv = iucv_sk(sk);
459 
460 	sock_init_data(sock, sk);
461 	INIT_LIST_HEAD(&iucv->accept_q);
462 	spin_lock_init(&iucv->accept_q_lock);
463 	skb_queue_head_init(&iucv->send_skb_q);
464 	INIT_LIST_HEAD(&iucv->message_q.list);
465 	spin_lock_init(&iucv->message_q.lock);
466 	skb_queue_head_init(&iucv->backlog_skb_q);
467 	iucv->send_tag = 0;
468 	atomic_set(&iucv->pendings, 0);
469 	iucv->flags = 0;
470 	iucv->msglimit = 0;
471 	atomic_set(&iucv->skbs_in_xmit, 0);
472 	atomic_set(&iucv->msg_sent, 0);
473 	atomic_set(&iucv->msg_recv, 0);
474 	iucv->path = NULL;
475 	iucv->sk_txnotify = afiucv_hs_callback_txnotify;
476 	memset(&iucv->src_user_id , 0, 32);
477 	if (pr_iucv)
478 		iucv->transport = AF_IUCV_TRANS_IUCV;
479 	else
480 		iucv->transport = AF_IUCV_TRANS_HIPER;
481 
482 	sk->sk_destruct = iucv_sock_destruct;
483 	sk->sk_sndtimeo = IUCV_CONN_TIMEOUT;
484 
485 	sock_reset_flag(sk, SOCK_ZAPPED);
486 
487 	sk->sk_protocol = proto;
488 	sk->sk_state	= IUCV_OPEN;
489 
490 	iucv_sock_link(&iucv_sk_list, sk);
491 	return sk;
492 }
493 
494 static void iucv_accept_enqueue(struct sock *parent, struct sock *sk)
495 {
496 	unsigned long flags;
497 	struct iucv_sock *par = iucv_sk(parent);
498 
499 	sock_hold(sk);
500 	spin_lock_irqsave(&par->accept_q_lock, flags);
501 	list_add_tail(&iucv_sk(sk)->accept_q, &par->accept_q);
502 	spin_unlock_irqrestore(&par->accept_q_lock, flags);
503 	iucv_sk(sk)->parent = parent;
504 	sk_acceptq_added(parent);
505 }
506 
507 static void iucv_accept_unlink(struct sock *sk)
508 {
509 	unsigned long flags;
510 	struct iucv_sock *par = iucv_sk(iucv_sk(sk)->parent);
511 
512 	spin_lock_irqsave(&par->accept_q_lock, flags);
513 	list_del_init(&iucv_sk(sk)->accept_q);
514 	spin_unlock_irqrestore(&par->accept_q_lock, flags);
515 	sk_acceptq_removed(iucv_sk(sk)->parent);
516 	iucv_sk(sk)->parent = NULL;
517 	sock_put(sk);
518 }
519 
520 static struct sock *iucv_accept_dequeue(struct sock *parent,
521 					struct socket *newsock)
522 {
523 	struct iucv_sock *isk, *n;
524 	struct sock *sk;
525 
526 	list_for_each_entry_safe(isk, n, &iucv_sk(parent)->accept_q, accept_q) {
527 		sk = (struct sock *) isk;
528 		lock_sock(sk);
529 
530 		if (sk->sk_state == IUCV_CLOSED) {
531 			iucv_accept_unlink(sk);
532 			release_sock(sk);
533 			continue;
534 		}
535 
536 		if (sk->sk_state == IUCV_CONNECTED ||
537 		    sk->sk_state == IUCV_DISCONN ||
538 		    !newsock) {
539 			iucv_accept_unlink(sk);
540 			if (newsock)
541 				sock_graft(sk, newsock);
542 
543 			release_sock(sk);
544 			return sk;
545 		}
546 
547 		release_sock(sk);
548 	}
549 	return NULL;
550 }
551 
552 static void __iucv_auto_name(struct iucv_sock *iucv)
553 {
554 	char name[12];
555 
556 	sprintf(name, "%08x", atomic_inc_return(&iucv_sk_list.autobind_name));
557 	while (__iucv_get_sock_by_name(name)) {
558 		sprintf(name, "%08x",
559 			atomic_inc_return(&iucv_sk_list.autobind_name));
560 	}
561 	memcpy(iucv->src_name, name, 8);
562 }
563 
564 /* Bind an unbound socket */
565 static int iucv_sock_bind(struct socket *sock, struct sockaddr *addr,
566 			  int addr_len)
567 {
568 	DECLARE_SOCKADDR(struct sockaddr_iucv *, sa, addr);
569 	char uid[sizeof(sa->siucv_user_id)];
570 	struct sock *sk = sock->sk;
571 	struct iucv_sock *iucv;
572 	int err = 0;
573 	struct net_device *dev;
574 
575 	/* Verify the input sockaddr */
576 	if (addr_len < sizeof(struct sockaddr_iucv) ||
577 	    addr->sa_family != AF_IUCV)
578 		return -EINVAL;
579 
580 	lock_sock(sk);
581 	if (sk->sk_state != IUCV_OPEN) {
582 		err = -EBADFD;
583 		goto done;
584 	}
585 
586 	write_lock_bh(&iucv_sk_list.lock);
587 
588 	iucv = iucv_sk(sk);
589 	if (__iucv_get_sock_by_name(sa->siucv_name)) {
590 		err = -EADDRINUSE;
591 		goto done_unlock;
592 	}
593 	if (iucv->path)
594 		goto done_unlock;
595 
596 	/* Bind the socket */
597 	if (pr_iucv)
598 		if (!memcmp(sa->siucv_user_id, iucv_userid, 8))
599 			goto vm_bind; /* VM IUCV transport */
600 
601 	/* try hiper transport */
602 	memcpy(uid, sa->siucv_user_id, sizeof(uid));
603 	ASCEBC(uid, 8);
604 	rcu_read_lock();
605 	for_each_netdev_rcu(&init_net, dev) {
606 		if (!memcmp(dev->perm_addr, uid, 8)) {
607 			memcpy(iucv->src_user_id, sa->siucv_user_id, 8);
608 			/* Check for uninitialized siucv_name */
609 			if (strncmp(sa->siucv_name, "        ", 8) == 0)
610 				__iucv_auto_name(iucv);
611 			else
612 				memcpy(iucv->src_name, sa->siucv_name, 8);
613 			sk->sk_bound_dev_if = dev->ifindex;
614 			iucv->hs_dev = dev;
615 			dev_hold(dev);
616 			sk->sk_state = IUCV_BOUND;
617 			iucv->transport = AF_IUCV_TRANS_HIPER;
618 			if (!iucv->msglimit)
619 				iucv->msglimit = IUCV_HIPER_MSGLIM_DEFAULT;
620 			rcu_read_unlock();
621 			goto done_unlock;
622 		}
623 	}
624 	rcu_read_unlock();
625 vm_bind:
626 	if (pr_iucv) {
627 		/* use local userid for backward compat */
628 		memcpy(iucv->src_name, sa->siucv_name, 8);
629 		memcpy(iucv->src_user_id, iucv_userid, 8);
630 		sk->sk_state = IUCV_BOUND;
631 		iucv->transport = AF_IUCV_TRANS_IUCV;
632 		sk->sk_allocation |= GFP_DMA;
633 		if (!iucv->msglimit)
634 			iucv->msglimit = IUCV_QUEUELEN_DEFAULT;
635 		goto done_unlock;
636 	}
637 	/* found no dev to bind */
638 	err = -ENODEV;
639 done_unlock:
640 	/* Release the socket list lock */
641 	write_unlock_bh(&iucv_sk_list.lock);
642 done:
643 	release_sock(sk);
644 	return err;
645 }
646 
647 /* Automatically bind an unbound socket */
648 static int iucv_sock_autobind(struct sock *sk)
649 {
650 	struct iucv_sock *iucv = iucv_sk(sk);
651 	int err = 0;
652 
653 	if (unlikely(!pr_iucv))
654 		return -EPROTO;
655 
656 	memcpy(iucv->src_user_id, iucv_userid, 8);
657 	iucv->transport = AF_IUCV_TRANS_IUCV;
658 	sk->sk_allocation |= GFP_DMA;
659 
660 	write_lock_bh(&iucv_sk_list.lock);
661 	__iucv_auto_name(iucv);
662 	write_unlock_bh(&iucv_sk_list.lock);
663 
664 	if (!iucv->msglimit)
665 		iucv->msglimit = IUCV_QUEUELEN_DEFAULT;
666 
667 	return err;
668 }
669 
670 static int afiucv_path_connect(struct socket *sock, struct sockaddr *addr)
671 {
672 	DECLARE_SOCKADDR(struct sockaddr_iucv *, sa, addr);
673 	struct sock *sk = sock->sk;
674 	struct iucv_sock *iucv = iucv_sk(sk);
675 	unsigned char user_data[16];
676 	int err;
677 
678 	high_nmcpy(user_data, sa->siucv_name);
679 	low_nmcpy(user_data, iucv->src_name);
680 	ASCEBC(user_data, sizeof(user_data));
681 
682 	/* Create path. */
683 	iucv->path = iucv_path_alloc(iucv->msglimit,
684 				     IUCV_IPRMDATA, GFP_KERNEL);
685 	if (!iucv->path) {
686 		err = -ENOMEM;
687 		goto done;
688 	}
689 	err = pr_iucv->path_connect(iucv->path, &af_iucv_handler,
690 				    sa->siucv_user_id, NULL, user_data,
691 				    sk);
692 	if (err) {
693 		iucv_path_free(iucv->path);
694 		iucv->path = NULL;
695 		switch (err) {
696 		case 0x0b:	/* Target communicator is not logged on */
697 			err = -ENETUNREACH;
698 			break;
699 		case 0x0d:	/* Max connections for this guest exceeded */
700 		case 0x0e:	/* Max connections for target guest exceeded */
701 			err = -EAGAIN;
702 			break;
703 		case 0x0f:	/* Missing IUCV authorization */
704 			err = -EACCES;
705 			break;
706 		default:
707 			err = -ECONNREFUSED;
708 			break;
709 		}
710 	}
711 done:
712 	return err;
713 }
714 
715 /* Connect an unconnected socket */
716 static int iucv_sock_connect(struct socket *sock, struct sockaddr *addr,
717 			     int alen, int flags)
718 {
719 	DECLARE_SOCKADDR(struct sockaddr_iucv *, sa, addr);
720 	struct sock *sk = sock->sk;
721 	struct iucv_sock *iucv = iucv_sk(sk);
722 	int err;
723 
724 	if (alen < sizeof(struct sockaddr_iucv) || addr->sa_family != AF_IUCV)
725 		return -EINVAL;
726 
727 	if (sk->sk_state != IUCV_OPEN && sk->sk_state != IUCV_BOUND)
728 		return -EBADFD;
729 
730 	if (sk->sk_state == IUCV_OPEN &&
731 	    iucv->transport == AF_IUCV_TRANS_HIPER)
732 		return -EBADFD; /* explicit bind required */
733 
734 	if (sk->sk_type != SOCK_STREAM && sk->sk_type != SOCK_SEQPACKET)
735 		return -EINVAL;
736 
737 	if (sk->sk_state == IUCV_OPEN) {
738 		err = iucv_sock_autobind(sk);
739 		if (unlikely(err))
740 			return err;
741 	}
742 
743 	lock_sock(sk);
744 
745 	/* Set the destination information */
746 	memcpy(iucv->dst_user_id, sa->siucv_user_id, 8);
747 	memcpy(iucv->dst_name, sa->siucv_name, 8);
748 
749 	if (iucv->transport == AF_IUCV_TRANS_HIPER)
750 		err = iucv_send_ctrl(sock->sk, AF_IUCV_FLAG_SYN);
751 	else
752 		err = afiucv_path_connect(sock, addr);
753 	if (err)
754 		goto done;
755 
756 	if (sk->sk_state != IUCV_CONNECTED)
757 		err = iucv_sock_wait(sk, iucv_sock_in_state(sk, IUCV_CONNECTED,
758 							    IUCV_DISCONN),
759 				     sock_sndtimeo(sk, flags & O_NONBLOCK));
760 
761 	if (sk->sk_state == IUCV_DISCONN || sk->sk_state == IUCV_CLOSED)
762 		err = -ECONNREFUSED;
763 
764 	if (err && iucv->transport == AF_IUCV_TRANS_IUCV)
765 		iucv_sever_path(sk, 0);
766 
767 done:
768 	release_sock(sk);
769 	return err;
770 }
771 
772 /* Move a socket into listening state. */
773 static int iucv_sock_listen(struct socket *sock, int backlog)
774 {
775 	struct sock *sk = sock->sk;
776 	int err;
777 
778 	lock_sock(sk);
779 
780 	err = -EINVAL;
781 	if (sk->sk_state != IUCV_BOUND)
782 		goto done;
783 
784 	if (sock->type != SOCK_STREAM && sock->type != SOCK_SEQPACKET)
785 		goto done;
786 
787 	sk->sk_max_ack_backlog = backlog;
788 	sk->sk_ack_backlog = 0;
789 	sk->sk_state = IUCV_LISTEN;
790 	err = 0;
791 
792 done:
793 	release_sock(sk);
794 	return err;
795 }
796 
797 /* Accept a pending connection */
798 static int iucv_sock_accept(struct socket *sock, struct socket *newsock,
799 			    int flags, bool kern)
800 {
801 	DECLARE_WAITQUEUE(wait, current);
802 	struct sock *sk = sock->sk, *nsk;
803 	long timeo;
804 	int err = 0;
805 
806 	lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
807 
808 	if (sk->sk_state != IUCV_LISTEN) {
809 		err = -EBADFD;
810 		goto done;
811 	}
812 
813 	timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
814 
815 	/* Wait for an incoming connection */
816 	add_wait_queue_exclusive(sk_sleep(sk), &wait);
817 	while (!(nsk = iucv_accept_dequeue(sk, newsock))) {
818 		set_current_state(TASK_INTERRUPTIBLE);
819 		if (!timeo) {
820 			err = -EAGAIN;
821 			break;
822 		}
823 
824 		release_sock(sk);
825 		timeo = schedule_timeout(timeo);
826 		lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
827 
828 		if (sk->sk_state != IUCV_LISTEN) {
829 			err = -EBADFD;
830 			break;
831 		}
832 
833 		if (signal_pending(current)) {
834 			err = sock_intr_errno(timeo);
835 			break;
836 		}
837 	}
838 
839 	set_current_state(TASK_RUNNING);
840 	remove_wait_queue(sk_sleep(sk), &wait);
841 
842 	if (err)
843 		goto done;
844 
845 	newsock->state = SS_CONNECTED;
846 
847 done:
848 	release_sock(sk);
849 	return err;
850 }
851 
852 static int iucv_sock_getname(struct socket *sock, struct sockaddr *addr,
853 			     int peer)
854 {
855 	DECLARE_SOCKADDR(struct sockaddr_iucv *, siucv, addr);
856 	struct sock *sk = sock->sk;
857 	struct iucv_sock *iucv = iucv_sk(sk);
858 
859 	addr->sa_family = AF_IUCV;
860 
861 	if (peer) {
862 		memcpy(siucv->siucv_user_id, iucv->dst_user_id, 8);
863 		memcpy(siucv->siucv_name, iucv->dst_name, 8);
864 	} else {
865 		memcpy(siucv->siucv_user_id, iucv->src_user_id, 8);
866 		memcpy(siucv->siucv_name, iucv->src_name, 8);
867 	}
868 	memset(&siucv->siucv_port, 0, sizeof(siucv->siucv_port));
869 	memset(&siucv->siucv_addr, 0, sizeof(siucv->siucv_addr));
870 	memset(&siucv->siucv_nodeid, 0, sizeof(siucv->siucv_nodeid));
871 
872 	return sizeof(struct sockaddr_iucv);
873 }
874 
875 /**
876  * iucv_send_iprm() - Send socket data in parameter list of an iucv message.
877  * @path:	IUCV path
878  * @msg:	Pointer to a struct iucv_message
879  * @skb:	The socket data to send, skb->len MUST BE <= 7
880  *
881  * Send the socket data in the parameter list in the iucv message
882  * (IUCV_IPRMDATA). The socket data is stored at index 0 to 6 in the parameter
883  * list and the socket data len at index 7 (last byte).
884  * See also iucv_msg_length().
885  *
886  * Returns the error code from the iucv_message_send() call.
887  */
888 static int iucv_send_iprm(struct iucv_path *path, struct iucv_message *msg,
889 			  struct sk_buff *skb)
890 {
891 	u8 prmdata[8];
892 
893 	memcpy(prmdata, (void *) skb->data, skb->len);
894 	prmdata[7] = 0xff - (u8) skb->len;
895 	return pr_iucv->message_send(path, msg, IUCV_IPRMDATA, 0,
896 				 (void *) prmdata, 8);
897 }
898 
899 static int iucv_sock_sendmsg(struct socket *sock, struct msghdr *msg,
900 			     size_t len)
901 {
902 	struct sock *sk = sock->sk;
903 	struct iucv_sock *iucv = iucv_sk(sk);
904 	size_t headroom = 0;
905 	size_t linear;
906 	struct sk_buff *skb;
907 	struct iucv_message txmsg = {0};
908 	struct cmsghdr *cmsg;
909 	int cmsg_done;
910 	long timeo;
911 	char user_id[9];
912 	char appl_id[9];
913 	int err;
914 	int noblock = msg->msg_flags & MSG_DONTWAIT;
915 
916 	err = sock_error(sk);
917 	if (err)
918 		return err;
919 
920 	if (msg->msg_flags & MSG_OOB)
921 		return -EOPNOTSUPP;
922 
923 	/* SOCK_SEQPACKET: we do not support segmented records */
924 	if (sk->sk_type == SOCK_SEQPACKET && !(msg->msg_flags & MSG_EOR))
925 		return -EOPNOTSUPP;
926 
927 	lock_sock(sk);
928 
929 	if (sk->sk_shutdown & SEND_SHUTDOWN) {
930 		err = -EPIPE;
931 		goto out;
932 	}
933 
934 	/* Return if the socket is not in connected state */
935 	if (sk->sk_state != IUCV_CONNECTED) {
936 		err = -ENOTCONN;
937 		goto out;
938 	}
939 
940 	/* initialize defaults */
941 	cmsg_done   = 0;	/* check for duplicate headers */
942 
943 	/* iterate over control messages */
944 	for_each_cmsghdr(cmsg, msg) {
945 		if (!CMSG_OK(msg, cmsg)) {
946 			err = -EINVAL;
947 			goto out;
948 		}
949 
950 		if (cmsg->cmsg_level != SOL_IUCV)
951 			continue;
952 
953 		if (cmsg->cmsg_type & cmsg_done) {
954 			err = -EINVAL;
955 			goto out;
956 		}
957 		cmsg_done |= cmsg->cmsg_type;
958 
959 		switch (cmsg->cmsg_type) {
960 		case SCM_IUCV_TRGCLS:
961 			if (cmsg->cmsg_len != CMSG_LEN(TRGCLS_SIZE)) {
962 				err = -EINVAL;
963 				goto out;
964 			}
965 
966 			/* set iucv message target class */
967 			memcpy(&txmsg.class,
968 				(void *) CMSG_DATA(cmsg), TRGCLS_SIZE);
969 
970 			break;
971 
972 		default:
973 			err = -EINVAL;
974 			goto out;
975 		}
976 	}
977 
978 	/* allocate one skb for each iucv message:
979 	 * this is fine for SOCK_SEQPACKET (unless we want to support
980 	 * segmented records using the MSG_EOR flag), but
981 	 * for SOCK_STREAM we might want to improve it in future */
982 	if (iucv->transport == AF_IUCV_TRANS_HIPER) {
983 		headroom = sizeof(struct af_iucv_trans_hdr) +
984 			   LL_RESERVED_SPACE(iucv->hs_dev);
985 		linear = min(len, PAGE_SIZE - headroom);
986 	} else {
987 		if (len < PAGE_SIZE) {
988 			linear = len;
989 		} else {
990 			/* In nonlinear "classic" iucv skb,
991 			 * reserve space for iucv_array
992 			 */
993 			headroom = sizeof(struct iucv_array) *
994 				   (MAX_SKB_FRAGS + 1);
995 			linear = PAGE_SIZE - headroom;
996 		}
997 	}
998 	skb = sock_alloc_send_pskb(sk, headroom + linear, len - linear,
999 				   noblock, &err, 0);
1000 	if (!skb)
1001 		goto out;
1002 	if (headroom)
1003 		skb_reserve(skb, headroom);
1004 	skb_put(skb, linear);
1005 	skb->len = len;
1006 	skb->data_len = len - linear;
1007 	err = skb_copy_datagram_from_iter(skb, 0, &msg->msg_iter, len);
1008 	if (err)
1009 		goto fail;
1010 
1011 	/* wait if outstanding messages for iucv path has reached */
1012 	timeo = sock_sndtimeo(sk, noblock);
1013 	err = iucv_sock_wait(sk, iucv_below_msglim(sk), timeo);
1014 	if (err)
1015 		goto fail;
1016 
1017 	/* return -ECONNRESET if the socket is no longer connected */
1018 	if (sk->sk_state != IUCV_CONNECTED) {
1019 		err = -ECONNRESET;
1020 		goto fail;
1021 	}
1022 
1023 	/* increment and save iucv message tag for msg_completion cbk */
1024 	txmsg.tag = iucv->send_tag++;
1025 	IUCV_SKB_CB(skb)->tag = txmsg.tag;
1026 
1027 	if (iucv->transport == AF_IUCV_TRANS_HIPER) {
1028 		atomic_inc(&iucv->msg_sent);
1029 		err = afiucv_hs_send(&txmsg, sk, skb, 0);
1030 		if (err) {
1031 			atomic_dec(&iucv->msg_sent);
1032 			goto out;
1033 		}
1034 	} else { /* Classic VM IUCV transport */
1035 		skb_queue_tail(&iucv->send_skb_q, skb);
1036 		atomic_inc(&iucv->skbs_in_xmit);
1037 
1038 		if (((iucv->path->flags & IUCV_IPRMDATA) & iucv->flags) &&
1039 		    skb->len <= 7) {
1040 			err = iucv_send_iprm(iucv->path, &txmsg, skb);
1041 
1042 			/* on success: there is no message_complete callback */
1043 			/* for an IPRMDATA msg; remove skb from send queue   */
1044 			if (err == 0) {
1045 				atomic_dec(&iucv->skbs_in_xmit);
1046 				skb_unlink(skb, &iucv->send_skb_q);
1047 				consume_skb(skb);
1048 			}
1049 
1050 			/* this error should never happen since the	*/
1051 			/* IUCV_IPRMDATA path flag is set... sever path */
1052 			if (err == 0x15) {
1053 				pr_iucv->path_sever(iucv->path, NULL);
1054 				atomic_dec(&iucv->skbs_in_xmit);
1055 				skb_unlink(skb, &iucv->send_skb_q);
1056 				err = -EPIPE;
1057 				goto fail;
1058 			}
1059 		} else if (skb_is_nonlinear(skb)) {
1060 			struct iucv_array *iba = (struct iucv_array *)skb->head;
1061 			int i;
1062 
1063 			/* skip iucv_array lying in the headroom */
1064 			iba[0].address = (u32)(addr_t)skb->data;
1065 			iba[0].length = (u32)skb_headlen(skb);
1066 			for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1067 				skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1068 
1069 				iba[i + 1].address =
1070 					(u32)(addr_t)skb_frag_address(frag);
1071 				iba[i + 1].length = (u32)skb_frag_size(frag);
1072 			}
1073 			err = pr_iucv->message_send(iucv->path, &txmsg,
1074 						    IUCV_IPBUFLST, 0,
1075 						    (void *)iba, skb->len);
1076 		} else { /* non-IPRM Linear skb */
1077 			err = pr_iucv->message_send(iucv->path, &txmsg,
1078 					0, 0, (void *)skb->data, skb->len);
1079 		}
1080 		if (err) {
1081 			if (err == 3) {
1082 				user_id[8] = 0;
1083 				memcpy(user_id, iucv->dst_user_id, 8);
1084 				appl_id[8] = 0;
1085 				memcpy(appl_id, iucv->dst_name, 8);
1086 				pr_err(
1087 		"Application %s on z/VM guest %s exceeds message limit\n",
1088 					appl_id, user_id);
1089 				err = -EAGAIN;
1090 			} else {
1091 				err = -EPIPE;
1092 			}
1093 
1094 			atomic_dec(&iucv->skbs_in_xmit);
1095 			skb_unlink(skb, &iucv->send_skb_q);
1096 			goto fail;
1097 		}
1098 	}
1099 
1100 	release_sock(sk);
1101 	return len;
1102 
1103 fail:
1104 	kfree_skb(skb);
1105 out:
1106 	release_sock(sk);
1107 	return err;
1108 }
1109 
1110 static struct sk_buff *alloc_iucv_recv_skb(unsigned long len)
1111 {
1112 	size_t headroom, linear;
1113 	struct sk_buff *skb;
1114 	int err;
1115 
1116 	if (len < PAGE_SIZE) {
1117 		headroom = 0;
1118 		linear = len;
1119 	} else {
1120 		headroom = sizeof(struct iucv_array) * (MAX_SKB_FRAGS + 1);
1121 		linear = PAGE_SIZE - headroom;
1122 	}
1123 	skb = alloc_skb_with_frags(headroom + linear, len - linear,
1124 				   0, &err, GFP_ATOMIC | GFP_DMA);
1125 	WARN_ONCE(!skb,
1126 		  "alloc of recv iucv skb len=%lu failed with errcode=%d\n",
1127 		  len, err);
1128 	if (skb) {
1129 		if (headroom)
1130 			skb_reserve(skb, headroom);
1131 		skb_put(skb, linear);
1132 		skb->len = len;
1133 		skb->data_len = len - linear;
1134 	}
1135 	return skb;
1136 }
1137 
1138 /* iucv_process_message() - Receive a single outstanding IUCV message
1139  *
1140  * Locking: must be called with message_q.lock held
1141  */
1142 static void iucv_process_message(struct sock *sk, struct sk_buff *skb,
1143 				 struct iucv_path *path,
1144 				 struct iucv_message *msg)
1145 {
1146 	int rc;
1147 	unsigned int len;
1148 
1149 	len = iucv_msg_length(msg);
1150 
1151 	/* store msg target class in the second 4 bytes of skb ctrl buffer */
1152 	/* Note: the first 4 bytes are reserved for msg tag */
1153 	IUCV_SKB_CB(skb)->class = msg->class;
1154 
1155 	/* check for special IPRM messages (e.g. iucv_sock_shutdown) */
1156 	if ((msg->flags & IUCV_IPRMDATA) && len > 7) {
1157 		if (memcmp(msg->rmmsg, iprm_shutdown, 8) == 0) {
1158 			skb->data = NULL;
1159 			skb->len = 0;
1160 		}
1161 	} else {
1162 		if (skb_is_nonlinear(skb)) {
1163 			struct iucv_array *iba = (struct iucv_array *)skb->head;
1164 			int i;
1165 
1166 			iba[0].address = (u32)(addr_t)skb->data;
1167 			iba[0].length = (u32)skb_headlen(skb);
1168 			for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1169 				skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1170 
1171 				iba[i + 1].address =
1172 					(u32)(addr_t)skb_frag_address(frag);
1173 				iba[i + 1].length = (u32)skb_frag_size(frag);
1174 			}
1175 			rc = pr_iucv->message_receive(path, msg,
1176 					      IUCV_IPBUFLST,
1177 					      (void *)iba, len, NULL);
1178 		} else {
1179 			rc = pr_iucv->message_receive(path, msg,
1180 					      msg->flags & IUCV_IPRMDATA,
1181 					      skb->data, len, NULL);
1182 		}
1183 		if (rc) {
1184 			kfree_skb(skb);
1185 			return;
1186 		}
1187 		WARN_ON_ONCE(skb->len != len);
1188 	}
1189 
1190 	IUCV_SKB_CB(skb)->offset = 0;
1191 	if (sk_filter(sk, skb)) {
1192 		atomic_inc(&sk->sk_drops);	/* skb rejected by filter */
1193 		kfree_skb(skb);
1194 		return;
1195 	}
1196 	if (__sock_queue_rcv_skb(sk, skb))	/* handle rcv queue full */
1197 		skb_queue_tail(&iucv_sk(sk)->backlog_skb_q, skb);
1198 }
1199 
1200 /* iucv_process_message_q() - Process outstanding IUCV messages
1201  *
1202  * Locking: must be called with message_q.lock held
1203  */
1204 static void iucv_process_message_q(struct sock *sk)
1205 {
1206 	struct iucv_sock *iucv = iucv_sk(sk);
1207 	struct sk_buff *skb;
1208 	struct sock_msg_q *p, *n;
1209 
1210 	list_for_each_entry_safe(p, n, &iucv->message_q.list, list) {
1211 		skb = alloc_iucv_recv_skb(iucv_msg_length(&p->msg));
1212 		if (!skb)
1213 			break;
1214 		iucv_process_message(sk, skb, p->path, &p->msg);
1215 		list_del(&p->list);
1216 		kfree(p);
1217 		if (!skb_queue_empty(&iucv->backlog_skb_q))
1218 			break;
1219 	}
1220 }
1221 
1222 static int iucv_sock_recvmsg(struct socket *sock, struct msghdr *msg,
1223 			     size_t len, int flags)
1224 {
1225 	int noblock = flags & MSG_DONTWAIT;
1226 	struct sock *sk = sock->sk;
1227 	struct iucv_sock *iucv = iucv_sk(sk);
1228 	unsigned int copied, rlen;
1229 	struct sk_buff *skb, *rskb, *cskb;
1230 	int err = 0;
1231 	u32 offset;
1232 
1233 	if ((sk->sk_state == IUCV_DISCONN) &&
1234 	    skb_queue_empty(&iucv->backlog_skb_q) &&
1235 	    skb_queue_empty(&sk->sk_receive_queue) &&
1236 	    list_empty(&iucv->message_q.list))
1237 		return 0;
1238 
1239 	if (flags & (MSG_OOB))
1240 		return -EOPNOTSUPP;
1241 
1242 	/* receive/dequeue next skb:
1243 	 * the function understands MSG_PEEK and, thus, does not dequeue skb */
1244 	skb = skb_recv_datagram(sk, flags, noblock, &err);
1245 	if (!skb) {
1246 		if (sk->sk_shutdown & RCV_SHUTDOWN)
1247 			return 0;
1248 		return err;
1249 	}
1250 
1251 	offset = IUCV_SKB_CB(skb)->offset;
1252 	rlen   = skb->len - offset;		/* real length of skb */
1253 	copied = min_t(unsigned int, rlen, len);
1254 	if (!rlen)
1255 		sk->sk_shutdown = sk->sk_shutdown | RCV_SHUTDOWN;
1256 
1257 	cskb = skb;
1258 	if (skb_copy_datagram_msg(cskb, offset, msg, copied)) {
1259 		if (!(flags & MSG_PEEK))
1260 			skb_queue_head(&sk->sk_receive_queue, skb);
1261 		return -EFAULT;
1262 	}
1263 
1264 	/* SOCK_SEQPACKET: set MSG_TRUNC if recv buf size is too small */
1265 	if (sk->sk_type == SOCK_SEQPACKET) {
1266 		if (copied < rlen)
1267 			msg->msg_flags |= MSG_TRUNC;
1268 		/* each iucv message contains a complete record */
1269 		msg->msg_flags |= MSG_EOR;
1270 	}
1271 
1272 	/* create control message to store iucv msg target class:
1273 	 * get the trgcls from the control buffer of the skb due to
1274 	 * fragmentation of original iucv message. */
1275 	err = put_cmsg(msg, SOL_IUCV, SCM_IUCV_TRGCLS,
1276 		       sizeof(IUCV_SKB_CB(skb)->class),
1277 		       (void *)&IUCV_SKB_CB(skb)->class);
1278 	if (err) {
1279 		if (!(flags & MSG_PEEK))
1280 			skb_queue_head(&sk->sk_receive_queue, skb);
1281 		return err;
1282 	}
1283 
1284 	/* Mark read part of skb as used */
1285 	if (!(flags & MSG_PEEK)) {
1286 
1287 		/* SOCK_STREAM: re-queue skb if it contains unreceived data */
1288 		if (sk->sk_type == SOCK_STREAM) {
1289 			if (copied < rlen) {
1290 				IUCV_SKB_CB(skb)->offset = offset + copied;
1291 				skb_queue_head(&sk->sk_receive_queue, skb);
1292 				goto done;
1293 			}
1294 		}
1295 
1296 		consume_skb(skb);
1297 		if (iucv->transport == AF_IUCV_TRANS_HIPER) {
1298 			atomic_inc(&iucv->msg_recv);
1299 			if (atomic_read(&iucv->msg_recv) > iucv->msglimit) {
1300 				WARN_ON(1);
1301 				iucv_sock_close(sk);
1302 				return -EFAULT;
1303 			}
1304 		}
1305 
1306 		/* Queue backlog skbs */
1307 		spin_lock_bh(&iucv->message_q.lock);
1308 		rskb = skb_dequeue(&iucv->backlog_skb_q);
1309 		while (rskb) {
1310 			IUCV_SKB_CB(rskb)->offset = 0;
1311 			if (__sock_queue_rcv_skb(sk, rskb)) {
1312 				/* handle rcv queue full */
1313 				skb_queue_head(&iucv->backlog_skb_q,
1314 						rskb);
1315 				break;
1316 			}
1317 			rskb = skb_dequeue(&iucv->backlog_skb_q);
1318 		}
1319 		if (skb_queue_empty(&iucv->backlog_skb_q)) {
1320 			if (!list_empty(&iucv->message_q.list))
1321 				iucv_process_message_q(sk);
1322 			if (atomic_read(&iucv->msg_recv) >=
1323 							iucv->msglimit / 2) {
1324 				err = iucv_send_ctrl(sk, AF_IUCV_FLAG_WIN);
1325 				if (err) {
1326 					sk->sk_state = IUCV_DISCONN;
1327 					sk->sk_state_change(sk);
1328 				}
1329 			}
1330 		}
1331 		spin_unlock_bh(&iucv->message_q.lock);
1332 	}
1333 
1334 done:
1335 	/* SOCK_SEQPACKET: return real length if MSG_TRUNC is set */
1336 	if (sk->sk_type == SOCK_SEQPACKET && (flags & MSG_TRUNC))
1337 		copied = rlen;
1338 
1339 	return copied;
1340 }
1341 
1342 static inline __poll_t iucv_accept_poll(struct sock *parent)
1343 {
1344 	struct iucv_sock *isk, *n;
1345 	struct sock *sk;
1346 
1347 	list_for_each_entry_safe(isk, n, &iucv_sk(parent)->accept_q, accept_q) {
1348 		sk = (struct sock *) isk;
1349 
1350 		if (sk->sk_state == IUCV_CONNECTED)
1351 			return EPOLLIN | EPOLLRDNORM;
1352 	}
1353 
1354 	return 0;
1355 }
1356 
1357 static __poll_t iucv_sock_poll(struct file *file, struct socket *sock,
1358 			       poll_table *wait)
1359 {
1360 	struct sock *sk = sock->sk;
1361 	__poll_t mask = 0;
1362 
1363 	sock_poll_wait(file, sock, wait);
1364 
1365 	if (sk->sk_state == IUCV_LISTEN)
1366 		return iucv_accept_poll(sk);
1367 
1368 	if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
1369 		mask |= EPOLLERR |
1370 			(sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? EPOLLPRI : 0);
1371 
1372 	if (sk->sk_shutdown & RCV_SHUTDOWN)
1373 		mask |= EPOLLRDHUP;
1374 
1375 	if (sk->sk_shutdown == SHUTDOWN_MASK)
1376 		mask |= EPOLLHUP;
1377 
1378 	if (!skb_queue_empty(&sk->sk_receive_queue) ||
1379 	    (sk->sk_shutdown & RCV_SHUTDOWN))
1380 		mask |= EPOLLIN | EPOLLRDNORM;
1381 
1382 	if (sk->sk_state == IUCV_CLOSED)
1383 		mask |= EPOLLHUP;
1384 
1385 	if (sk->sk_state == IUCV_DISCONN)
1386 		mask |= EPOLLIN;
1387 
1388 	if (sock_writeable(sk) && iucv_below_msglim(sk))
1389 		mask |= EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND;
1390 	else
1391 		sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
1392 
1393 	return mask;
1394 }
1395 
1396 static int iucv_sock_shutdown(struct socket *sock, int how)
1397 {
1398 	struct sock *sk = sock->sk;
1399 	struct iucv_sock *iucv = iucv_sk(sk);
1400 	struct iucv_message txmsg;
1401 	int err = 0;
1402 
1403 	how++;
1404 
1405 	if ((how & ~SHUTDOWN_MASK) || !how)
1406 		return -EINVAL;
1407 
1408 	lock_sock(sk);
1409 	switch (sk->sk_state) {
1410 	case IUCV_LISTEN:
1411 	case IUCV_DISCONN:
1412 	case IUCV_CLOSING:
1413 	case IUCV_CLOSED:
1414 		err = -ENOTCONN;
1415 		goto fail;
1416 	default:
1417 		break;
1418 	}
1419 
1420 	if ((how == SEND_SHUTDOWN || how == SHUTDOWN_MASK) &&
1421 	    sk->sk_state == IUCV_CONNECTED) {
1422 		if (iucv->transport == AF_IUCV_TRANS_IUCV) {
1423 			txmsg.class = 0;
1424 			txmsg.tag = 0;
1425 			err = pr_iucv->message_send(iucv->path, &txmsg,
1426 				IUCV_IPRMDATA, 0, (void *) iprm_shutdown, 8);
1427 			if (err) {
1428 				switch (err) {
1429 				case 1:
1430 					err = -ENOTCONN;
1431 					break;
1432 				case 2:
1433 					err = -ECONNRESET;
1434 					break;
1435 				default:
1436 					err = -ENOTCONN;
1437 					break;
1438 				}
1439 			}
1440 		} else
1441 			iucv_send_ctrl(sk, AF_IUCV_FLAG_SHT);
1442 	}
1443 
1444 	sk->sk_shutdown |= how;
1445 	if (how == RCV_SHUTDOWN || how == SHUTDOWN_MASK) {
1446 		if ((iucv->transport == AF_IUCV_TRANS_IUCV) &&
1447 		    iucv->path) {
1448 			err = pr_iucv->path_quiesce(iucv->path, NULL);
1449 			if (err)
1450 				err = -ENOTCONN;
1451 /*			skb_queue_purge(&sk->sk_receive_queue); */
1452 		}
1453 		skb_queue_purge(&sk->sk_receive_queue);
1454 	}
1455 
1456 	/* Wake up anyone sleeping in poll */
1457 	sk->sk_state_change(sk);
1458 
1459 fail:
1460 	release_sock(sk);
1461 	return err;
1462 }
1463 
1464 static int iucv_sock_release(struct socket *sock)
1465 {
1466 	struct sock *sk = sock->sk;
1467 	int err = 0;
1468 
1469 	if (!sk)
1470 		return 0;
1471 
1472 	iucv_sock_close(sk);
1473 
1474 	sock_orphan(sk);
1475 	iucv_sock_kill(sk);
1476 	return err;
1477 }
1478 
1479 /* getsockopt and setsockopt */
1480 static int iucv_sock_setsockopt(struct socket *sock, int level, int optname,
1481 				sockptr_t optval, unsigned int optlen)
1482 {
1483 	struct sock *sk = sock->sk;
1484 	struct iucv_sock *iucv = iucv_sk(sk);
1485 	int val;
1486 	int rc;
1487 
1488 	if (level != SOL_IUCV)
1489 		return -ENOPROTOOPT;
1490 
1491 	if (optlen < sizeof(int))
1492 		return -EINVAL;
1493 
1494 	if (copy_from_sockptr(&val, optval, sizeof(int)))
1495 		return -EFAULT;
1496 
1497 	rc = 0;
1498 
1499 	lock_sock(sk);
1500 	switch (optname) {
1501 	case SO_IPRMDATA_MSG:
1502 		if (val)
1503 			iucv->flags |= IUCV_IPRMDATA;
1504 		else
1505 			iucv->flags &= ~IUCV_IPRMDATA;
1506 		break;
1507 	case SO_MSGLIMIT:
1508 		switch (sk->sk_state) {
1509 		case IUCV_OPEN:
1510 		case IUCV_BOUND:
1511 			if (val < 1 || val > U16_MAX)
1512 				rc = -EINVAL;
1513 			else
1514 				iucv->msglimit = val;
1515 			break;
1516 		default:
1517 			rc = -EINVAL;
1518 			break;
1519 		}
1520 		break;
1521 	default:
1522 		rc = -ENOPROTOOPT;
1523 		break;
1524 	}
1525 	release_sock(sk);
1526 
1527 	return rc;
1528 }
1529 
1530 static int iucv_sock_getsockopt(struct socket *sock, int level, int optname,
1531 				char __user *optval, int __user *optlen)
1532 {
1533 	struct sock *sk = sock->sk;
1534 	struct iucv_sock *iucv = iucv_sk(sk);
1535 	unsigned int val;
1536 	int len;
1537 
1538 	if (level != SOL_IUCV)
1539 		return -ENOPROTOOPT;
1540 
1541 	if (get_user(len, optlen))
1542 		return -EFAULT;
1543 
1544 	if (len < 0)
1545 		return -EINVAL;
1546 
1547 	len = min_t(unsigned int, len, sizeof(int));
1548 
1549 	switch (optname) {
1550 	case SO_IPRMDATA_MSG:
1551 		val = (iucv->flags & IUCV_IPRMDATA) ? 1 : 0;
1552 		break;
1553 	case SO_MSGLIMIT:
1554 		lock_sock(sk);
1555 		val = (iucv->path != NULL) ? iucv->path->msglim	/* connected */
1556 					   : iucv->msglimit;	/* default */
1557 		release_sock(sk);
1558 		break;
1559 	case SO_MSGSIZE:
1560 		if (sk->sk_state == IUCV_OPEN)
1561 			return -EBADFD;
1562 		val = (iucv->hs_dev) ? iucv->hs_dev->mtu -
1563 				sizeof(struct af_iucv_trans_hdr) - ETH_HLEN :
1564 				0x7fffffff;
1565 		break;
1566 	default:
1567 		return -ENOPROTOOPT;
1568 	}
1569 
1570 	if (put_user(len, optlen))
1571 		return -EFAULT;
1572 	if (copy_to_user(optval, &val, len))
1573 		return -EFAULT;
1574 
1575 	return 0;
1576 }
1577 
1578 
1579 /* Callback wrappers - called from iucv base support */
1580 static int iucv_callback_connreq(struct iucv_path *path,
1581 				 u8 ipvmid[8], u8 ipuser[16])
1582 {
1583 	unsigned char user_data[16];
1584 	unsigned char nuser_data[16];
1585 	unsigned char src_name[8];
1586 	struct sock *sk, *nsk;
1587 	struct iucv_sock *iucv, *niucv;
1588 	int err;
1589 
1590 	memcpy(src_name, ipuser, 8);
1591 	EBCASC(src_name, 8);
1592 	/* Find out if this path belongs to af_iucv. */
1593 	read_lock(&iucv_sk_list.lock);
1594 	iucv = NULL;
1595 	sk = NULL;
1596 	sk_for_each(sk, &iucv_sk_list.head)
1597 		if (sk->sk_state == IUCV_LISTEN &&
1598 		    !memcmp(&iucv_sk(sk)->src_name, src_name, 8)) {
1599 			/*
1600 			 * Found a listening socket with
1601 			 * src_name == ipuser[0-7].
1602 			 */
1603 			iucv = iucv_sk(sk);
1604 			break;
1605 		}
1606 	read_unlock(&iucv_sk_list.lock);
1607 	if (!iucv)
1608 		/* No socket found, not one of our paths. */
1609 		return -EINVAL;
1610 
1611 	bh_lock_sock(sk);
1612 
1613 	/* Check if parent socket is listening */
1614 	low_nmcpy(user_data, iucv->src_name);
1615 	high_nmcpy(user_data, iucv->dst_name);
1616 	ASCEBC(user_data, sizeof(user_data));
1617 	if (sk->sk_state != IUCV_LISTEN) {
1618 		err = pr_iucv->path_sever(path, user_data);
1619 		iucv_path_free(path);
1620 		goto fail;
1621 	}
1622 
1623 	/* Check for backlog size */
1624 	if (sk_acceptq_is_full(sk)) {
1625 		err = pr_iucv->path_sever(path, user_data);
1626 		iucv_path_free(path);
1627 		goto fail;
1628 	}
1629 
1630 	/* Create the new socket */
1631 	nsk = iucv_sock_alloc(NULL, sk->sk_protocol, GFP_ATOMIC, 0);
1632 	if (!nsk) {
1633 		err = pr_iucv->path_sever(path, user_data);
1634 		iucv_path_free(path);
1635 		goto fail;
1636 	}
1637 
1638 	niucv = iucv_sk(nsk);
1639 	iucv_sock_init(nsk, sk);
1640 	niucv->transport = AF_IUCV_TRANS_IUCV;
1641 	nsk->sk_allocation |= GFP_DMA;
1642 
1643 	/* Set the new iucv_sock */
1644 	memcpy(niucv->dst_name, ipuser + 8, 8);
1645 	EBCASC(niucv->dst_name, 8);
1646 	memcpy(niucv->dst_user_id, ipvmid, 8);
1647 	memcpy(niucv->src_name, iucv->src_name, 8);
1648 	memcpy(niucv->src_user_id, iucv->src_user_id, 8);
1649 	niucv->path = path;
1650 
1651 	/* Call iucv_accept */
1652 	high_nmcpy(nuser_data, ipuser + 8);
1653 	memcpy(nuser_data + 8, niucv->src_name, 8);
1654 	ASCEBC(nuser_data + 8, 8);
1655 
1656 	/* set message limit for path based on msglimit of accepting socket */
1657 	niucv->msglimit = iucv->msglimit;
1658 	path->msglim = iucv->msglimit;
1659 	err = pr_iucv->path_accept(path, &af_iucv_handler, nuser_data, nsk);
1660 	if (err) {
1661 		iucv_sever_path(nsk, 1);
1662 		iucv_sock_kill(nsk);
1663 		goto fail;
1664 	}
1665 
1666 	iucv_accept_enqueue(sk, nsk);
1667 
1668 	/* Wake up accept */
1669 	nsk->sk_state = IUCV_CONNECTED;
1670 	sk->sk_data_ready(sk);
1671 	err = 0;
1672 fail:
1673 	bh_unlock_sock(sk);
1674 	return 0;
1675 }
1676 
1677 static void iucv_callback_connack(struct iucv_path *path, u8 ipuser[16])
1678 {
1679 	struct sock *sk = path->private;
1680 
1681 	sk->sk_state = IUCV_CONNECTED;
1682 	sk->sk_state_change(sk);
1683 }
1684 
1685 static void iucv_callback_rx(struct iucv_path *path, struct iucv_message *msg)
1686 {
1687 	struct sock *sk = path->private;
1688 	struct iucv_sock *iucv = iucv_sk(sk);
1689 	struct sk_buff *skb;
1690 	struct sock_msg_q *save_msg;
1691 	int len;
1692 
1693 	if (sk->sk_shutdown & RCV_SHUTDOWN) {
1694 		pr_iucv->message_reject(path, msg);
1695 		return;
1696 	}
1697 
1698 	spin_lock(&iucv->message_q.lock);
1699 
1700 	if (!list_empty(&iucv->message_q.list) ||
1701 	    !skb_queue_empty(&iucv->backlog_skb_q))
1702 		goto save_message;
1703 
1704 	len = atomic_read(&sk->sk_rmem_alloc);
1705 	len += SKB_TRUESIZE(iucv_msg_length(msg));
1706 	if (len > sk->sk_rcvbuf)
1707 		goto save_message;
1708 
1709 	skb = alloc_iucv_recv_skb(iucv_msg_length(msg));
1710 	if (!skb)
1711 		goto save_message;
1712 
1713 	iucv_process_message(sk, skb, path, msg);
1714 	goto out_unlock;
1715 
1716 save_message:
1717 	save_msg = kzalloc(sizeof(struct sock_msg_q), GFP_ATOMIC | GFP_DMA);
1718 	if (!save_msg)
1719 		goto out_unlock;
1720 	save_msg->path = path;
1721 	save_msg->msg = *msg;
1722 
1723 	list_add_tail(&save_msg->list, &iucv->message_q.list);
1724 
1725 out_unlock:
1726 	spin_unlock(&iucv->message_q.lock);
1727 }
1728 
1729 static void iucv_callback_txdone(struct iucv_path *path,
1730 				 struct iucv_message *msg)
1731 {
1732 	struct sock *sk = path->private;
1733 	struct sk_buff *this = NULL;
1734 	struct sk_buff_head *list;
1735 	struct sk_buff *list_skb;
1736 	struct iucv_sock *iucv;
1737 	unsigned long flags;
1738 
1739 	iucv = iucv_sk(sk);
1740 	list = &iucv->send_skb_q;
1741 
1742 	bh_lock_sock(sk);
1743 
1744 	spin_lock_irqsave(&list->lock, flags);
1745 	skb_queue_walk(list, list_skb) {
1746 		if (msg->tag == IUCV_SKB_CB(list_skb)->tag) {
1747 			this = list_skb;
1748 			break;
1749 		}
1750 	}
1751 	if (this) {
1752 		atomic_dec(&iucv->skbs_in_xmit);
1753 		__skb_unlink(this, list);
1754 	}
1755 
1756 	spin_unlock_irqrestore(&list->lock, flags);
1757 
1758 	if (this) {
1759 		consume_skb(this);
1760 		/* wake up any process waiting for sending */
1761 		iucv_sock_wake_msglim(sk);
1762 	}
1763 
1764 	if (sk->sk_state == IUCV_CLOSING) {
1765 		if (atomic_read(&iucv->skbs_in_xmit) == 0) {
1766 			sk->sk_state = IUCV_CLOSED;
1767 			sk->sk_state_change(sk);
1768 		}
1769 	}
1770 	bh_unlock_sock(sk);
1771 
1772 }
1773 
1774 static void iucv_callback_connrej(struct iucv_path *path, u8 ipuser[16])
1775 {
1776 	struct sock *sk = path->private;
1777 
1778 	if (sk->sk_state == IUCV_CLOSED)
1779 		return;
1780 
1781 	bh_lock_sock(sk);
1782 	iucv_sever_path(sk, 1);
1783 	sk->sk_state = IUCV_DISCONN;
1784 
1785 	sk->sk_state_change(sk);
1786 	bh_unlock_sock(sk);
1787 }
1788 
1789 /* called if the other communication side shuts down its RECV direction;
1790  * in turn, the callback sets SEND_SHUTDOWN to disable sending of data.
1791  */
1792 static void iucv_callback_shutdown(struct iucv_path *path, u8 ipuser[16])
1793 {
1794 	struct sock *sk = path->private;
1795 
1796 	bh_lock_sock(sk);
1797 	if (sk->sk_state != IUCV_CLOSED) {
1798 		sk->sk_shutdown |= SEND_SHUTDOWN;
1799 		sk->sk_state_change(sk);
1800 	}
1801 	bh_unlock_sock(sk);
1802 }
1803 
1804 static struct iucv_handler af_iucv_handler = {
1805 	.path_pending		= iucv_callback_connreq,
1806 	.path_complete		= iucv_callback_connack,
1807 	.path_severed		= iucv_callback_connrej,
1808 	.message_pending	= iucv_callback_rx,
1809 	.message_complete	= iucv_callback_txdone,
1810 	.path_quiesced		= iucv_callback_shutdown,
1811 };
1812 
1813 /***************** HiperSockets transport callbacks ********************/
1814 static void afiucv_swap_src_dest(struct sk_buff *skb)
1815 {
1816 	struct af_iucv_trans_hdr *trans_hdr = iucv_trans_hdr(skb);
1817 	char tmpID[8];
1818 	char tmpName[8];
1819 
1820 	ASCEBC(trans_hdr->destUserID, sizeof(trans_hdr->destUserID));
1821 	ASCEBC(trans_hdr->destAppName, sizeof(trans_hdr->destAppName));
1822 	ASCEBC(trans_hdr->srcUserID, sizeof(trans_hdr->srcUserID));
1823 	ASCEBC(trans_hdr->srcAppName, sizeof(trans_hdr->srcAppName));
1824 	memcpy(tmpID, trans_hdr->srcUserID, 8);
1825 	memcpy(tmpName, trans_hdr->srcAppName, 8);
1826 	memcpy(trans_hdr->srcUserID, trans_hdr->destUserID, 8);
1827 	memcpy(trans_hdr->srcAppName, trans_hdr->destAppName, 8);
1828 	memcpy(trans_hdr->destUserID, tmpID, 8);
1829 	memcpy(trans_hdr->destAppName, tmpName, 8);
1830 	skb_push(skb, ETH_HLEN);
1831 	memset(skb->data, 0, ETH_HLEN);
1832 }
1833 
1834 /**
1835  * afiucv_hs_callback_syn - react on received SYN
1836  **/
1837 static int afiucv_hs_callback_syn(struct sock *sk, struct sk_buff *skb)
1838 {
1839 	struct af_iucv_trans_hdr *trans_hdr = iucv_trans_hdr(skb);
1840 	struct sock *nsk;
1841 	struct iucv_sock *iucv, *niucv;
1842 	int err;
1843 
1844 	iucv = iucv_sk(sk);
1845 	if (!iucv) {
1846 		/* no sock - connection refused */
1847 		afiucv_swap_src_dest(skb);
1848 		trans_hdr->flags = AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_FIN;
1849 		err = dev_queue_xmit(skb);
1850 		goto out;
1851 	}
1852 
1853 	nsk = iucv_sock_alloc(NULL, sk->sk_protocol, GFP_ATOMIC, 0);
1854 	bh_lock_sock(sk);
1855 	if ((sk->sk_state != IUCV_LISTEN) ||
1856 	    sk_acceptq_is_full(sk) ||
1857 	    !nsk) {
1858 		/* error on server socket - connection refused */
1859 		afiucv_swap_src_dest(skb);
1860 		trans_hdr->flags = AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_FIN;
1861 		err = dev_queue_xmit(skb);
1862 		iucv_sock_kill(nsk);
1863 		bh_unlock_sock(sk);
1864 		goto out;
1865 	}
1866 
1867 	niucv = iucv_sk(nsk);
1868 	iucv_sock_init(nsk, sk);
1869 	niucv->transport = AF_IUCV_TRANS_HIPER;
1870 	niucv->msglimit = iucv->msglimit;
1871 	if (!trans_hdr->window)
1872 		niucv->msglimit_peer = IUCV_HIPER_MSGLIM_DEFAULT;
1873 	else
1874 		niucv->msglimit_peer = trans_hdr->window;
1875 	memcpy(niucv->dst_name, trans_hdr->srcAppName, 8);
1876 	memcpy(niucv->dst_user_id, trans_hdr->srcUserID, 8);
1877 	memcpy(niucv->src_name, iucv->src_name, 8);
1878 	memcpy(niucv->src_user_id, iucv->src_user_id, 8);
1879 	nsk->sk_bound_dev_if = sk->sk_bound_dev_if;
1880 	niucv->hs_dev = iucv->hs_dev;
1881 	dev_hold(niucv->hs_dev);
1882 	afiucv_swap_src_dest(skb);
1883 	trans_hdr->flags = AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_ACK;
1884 	trans_hdr->window = niucv->msglimit;
1885 	/* if receiver acks the xmit connection is established */
1886 	err = dev_queue_xmit(skb);
1887 	if (!err) {
1888 		iucv_accept_enqueue(sk, nsk);
1889 		nsk->sk_state = IUCV_CONNECTED;
1890 		sk->sk_data_ready(sk);
1891 	} else
1892 		iucv_sock_kill(nsk);
1893 	bh_unlock_sock(sk);
1894 
1895 out:
1896 	return NET_RX_SUCCESS;
1897 }
1898 
1899 /**
1900  * afiucv_hs_callback_synack() - react on received SYN-ACK
1901  **/
1902 static int afiucv_hs_callback_synack(struct sock *sk, struct sk_buff *skb)
1903 {
1904 	struct iucv_sock *iucv = iucv_sk(sk);
1905 
1906 	if (!iucv || sk->sk_state != IUCV_BOUND) {
1907 		kfree_skb(skb);
1908 		return NET_RX_SUCCESS;
1909 	}
1910 
1911 	bh_lock_sock(sk);
1912 	iucv->msglimit_peer = iucv_trans_hdr(skb)->window;
1913 	sk->sk_state = IUCV_CONNECTED;
1914 	sk->sk_state_change(sk);
1915 	bh_unlock_sock(sk);
1916 	consume_skb(skb);
1917 	return NET_RX_SUCCESS;
1918 }
1919 
1920 /**
1921  * afiucv_hs_callback_synfin() - react on received SYN_FIN
1922  **/
1923 static int afiucv_hs_callback_synfin(struct sock *sk, struct sk_buff *skb)
1924 {
1925 	struct iucv_sock *iucv = iucv_sk(sk);
1926 
1927 	if (!iucv || sk->sk_state != IUCV_BOUND) {
1928 		kfree_skb(skb);
1929 		return NET_RX_SUCCESS;
1930 	}
1931 
1932 	bh_lock_sock(sk);
1933 	sk->sk_state = IUCV_DISCONN;
1934 	sk->sk_state_change(sk);
1935 	bh_unlock_sock(sk);
1936 	consume_skb(skb);
1937 	return NET_RX_SUCCESS;
1938 }
1939 
1940 /**
1941  * afiucv_hs_callback_fin() - react on received FIN
1942  **/
1943 static int afiucv_hs_callback_fin(struct sock *sk, struct sk_buff *skb)
1944 {
1945 	struct iucv_sock *iucv = iucv_sk(sk);
1946 
1947 	/* other end of connection closed */
1948 	if (!iucv) {
1949 		kfree_skb(skb);
1950 		return NET_RX_SUCCESS;
1951 	}
1952 
1953 	bh_lock_sock(sk);
1954 	if (sk->sk_state == IUCV_CONNECTED) {
1955 		sk->sk_state = IUCV_DISCONN;
1956 		sk->sk_state_change(sk);
1957 	}
1958 	bh_unlock_sock(sk);
1959 	consume_skb(skb);
1960 	return NET_RX_SUCCESS;
1961 }
1962 
1963 /**
1964  * afiucv_hs_callback_win() - react on received WIN
1965  **/
1966 static int afiucv_hs_callback_win(struct sock *sk, struct sk_buff *skb)
1967 {
1968 	struct iucv_sock *iucv = iucv_sk(sk);
1969 
1970 	if (!iucv)
1971 		return NET_RX_SUCCESS;
1972 
1973 	if (sk->sk_state != IUCV_CONNECTED)
1974 		return NET_RX_SUCCESS;
1975 
1976 	atomic_sub(iucv_trans_hdr(skb)->window, &iucv->msg_sent);
1977 	iucv_sock_wake_msglim(sk);
1978 	return NET_RX_SUCCESS;
1979 }
1980 
1981 /**
1982  * afiucv_hs_callback_rx() - react on received data
1983  **/
1984 static int afiucv_hs_callback_rx(struct sock *sk, struct sk_buff *skb)
1985 {
1986 	struct iucv_sock *iucv = iucv_sk(sk);
1987 
1988 	if (!iucv) {
1989 		kfree_skb(skb);
1990 		return NET_RX_SUCCESS;
1991 	}
1992 
1993 	if (sk->sk_state != IUCV_CONNECTED) {
1994 		kfree_skb(skb);
1995 		return NET_RX_SUCCESS;
1996 	}
1997 
1998 	if (sk->sk_shutdown & RCV_SHUTDOWN) {
1999 		kfree_skb(skb);
2000 		return NET_RX_SUCCESS;
2001 	}
2002 
2003 	/* write stuff from iucv_msg to skb cb */
2004 	skb_pull(skb, sizeof(struct af_iucv_trans_hdr));
2005 	skb_reset_transport_header(skb);
2006 	skb_reset_network_header(skb);
2007 	IUCV_SKB_CB(skb)->offset = 0;
2008 	if (sk_filter(sk, skb)) {
2009 		atomic_inc(&sk->sk_drops);	/* skb rejected by filter */
2010 		kfree_skb(skb);
2011 		return NET_RX_SUCCESS;
2012 	}
2013 
2014 	spin_lock(&iucv->message_q.lock);
2015 	if (skb_queue_empty(&iucv->backlog_skb_q)) {
2016 		if (__sock_queue_rcv_skb(sk, skb))
2017 			/* handle rcv queue full */
2018 			skb_queue_tail(&iucv->backlog_skb_q, skb);
2019 	} else
2020 		skb_queue_tail(&iucv_sk(sk)->backlog_skb_q, skb);
2021 	spin_unlock(&iucv->message_q.lock);
2022 	return NET_RX_SUCCESS;
2023 }
2024 
2025 /**
2026  * afiucv_hs_rcv() - base function for arriving data through HiperSockets
2027  *                   transport
2028  *                   called from netif RX softirq
2029  **/
2030 static int afiucv_hs_rcv(struct sk_buff *skb, struct net_device *dev,
2031 	struct packet_type *pt, struct net_device *orig_dev)
2032 {
2033 	struct sock *sk;
2034 	struct iucv_sock *iucv;
2035 	struct af_iucv_trans_hdr *trans_hdr;
2036 	int err = NET_RX_SUCCESS;
2037 	char nullstring[8];
2038 
2039 	if (!pskb_may_pull(skb, sizeof(*trans_hdr))) {
2040 		kfree_skb(skb);
2041 		return NET_RX_SUCCESS;
2042 	}
2043 
2044 	trans_hdr = iucv_trans_hdr(skb);
2045 	EBCASC(trans_hdr->destAppName, sizeof(trans_hdr->destAppName));
2046 	EBCASC(trans_hdr->destUserID, sizeof(trans_hdr->destUserID));
2047 	EBCASC(trans_hdr->srcAppName, sizeof(trans_hdr->srcAppName));
2048 	EBCASC(trans_hdr->srcUserID, sizeof(trans_hdr->srcUserID));
2049 	memset(nullstring, 0, sizeof(nullstring));
2050 	iucv = NULL;
2051 	sk = NULL;
2052 	read_lock(&iucv_sk_list.lock);
2053 	sk_for_each(sk, &iucv_sk_list.head) {
2054 		if (trans_hdr->flags == AF_IUCV_FLAG_SYN) {
2055 			if ((!memcmp(&iucv_sk(sk)->src_name,
2056 				     trans_hdr->destAppName, 8)) &&
2057 			    (!memcmp(&iucv_sk(sk)->src_user_id,
2058 				     trans_hdr->destUserID, 8)) &&
2059 			    (!memcmp(&iucv_sk(sk)->dst_name, nullstring, 8)) &&
2060 			    (!memcmp(&iucv_sk(sk)->dst_user_id,
2061 				     nullstring, 8))) {
2062 				iucv = iucv_sk(sk);
2063 				break;
2064 			}
2065 		} else {
2066 			if ((!memcmp(&iucv_sk(sk)->src_name,
2067 				     trans_hdr->destAppName, 8)) &&
2068 			    (!memcmp(&iucv_sk(sk)->src_user_id,
2069 				     trans_hdr->destUserID, 8)) &&
2070 			    (!memcmp(&iucv_sk(sk)->dst_name,
2071 				     trans_hdr->srcAppName, 8)) &&
2072 			    (!memcmp(&iucv_sk(sk)->dst_user_id,
2073 				     trans_hdr->srcUserID, 8))) {
2074 				iucv = iucv_sk(sk);
2075 				break;
2076 			}
2077 		}
2078 	}
2079 	read_unlock(&iucv_sk_list.lock);
2080 	if (!iucv)
2081 		sk = NULL;
2082 
2083 	/* no sock
2084 	how should we send with no sock
2085 	1) send without sock no send rc checking?
2086 	2) introduce default sock to handle this cases
2087 
2088 	 SYN -> send SYN|ACK in good case, send SYN|FIN in bad case
2089 	 data -> send FIN
2090 	 SYN|ACK, SYN|FIN, FIN -> no action? */
2091 
2092 	switch (trans_hdr->flags) {
2093 	case AF_IUCV_FLAG_SYN:
2094 		/* connect request */
2095 		err = afiucv_hs_callback_syn(sk, skb);
2096 		break;
2097 	case (AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_ACK):
2098 		/* connect request confirmed */
2099 		err = afiucv_hs_callback_synack(sk, skb);
2100 		break;
2101 	case (AF_IUCV_FLAG_SYN | AF_IUCV_FLAG_FIN):
2102 		/* connect request refused */
2103 		err = afiucv_hs_callback_synfin(sk, skb);
2104 		break;
2105 	case (AF_IUCV_FLAG_FIN):
2106 		/* close request */
2107 		err = afiucv_hs_callback_fin(sk, skb);
2108 		break;
2109 	case (AF_IUCV_FLAG_WIN):
2110 		err = afiucv_hs_callback_win(sk, skb);
2111 		if (skb->len == sizeof(struct af_iucv_trans_hdr)) {
2112 			consume_skb(skb);
2113 			break;
2114 		}
2115 		fallthrough;	/* and receive non-zero length data */
2116 	case (AF_IUCV_FLAG_SHT):
2117 		/* shutdown request */
2118 		fallthrough;	/* and receive zero length data */
2119 	case 0:
2120 		/* plain data frame */
2121 		IUCV_SKB_CB(skb)->class = trans_hdr->iucv_hdr.class;
2122 		err = afiucv_hs_callback_rx(sk, skb);
2123 		break;
2124 	default:
2125 		kfree_skb(skb);
2126 	}
2127 
2128 	return err;
2129 }
2130 
2131 /**
2132  * afiucv_hs_callback_txnotify() - handle send notifications from HiperSockets
2133  *                                 transport
2134  **/
2135 static void afiucv_hs_callback_txnotify(struct sock *sk, enum iucv_tx_notify n)
2136 {
2137 	struct iucv_sock *iucv = iucv_sk(sk);
2138 
2139 	if (sock_flag(sk, SOCK_ZAPPED))
2140 		return;
2141 
2142 	switch (n) {
2143 	case TX_NOTIFY_OK:
2144 		atomic_dec(&iucv->skbs_in_xmit);
2145 		iucv_sock_wake_msglim(sk);
2146 		break;
2147 	case TX_NOTIFY_PENDING:
2148 		atomic_inc(&iucv->pendings);
2149 		break;
2150 	case TX_NOTIFY_DELAYED_OK:
2151 		atomic_dec(&iucv->skbs_in_xmit);
2152 		if (atomic_dec_return(&iucv->pendings) <= 0)
2153 			iucv_sock_wake_msglim(sk);
2154 		break;
2155 	default:
2156 		atomic_dec(&iucv->skbs_in_xmit);
2157 		if (sk->sk_state == IUCV_CONNECTED) {
2158 			sk->sk_state = IUCV_DISCONN;
2159 			sk->sk_state_change(sk);
2160 		}
2161 	}
2162 
2163 	if (sk->sk_state == IUCV_CLOSING) {
2164 		if (atomic_read(&iucv->skbs_in_xmit) == 0) {
2165 			sk->sk_state = IUCV_CLOSED;
2166 			sk->sk_state_change(sk);
2167 		}
2168 	}
2169 }
2170 
2171 /*
2172  * afiucv_netdev_event: handle netdev notifier chain events
2173  */
2174 static int afiucv_netdev_event(struct notifier_block *this,
2175 			       unsigned long event, void *ptr)
2176 {
2177 	struct net_device *event_dev = netdev_notifier_info_to_dev(ptr);
2178 	struct sock *sk;
2179 	struct iucv_sock *iucv;
2180 
2181 	switch (event) {
2182 	case NETDEV_REBOOT:
2183 	case NETDEV_GOING_DOWN:
2184 		sk_for_each(sk, &iucv_sk_list.head) {
2185 			iucv = iucv_sk(sk);
2186 			if ((iucv->hs_dev == event_dev) &&
2187 			    (sk->sk_state == IUCV_CONNECTED)) {
2188 				if (event == NETDEV_GOING_DOWN)
2189 					iucv_send_ctrl(sk, AF_IUCV_FLAG_FIN);
2190 				sk->sk_state = IUCV_DISCONN;
2191 				sk->sk_state_change(sk);
2192 			}
2193 		}
2194 		break;
2195 	case NETDEV_DOWN:
2196 	case NETDEV_UNREGISTER:
2197 	default:
2198 		break;
2199 	}
2200 	return NOTIFY_DONE;
2201 }
2202 
2203 static struct notifier_block afiucv_netdev_notifier = {
2204 	.notifier_call = afiucv_netdev_event,
2205 };
2206 
2207 static const struct proto_ops iucv_sock_ops = {
2208 	.family		= PF_IUCV,
2209 	.owner		= THIS_MODULE,
2210 	.release	= iucv_sock_release,
2211 	.bind		= iucv_sock_bind,
2212 	.connect	= iucv_sock_connect,
2213 	.listen		= iucv_sock_listen,
2214 	.accept		= iucv_sock_accept,
2215 	.getname	= iucv_sock_getname,
2216 	.sendmsg	= iucv_sock_sendmsg,
2217 	.recvmsg	= iucv_sock_recvmsg,
2218 	.poll		= iucv_sock_poll,
2219 	.ioctl		= sock_no_ioctl,
2220 	.mmap		= sock_no_mmap,
2221 	.socketpair	= sock_no_socketpair,
2222 	.shutdown	= iucv_sock_shutdown,
2223 	.setsockopt	= iucv_sock_setsockopt,
2224 	.getsockopt	= iucv_sock_getsockopt,
2225 };
2226 
2227 static int iucv_sock_create(struct net *net, struct socket *sock, int protocol,
2228 			    int kern)
2229 {
2230 	struct sock *sk;
2231 
2232 	if (protocol && protocol != PF_IUCV)
2233 		return -EPROTONOSUPPORT;
2234 
2235 	sock->state = SS_UNCONNECTED;
2236 
2237 	switch (sock->type) {
2238 	case SOCK_STREAM:
2239 	case SOCK_SEQPACKET:
2240 		/* currently, proto ops can handle both sk types */
2241 		sock->ops = &iucv_sock_ops;
2242 		break;
2243 	default:
2244 		return -ESOCKTNOSUPPORT;
2245 	}
2246 
2247 	sk = iucv_sock_alloc(sock, protocol, GFP_KERNEL, kern);
2248 	if (!sk)
2249 		return -ENOMEM;
2250 
2251 	iucv_sock_init(sk, NULL);
2252 
2253 	return 0;
2254 }
2255 
2256 static const struct net_proto_family iucv_sock_family_ops = {
2257 	.family	= AF_IUCV,
2258 	.owner	= THIS_MODULE,
2259 	.create	= iucv_sock_create,
2260 };
2261 
2262 static struct packet_type iucv_packet_type = {
2263 	.type = cpu_to_be16(ETH_P_AF_IUCV),
2264 	.func = afiucv_hs_rcv,
2265 };
2266 
2267 static int __init afiucv_init(void)
2268 {
2269 	int err;
2270 
2271 	if (MACHINE_IS_VM && IS_ENABLED(CONFIG_IUCV)) {
2272 		cpcmd("QUERY USERID", iucv_userid, sizeof(iucv_userid), &err);
2273 		if (unlikely(err)) {
2274 			WARN_ON(err);
2275 			err = -EPROTONOSUPPORT;
2276 			goto out;
2277 		}
2278 
2279 		pr_iucv = &iucv_if;
2280 	} else {
2281 		memset(&iucv_userid, 0, sizeof(iucv_userid));
2282 		pr_iucv = NULL;
2283 	}
2284 
2285 	err = proto_register(&iucv_proto, 0);
2286 	if (err)
2287 		goto out;
2288 	err = sock_register(&iucv_sock_family_ops);
2289 	if (err)
2290 		goto out_proto;
2291 
2292 	if (pr_iucv) {
2293 		err = pr_iucv->iucv_register(&af_iucv_handler, 0);
2294 		if (err)
2295 			goto out_sock;
2296 	}
2297 
2298 	err = register_netdevice_notifier(&afiucv_netdev_notifier);
2299 	if (err)
2300 		goto out_notifier;
2301 
2302 	dev_add_pack(&iucv_packet_type);
2303 	return 0;
2304 
2305 out_notifier:
2306 	if (pr_iucv)
2307 		pr_iucv->iucv_unregister(&af_iucv_handler, 0);
2308 out_sock:
2309 	sock_unregister(PF_IUCV);
2310 out_proto:
2311 	proto_unregister(&iucv_proto);
2312 out:
2313 	return err;
2314 }
2315 
2316 static void __exit afiucv_exit(void)
2317 {
2318 	if (pr_iucv)
2319 		pr_iucv->iucv_unregister(&af_iucv_handler, 0);
2320 
2321 	unregister_netdevice_notifier(&afiucv_netdev_notifier);
2322 	dev_remove_pack(&iucv_packet_type);
2323 	sock_unregister(PF_IUCV);
2324 	proto_unregister(&iucv_proto);
2325 }
2326 
2327 module_init(afiucv_init);
2328 module_exit(afiucv_exit);
2329 
2330 MODULE_AUTHOR("Jennifer Hunt <jenhunt@us.ibm.com>");
2331 MODULE_DESCRIPTION("IUCV Sockets ver " VERSION);
2332 MODULE_VERSION(VERSION);
2333 MODULE_LICENSE("GPL");
2334 MODULE_ALIAS_NETPROTO(PF_IUCV);
2335