xref: /linux/net/9p/trans_fd.c (revision 8fa5723aa7e053d498336b48448b292fc2e0458b)
1 /*
2  * linux/fs/9p/trans_fd.c
3  *
4  * Fd transport layer.  Includes deprecated socket layer.
5  *
6  *  Copyright (C) 2006 by Russ Cox <rsc@swtch.com>
7  *  Copyright (C) 2004-2005 by Latchesar Ionkov <lucho@ionkov.net>
8  *  Copyright (C) 2004-2008 by Eric Van Hensbergen <ericvh@gmail.com>
9  *  Copyright (C) 1997-2002 by Ron Minnich <rminnich@sarnoff.com>
10  *
11  *  This program is free software; you can redistribute it and/or modify
12  *  it under the terms of the GNU General Public License version 2
13  *  as published by the Free Software Foundation.
14  *
15  *  This program is distributed in the hope that it will be useful,
16  *  but WITHOUT ANY WARRANTY; without even the implied warranty of
17  *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
18  *  GNU General Public License for more details.
19  *
20  *  You should have received a copy of the GNU General Public License
21  *  along with this program; if not, write to:
22  *  Free Software Foundation
23  *  51 Franklin Street, Fifth Floor
24  *  Boston, MA  02111-1301  USA
25  *
26  */
27 
28 #include <linux/in.h>
29 #include <linux/module.h>
30 #include <linux/net.h>
31 #include <linux/ipv6.h>
32 #include <linux/kthread.h>
33 #include <linux/errno.h>
34 #include <linux/kernel.h>
35 #include <linux/un.h>
36 #include <linux/uaccess.h>
37 #include <linux/inet.h>
38 #include <linux/idr.h>
39 #include <linux/file.h>
40 #include <linux/parser.h>
41 #include <net/9p/9p.h>
42 #include <net/9p/client.h>
43 #include <net/9p/transport.h>
44 
45 #define P9_PORT 564
46 #define MAX_SOCK_BUF (64*1024)
47 #define MAXPOLLWADDR	2
48 
49 /**
50  * struct p9_fd_opts - per-transport options
51  * @rfd: file descriptor for reading (trans=fd)
52  * @wfd: file descriptor for writing (trans=fd)
53  * @port: port to connect to (trans=tcp)
54  *
55  */
56 
57 struct p9_fd_opts {
58 	int rfd;
59 	int wfd;
60 	u16 port;
61 };
62 
63 /**
64  * struct p9_trans_fd - transport state
65  * @rd: reference to file to read from
66  * @wr: reference of file to write to
67  * @conn: connection state reference
68  *
69  */
70 
71 struct p9_trans_fd {
72 	struct file *rd;
73 	struct file *wr;
74 	struct p9_conn *conn;
75 };
76 
77 /*
78   * Option Parsing (code inspired by NFS code)
79   *  - a little lazy - parse all fd-transport options
80   */
81 
82 enum {
83 	/* Options that take integer arguments */
84 	Opt_port, Opt_rfdno, Opt_wfdno, Opt_err,
85 };
86 
87 static const match_table_t tokens = {
88 	{Opt_port, "port=%u"},
89 	{Opt_rfdno, "rfdno=%u"},
90 	{Opt_wfdno, "wfdno=%u"},
91 	{Opt_err, NULL},
92 };
93 
94 enum {
95 	Rworksched = 1,		/* read work scheduled or running */
96 	Rpending = 2,		/* can read */
97 	Wworksched = 4,		/* write work scheduled or running */
98 	Wpending = 8,		/* can write */
99 };
100 
101 struct p9_poll_wait {
102 	struct p9_conn *conn;
103 	wait_queue_t wait;
104 	wait_queue_head_t *wait_addr;
105 };
106 
107 /**
108  * struct p9_conn - fd mux connection state information
109  * @mux_list: list link for mux to manage multiple connections (?)
110  * @client: reference to client instance for this connection
111  * @err: error state
112  * @req_list: accounting for requests which have been sent
113  * @unsent_req_list: accounting for requests that haven't been sent
114  * @req: current request being processed (if any)
115  * @tmp_buf: temporary buffer to read in header
116  * @rsize: amount to read for current frame
117  * @rpos: read position in current frame
118  * @rbuf: current read buffer
119  * @wpos: write position for current frame
120  * @wsize: amount of data to write for current frame
121  * @wbuf: current write buffer
122  * @poll_wait: array of wait_q's for various worker threads
123  * @poll_waddr: ????
124  * @pt: poll state
125  * @rq: current read work
126  * @wq: current write work
127  * @wsched: ????
128  *
129  */
130 
131 struct p9_conn {
132 	struct list_head mux_list;
133 	struct p9_client *client;
134 	int err;
135 	struct list_head req_list;
136 	struct list_head unsent_req_list;
137 	struct p9_req_t *req;
138 	char tmp_buf[7];
139 	int rsize;
140 	int rpos;
141 	char *rbuf;
142 	int wpos;
143 	int wsize;
144 	char *wbuf;
145 	struct list_head poll_pending_link;
146 	struct p9_poll_wait poll_wait[MAXPOLLWADDR];
147 	poll_table pt;
148 	struct work_struct rq;
149 	struct work_struct wq;
150 	unsigned long wsched;
151 };
152 
153 static DEFINE_SPINLOCK(p9_poll_lock);
154 static LIST_HEAD(p9_poll_pending_list);
155 static struct workqueue_struct *p9_mux_wq;
156 static struct task_struct *p9_poll_task;
157 
158 static void p9_mux_poll_stop(struct p9_conn *m)
159 {
160 	unsigned long flags;
161 	int i;
162 
163 	for (i = 0; i < ARRAY_SIZE(m->poll_wait); i++) {
164 		struct p9_poll_wait *pwait = &m->poll_wait[i];
165 
166 		if (pwait->wait_addr) {
167 			remove_wait_queue(pwait->wait_addr, &pwait->wait);
168 			pwait->wait_addr = NULL;
169 		}
170 	}
171 
172 	spin_lock_irqsave(&p9_poll_lock, flags);
173 	list_del_init(&m->poll_pending_link);
174 	spin_unlock_irqrestore(&p9_poll_lock, flags);
175 }
176 
177 /**
178  * p9_conn_cancel - cancel all pending requests with error
179  * @m: mux data
180  * @err: error code
181  *
182  */
183 
184 static void p9_conn_cancel(struct p9_conn *m, int err)
185 {
186 	struct p9_req_t *req, *rtmp;
187 	unsigned long flags;
188 	LIST_HEAD(cancel_list);
189 
190 	P9_DPRINTK(P9_DEBUG_ERROR, "mux %p err %d\n", m, err);
191 
192 	spin_lock_irqsave(&m->client->lock, flags);
193 
194 	if (m->err) {
195 		spin_unlock_irqrestore(&m->client->lock, flags);
196 		return;
197 	}
198 
199 	m->err = err;
200 
201 	list_for_each_entry_safe(req, rtmp, &m->req_list, req_list) {
202 		req->status = REQ_STATUS_ERROR;
203 		if (!req->t_err)
204 			req->t_err = err;
205 		list_move(&req->req_list, &cancel_list);
206 	}
207 	list_for_each_entry_safe(req, rtmp, &m->unsent_req_list, req_list) {
208 		req->status = REQ_STATUS_ERROR;
209 		if (!req->t_err)
210 			req->t_err = err;
211 		list_move(&req->req_list, &cancel_list);
212 	}
213 	spin_unlock_irqrestore(&m->client->lock, flags);
214 
215 	list_for_each_entry_safe(req, rtmp, &cancel_list, req_list) {
216 		list_del(&req->req_list);
217 		P9_DPRINTK(P9_DEBUG_ERROR, "call back req %p\n", req);
218 		p9_client_cb(m->client, req);
219 	}
220 }
221 
222 static unsigned int
223 p9_fd_poll(struct p9_client *client, struct poll_table_struct *pt)
224 {
225 	int ret, n;
226 	struct p9_trans_fd *ts = NULL;
227 
228 	if (client && client->status == Connected)
229 		ts = client->trans;
230 
231 	if (!ts)
232 		return -EREMOTEIO;
233 
234 	if (!ts->rd->f_op || !ts->rd->f_op->poll)
235 		return -EIO;
236 
237 	if (!ts->wr->f_op || !ts->wr->f_op->poll)
238 		return -EIO;
239 
240 	ret = ts->rd->f_op->poll(ts->rd, pt);
241 	if (ret < 0)
242 		return ret;
243 
244 	if (ts->rd != ts->wr) {
245 		n = ts->wr->f_op->poll(ts->wr, pt);
246 		if (n < 0)
247 			return n;
248 		ret = (ret & ~POLLOUT) | (n & ~POLLIN);
249 	}
250 
251 	return ret;
252 }
253 
254 /**
255  * p9_fd_read- read from a fd
256  * @client: client instance
257  * @v: buffer to receive data into
258  * @len: size of receive buffer
259  *
260  */
261 
262 static int p9_fd_read(struct p9_client *client, void *v, int len)
263 {
264 	int ret;
265 	struct p9_trans_fd *ts = NULL;
266 
267 	if (client && client->status != Disconnected)
268 		ts = client->trans;
269 
270 	if (!ts)
271 		return -EREMOTEIO;
272 
273 	if (!(ts->rd->f_flags & O_NONBLOCK))
274 		P9_DPRINTK(P9_DEBUG_ERROR, "blocking read ...\n");
275 
276 	ret = kernel_read(ts->rd, ts->rd->f_pos, v, len);
277 	if (ret <= 0 && ret != -ERESTARTSYS && ret != -EAGAIN)
278 		client->status = Disconnected;
279 	return ret;
280 }
281 
282 /**
283  * p9_read_work - called when there is some data to be read from a transport
284  * @work: container of work to be done
285  *
286  */
287 
288 static void p9_read_work(struct work_struct *work)
289 {
290 	int n, err;
291 	struct p9_conn *m;
292 
293 	m = container_of(work, struct p9_conn, rq);
294 
295 	if (m->err < 0)
296 		return;
297 
298 	P9_DPRINTK(P9_DEBUG_TRANS, "start mux %p pos %d\n", m, m->rpos);
299 
300 	if (!m->rbuf) {
301 		m->rbuf = m->tmp_buf;
302 		m->rpos = 0;
303 		m->rsize = 7; /* start by reading header */
304 	}
305 
306 	clear_bit(Rpending, &m->wsched);
307 	P9_DPRINTK(P9_DEBUG_TRANS, "read mux %p pos %d size: %d = %d\n", m,
308 					m->rpos, m->rsize, m->rsize-m->rpos);
309 	err = p9_fd_read(m->client, m->rbuf + m->rpos,
310 						m->rsize - m->rpos);
311 	P9_DPRINTK(P9_DEBUG_TRANS, "mux %p got %d bytes\n", m, err);
312 	if (err == -EAGAIN) {
313 		clear_bit(Rworksched, &m->wsched);
314 		return;
315 	}
316 
317 	if (err <= 0)
318 		goto error;
319 
320 	m->rpos += err;
321 
322 	if ((!m->req) && (m->rpos == m->rsize)) { /* header read in */
323 		u16 tag;
324 		P9_DPRINTK(P9_DEBUG_TRANS, "got new header\n");
325 
326 		n = le32_to_cpu(*(__le32 *) m->rbuf); /* read packet size */
327 		if (n >= m->client->msize) {
328 			P9_DPRINTK(P9_DEBUG_ERROR,
329 				"requested packet size too big: %d\n", n);
330 			err = -EIO;
331 			goto error;
332 		}
333 
334 		tag = le16_to_cpu(*(__le16 *) (m->rbuf+5)); /* read tag */
335 		P9_DPRINTK(P9_DEBUG_TRANS,
336 			"mux %p pkt: size: %d bytes tag: %d\n", m, n, tag);
337 
338 		m->req = p9_tag_lookup(m->client, tag);
339 		if (!m->req) {
340 			P9_DPRINTK(P9_DEBUG_ERROR, "Unexpected packet tag %d\n",
341 								 tag);
342 			err = -EIO;
343 			goto error;
344 		}
345 
346 		if (m->req->rc == NULL) {
347 			m->req->rc = kmalloc(sizeof(struct p9_fcall) +
348 						m->client->msize, GFP_KERNEL);
349 			if (!m->req->rc) {
350 				m->req = NULL;
351 				err = -ENOMEM;
352 				goto error;
353 			}
354 		}
355 		m->rbuf = (char *)m->req->rc + sizeof(struct p9_fcall);
356 		memcpy(m->rbuf, m->tmp_buf, m->rsize);
357 		m->rsize = n;
358 	}
359 
360 	/* not an else because some packets (like clunk) have no payload */
361 	if ((m->req) && (m->rpos == m->rsize)) { /* packet is read in */
362 		P9_DPRINTK(P9_DEBUG_TRANS, "got new packet\n");
363 		spin_lock(&m->client->lock);
364 		list_del(&m->req->req_list);
365 		spin_unlock(&m->client->lock);
366 		p9_client_cb(m->client, m->req);
367 
368 		m->rbuf = NULL;
369 		m->rpos = 0;
370 		m->rsize = 0;
371 		m->req = NULL;
372 	}
373 
374 	if (!list_empty(&m->req_list)) {
375 		if (test_and_clear_bit(Rpending, &m->wsched))
376 			n = POLLIN;
377 		else
378 			n = p9_fd_poll(m->client, NULL);
379 
380 		if (n & POLLIN) {
381 			P9_DPRINTK(P9_DEBUG_TRANS, "sched read work %p\n", m);
382 			queue_work(p9_mux_wq, &m->rq);
383 		} else
384 			clear_bit(Rworksched, &m->wsched);
385 	} else
386 		clear_bit(Rworksched, &m->wsched);
387 
388 	return;
389 error:
390 	p9_conn_cancel(m, err);
391 	clear_bit(Rworksched, &m->wsched);
392 }
393 
394 /**
395  * p9_fd_write - write to a socket
396  * @client: client instance
397  * @v: buffer to send data from
398  * @len: size of send buffer
399  *
400  */
401 
402 static int p9_fd_write(struct p9_client *client, void *v, int len)
403 {
404 	int ret;
405 	mm_segment_t oldfs;
406 	struct p9_trans_fd *ts = NULL;
407 
408 	if (client && client->status != Disconnected)
409 		ts = client->trans;
410 
411 	if (!ts)
412 		return -EREMOTEIO;
413 
414 	if (!(ts->wr->f_flags & O_NONBLOCK))
415 		P9_DPRINTK(P9_DEBUG_ERROR, "blocking write ...\n");
416 
417 	oldfs = get_fs();
418 	set_fs(get_ds());
419 	/* The cast to a user pointer is valid due to the set_fs() */
420 	ret = vfs_write(ts->wr, (void __user *)v, len, &ts->wr->f_pos);
421 	set_fs(oldfs);
422 
423 	if (ret <= 0 && ret != -ERESTARTSYS && ret != -EAGAIN)
424 		client->status = Disconnected;
425 	return ret;
426 }
427 
428 /**
429  * p9_write_work - called when a transport can send some data
430  * @work: container for work to be done
431  *
432  */
433 
434 static void p9_write_work(struct work_struct *work)
435 {
436 	int n, err;
437 	struct p9_conn *m;
438 	struct p9_req_t *req;
439 
440 	m = container_of(work, struct p9_conn, wq);
441 
442 	if (m->err < 0) {
443 		clear_bit(Wworksched, &m->wsched);
444 		return;
445 	}
446 
447 	if (!m->wsize) {
448 		if (list_empty(&m->unsent_req_list)) {
449 			clear_bit(Wworksched, &m->wsched);
450 			return;
451 		}
452 
453 		spin_lock(&m->client->lock);
454 		req = list_entry(m->unsent_req_list.next, struct p9_req_t,
455 			       req_list);
456 		req->status = REQ_STATUS_SENT;
457 		list_move_tail(&req->req_list, &m->req_list);
458 
459 		m->wbuf = req->tc->sdata;
460 		m->wsize = req->tc->size;
461 		m->wpos = 0;
462 		spin_unlock(&m->client->lock);
463 	}
464 
465 	P9_DPRINTK(P9_DEBUG_TRANS, "mux %p pos %d size %d\n", m, m->wpos,
466 								m->wsize);
467 	clear_bit(Wpending, &m->wsched);
468 	err = p9_fd_write(m->client, m->wbuf + m->wpos, m->wsize - m->wpos);
469 	P9_DPRINTK(P9_DEBUG_TRANS, "mux %p sent %d bytes\n", m, err);
470 	if (err == -EAGAIN) {
471 		clear_bit(Wworksched, &m->wsched);
472 		return;
473 	}
474 
475 	if (err < 0)
476 		goto error;
477 	else if (err == 0) {
478 		err = -EREMOTEIO;
479 		goto error;
480 	}
481 
482 	m->wpos += err;
483 	if (m->wpos == m->wsize)
484 		m->wpos = m->wsize = 0;
485 
486 	if (m->wsize == 0 && !list_empty(&m->unsent_req_list)) {
487 		if (test_and_clear_bit(Wpending, &m->wsched))
488 			n = POLLOUT;
489 		else
490 			n = p9_fd_poll(m->client, NULL);
491 
492 		if (n & POLLOUT) {
493 			P9_DPRINTK(P9_DEBUG_TRANS, "sched write work %p\n", m);
494 			queue_work(p9_mux_wq, &m->wq);
495 		} else
496 			clear_bit(Wworksched, &m->wsched);
497 	} else
498 		clear_bit(Wworksched, &m->wsched);
499 
500 	return;
501 
502 error:
503 	p9_conn_cancel(m, err);
504 	clear_bit(Wworksched, &m->wsched);
505 }
506 
507 static int p9_pollwake(wait_queue_t *wait, unsigned mode, int sync, void *key)
508 {
509 	struct p9_poll_wait *pwait =
510 		container_of(wait, struct p9_poll_wait, wait);
511 	struct p9_conn *m = pwait->conn;
512 	unsigned long flags;
513 	DECLARE_WAITQUEUE(dummy_wait, p9_poll_task);
514 
515 	spin_lock_irqsave(&p9_poll_lock, flags);
516 	if (list_empty(&m->poll_pending_link))
517 		list_add_tail(&m->poll_pending_link, &p9_poll_pending_list);
518 	spin_unlock_irqrestore(&p9_poll_lock, flags);
519 
520 	/* perform the default wake up operation */
521 	return default_wake_function(&dummy_wait, mode, sync, key);
522 }
523 
524 /**
525  * p9_pollwait - add poll task to the wait queue
526  * @filp: file pointer being polled
527  * @wait_address: wait_q to block on
528  * @p: poll state
529  *
530  * called by files poll operation to add v9fs-poll task to files wait queue
531  */
532 
533 static void
534 p9_pollwait(struct file *filp, wait_queue_head_t *wait_address, poll_table *p)
535 {
536 	struct p9_conn *m = container_of(p, struct p9_conn, pt);
537 	struct p9_poll_wait *pwait = NULL;
538 	int i;
539 
540 	for (i = 0; i < ARRAY_SIZE(m->poll_wait); i++) {
541 		if (m->poll_wait[i].wait_addr == NULL) {
542 			pwait = &m->poll_wait[i];
543 			break;
544 		}
545 	}
546 
547 	if (!pwait) {
548 		P9_DPRINTK(P9_DEBUG_ERROR, "not enough wait_address slots\n");
549 		return;
550 	}
551 
552 	pwait->conn = m;
553 	pwait->wait_addr = wait_address;
554 	init_waitqueue_func_entry(&pwait->wait, p9_pollwake);
555 	add_wait_queue(wait_address, &pwait->wait);
556 }
557 
558 /**
559  * p9_conn_create - allocate and initialize the per-session mux data
560  * @client: client instance
561  *
562  * Note: Creates the polling task if this is the first session.
563  */
564 
565 static struct p9_conn *p9_conn_create(struct p9_client *client)
566 {
567 	int n;
568 	struct p9_conn *m;
569 
570 	P9_DPRINTK(P9_DEBUG_TRANS, "client %p msize %d\n", client,
571 								client->msize);
572 	m = kzalloc(sizeof(struct p9_conn), GFP_KERNEL);
573 	if (!m)
574 		return ERR_PTR(-ENOMEM);
575 
576 	INIT_LIST_HEAD(&m->mux_list);
577 	m->client = client;
578 
579 	INIT_LIST_HEAD(&m->req_list);
580 	INIT_LIST_HEAD(&m->unsent_req_list);
581 	INIT_WORK(&m->rq, p9_read_work);
582 	INIT_WORK(&m->wq, p9_write_work);
583 	INIT_LIST_HEAD(&m->poll_pending_link);
584 	init_poll_funcptr(&m->pt, p9_pollwait);
585 
586 	n = p9_fd_poll(client, &m->pt);
587 	if (n & POLLIN) {
588 		P9_DPRINTK(P9_DEBUG_TRANS, "mux %p can read\n", m);
589 		set_bit(Rpending, &m->wsched);
590 	}
591 
592 	if (n & POLLOUT) {
593 		P9_DPRINTK(P9_DEBUG_TRANS, "mux %p can write\n", m);
594 		set_bit(Wpending, &m->wsched);
595 	}
596 
597 	return m;
598 }
599 
600 /**
601  * p9_poll_mux - polls a mux and schedules read or write works if necessary
602  * @m: connection to poll
603  *
604  */
605 
606 static void p9_poll_mux(struct p9_conn *m)
607 {
608 	int n;
609 
610 	if (m->err < 0)
611 		return;
612 
613 	n = p9_fd_poll(m->client, NULL);
614 	if (n < 0 || n & (POLLERR | POLLHUP | POLLNVAL)) {
615 		P9_DPRINTK(P9_DEBUG_TRANS, "error mux %p err %d\n", m, n);
616 		if (n >= 0)
617 			n = -ECONNRESET;
618 		p9_conn_cancel(m, n);
619 	}
620 
621 	if (n & POLLIN) {
622 		set_bit(Rpending, &m->wsched);
623 		P9_DPRINTK(P9_DEBUG_TRANS, "mux %p can read\n", m);
624 		if (!test_and_set_bit(Rworksched, &m->wsched)) {
625 			P9_DPRINTK(P9_DEBUG_TRANS, "sched read work %p\n", m);
626 			queue_work(p9_mux_wq, &m->rq);
627 		}
628 	}
629 
630 	if (n & POLLOUT) {
631 		set_bit(Wpending, &m->wsched);
632 		P9_DPRINTK(P9_DEBUG_TRANS, "mux %p can write\n", m);
633 		if ((m->wsize || !list_empty(&m->unsent_req_list))
634 		    && !test_and_set_bit(Wworksched, &m->wsched)) {
635 			P9_DPRINTK(P9_DEBUG_TRANS, "sched write work %p\n", m);
636 			queue_work(p9_mux_wq, &m->wq);
637 		}
638 	}
639 }
640 
641 /**
642  * p9_fd_request - send 9P request
643  * The function can sleep until the request is scheduled for sending.
644  * The function can be interrupted. Return from the function is not
645  * a guarantee that the request is sent successfully.
646  *
647  * @client: client instance
648  * @req: request to be sent
649  *
650  */
651 
652 static int p9_fd_request(struct p9_client *client, struct p9_req_t *req)
653 {
654 	int n;
655 	struct p9_trans_fd *ts = client->trans;
656 	struct p9_conn *m = ts->conn;
657 
658 	P9_DPRINTK(P9_DEBUG_TRANS, "mux %p task %p tcall %p id %d\n", m,
659 						current, req->tc, req->tc->id);
660 	if (m->err < 0)
661 		return m->err;
662 
663 	spin_lock(&client->lock);
664 	req->status = REQ_STATUS_UNSENT;
665 	list_add_tail(&req->req_list, &m->unsent_req_list);
666 	spin_unlock(&client->lock);
667 
668 	if (test_and_clear_bit(Wpending, &m->wsched))
669 		n = POLLOUT;
670 	else
671 		n = p9_fd_poll(m->client, NULL);
672 
673 	if (n & POLLOUT && !test_and_set_bit(Wworksched, &m->wsched))
674 		queue_work(p9_mux_wq, &m->wq);
675 
676 	return 0;
677 }
678 
679 static int p9_fd_cancel(struct p9_client *client, struct p9_req_t *req)
680 {
681 	int ret = 1;
682 
683 	P9_DPRINTK(P9_DEBUG_TRANS, "client %p req %p\n", client, req);
684 
685 	spin_lock(&client->lock);
686 	list_del(&req->req_list);
687 
688 	if (req->status == REQ_STATUS_UNSENT) {
689 		req->status = REQ_STATUS_FLSHD;
690 		ret = 0;
691 	}
692 
693 	spin_unlock(&client->lock);
694 
695 	return ret;
696 }
697 
698 /**
699  * parse_options - parse mount options into session structure
700  * @options: options string passed from mount
701  * @opts: transport-specific structure to parse options into
702  *
703  * Returns 0 upon success, -ERRNO upon failure
704  */
705 
706 static int parse_opts(char *params, struct p9_fd_opts *opts)
707 {
708 	char *p;
709 	substring_t args[MAX_OPT_ARGS];
710 	int option;
711 	char *options;
712 	int ret;
713 
714 	opts->port = P9_PORT;
715 	opts->rfd = ~0;
716 	opts->wfd = ~0;
717 
718 	if (!params)
719 		return 0;
720 
721 	options = kstrdup(params, GFP_KERNEL);
722 	if (!options) {
723 		P9_DPRINTK(P9_DEBUG_ERROR,
724 				"failed to allocate copy of option string\n");
725 		return -ENOMEM;
726 	}
727 
728 	while ((p = strsep(&options, ",")) != NULL) {
729 		int token;
730 		int r;
731 		if (!*p)
732 			continue;
733 		token = match_token(p, tokens, args);
734 		r = match_int(&args[0], &option);
735 		if (r < 0) {
736 			P9_DPRINTK(P9_DEBUG_ERROR,
737 			 "integer field, but no integer?\n");
738 			ret = r;
739 			continue;
740 		}
741 		switch (token) {
742 		case Opt_port:
743 			opts->port = option;
744 			break;
745 		case Opt_rfdno:
746 			opts->rfd = option;
747 			break;
748 		case Opt_wfdno:
749 			opts->wfd = option;
750 			break;
751 		default:
752 			continue;
753 		}
754 	}
755 	kfree(options);
756 	return 0;
757 }
758 
759 static int p9_fd_open(struct p9_client *client, int rfd, int wfd)
760 {
761 	struct p9_trans_fd *ts = kmalloc(sizeof(struct p9_trans_fd),
762 					   GFP_KERNEL);
763 	if (!ts)
764 		return -ENOMEM;
765 
766 	ts->rd = fget(rfd);
767 	ts->wr = fget(wfd);
768 	if (!ts->rd || !ts->wr) {
769 		if (ts->rd)
770 			fput(ts->rd);
771 		if (ts->wr)
772 			fput(ts->wr);
773 		kfree(ts);
774 		return -EIO;
775 	}
776 
777 	client->trans = ts;
778 	client->status = Connected;
779 
780 	return 0;
781 }
782 
783 static int p9_socket_open(struct p9_client *client, struct socket *csocket)
784 {
785 	int fd, ret;
786 
787 	csocket->sk->sk_allocation = GFP_NOIO;
788 	fd = sock_map_fd(csocket, 0);
789 	if (fd < 0) {
790 		P9_EPRINTK(KERN_ERR, "p9_socket_open: failed to map fd\n");
791 		return fd;
792 	}
793 
794 	ret = p9_fd_open(client, fd, fd);
795 	if (ret < 0) {
796 		P9_EPRINTK(KERN_ERR, "p9_socket_open: failed to open fd\n");
797 		sockfd_put(csocket);
798 		return ret;
799 	}
800 
801 	((struct p9_trans_fd *)client->trans)->rd->f_flags |= O_NONBLOCK;
802 
803 	return 0;
804 }
805 
806 /**
807  * p9_mux_destroy - cancels all pending requests and frees mux resources
808  * @m: mux to destroy
809  *
810  */
811 
812 static void p9_conn_destroy(struct p9_conn *m)
813 {
814 	P9_DPRINTK(P9_DEBUG_TRANS, "mux %p prev %p next %p\n", m,
815 		m->mux_list.prev, m->mux_list.next);
816 
817 	p9_mux_poll_stop(m);
818 	cancel_work_sync(&m->rq);
819 	cancel_work_sync(&m->wq);
820 
821 	p9_conn_cancel(m, -ECONNRESET);
822 
823 	m->client = NULL;
824 	kfree(m);
825 }
826 
827 /**
828  * p9_fd_close - shutdown file descriptor transport
829  * @client: client instance
830  *
831  */
832 
833 static void p9_fd_close(struct p9_client *client)
834 {
835 	struct p9_trans_fd *ts;
836 
837 	if (!client)
838 		return;
839 
840 	ts = client->trans;
841 	if (!ts)
842 		return;
843 
844 	client->status = Disconnected;
845 
846 	p9_conn_destroy(ts->conn);
847 
848 	if (ts->rd)
849 		fput(ts->rd);
850 	if (ts->wr)
851 		fput(ts->wr);
852 
853 	kfree(ts);
854 }
855 
856 /*
857  * stolen from NFS - maybe should be made a generic function?
858  */
859 static inline int valid_ipaddr4(const char *buf)
860 {
861 	int rc, count, in[4];
862 
863 	rc = sscanf(buf, "%d.%d.%d.%d", &in[0], &in[1], &in[2], &in[3]);
864 	if (rc != 4)
865 		return -EINVAL;
866 	for (count = 0; count < 4; count++) {
867 		if (in[count] > 255)
868 			return -EINVAL;
869 	}
870 	return 0;
871 }
872 
873 static int
874 p9_fd_create_tcp(struct p9_client *client, const char *addr, char *args)
875 {
876 	int err;
877 	struct socket *csocket;
878 	struct sockaddr_in sin_server;
879 	struct p9_fd_opts opts;
880 	struct p9_trans_fd *p = NULL; /* this gets allocated in p9_fd_open */
881 
882 	err = parse_opts(args, &opts);
883 	if (err < 0)
884 		return err;
885 
886 	if (valid_ipaddr4(addr) < 0)
887 		return -EINVAL;
888 
889 	csocket = NULL;
890 
891 	sin_server.sin_family = AF_INET;
892 	sin_server.sin_addr.s_addr = in_aton(addr);
893 	sin_server.sin_port = htons(opts.port);
894 	sock_create_kern(PF_INET, SOCK_STREAM, IPPROTO_TCP, &csocket);
895 
896 	if (!csocket) {
897 		P9_EPRINTK(KERN_ERR, "p9_trans_tcp: problem creating socket\n");
898 		err = -EIO;
899 		goto error;
900 	}
901 
902 	err = csocket->ops->connect(csocket,
903 				    (struct sockaddr *)&sin_server,
904 				    sizeof(struct sockaddr_in), 0);
905 	if (err < 0) {
906 		P9_EPRINTK(KERN_ERR,
907 			"p9_trans_tcp: problem connecting socket to %s\n",
908 			addr);
909 		goto error;
910 	}
911 
912 	err = p9_socket_open(client, csocket);
913 	if (err < 0)
914 		goto error;
915 
916 	p = (struct p9_trans_fd *) client->trans;
917 	p->conn = p9_conn_create(client);
918 	if (IS_ERR(p->conn)) {
919 		err = PTR_ERR(p->conn);
920 		p->conn = NULL;
921 		goto error;
922 	}
923 
924 	return 0;
925 
926 error:
927 	if (csocket)
928 		sock_release(csocket);
929 
930 	kfree(p);
931 
932 	return err;
933 }
934 
935 static int
936 p9_fd_create_unix(struct p9_client *client, const char *addr, char *args)
937 {
938 	int err;
939 	struct socket *csocket;
940 	struct sockaddr_un sun_server;
941 	struct p9_trans_fd *p = NULL; /* this gets allocated in p9_fd_open */
942 
943 	csocket = NULL;
944 
945 	if (strlen(addr) > UNIX_PATH_MAX) {
946 		P9_EPRINTK(KERN_ERR, "p9_trans_unix: address too long: %s\n",
947 			addr);
948 		err = -ENAMETOOLONG;
949 		goto error;
950 	}
951 
952 	sun_server.sun_family = PF_UNIX;
953 	strcpy(sun_server.sun_path, addr);
954 	sock_create_kern(PF_UNIX, SOCK_STREAM, 0, &csocket);
955 	err = csocket->ops->connect(csocket, (struct sockaddr *)&sun_server,
956 			sizeof(struct sockaddr_un) - 1, 0);
957 	if (err < 0) {
958 		P9_EPRINTK(KERN_ERR,
959 			"p9_trans_unix: problem connecting socket: %s: %d\n",
960 			addr, err);
961 		goto error;
962 	}
963 
964 	err = p9_socket_open(client, csocket);
965 	if (err < 0)
966 		goto error;
967 
968 	p = (struct p9_trans_fd *) client->trans;
969 	p->conn = p9_conn_create(client);
970 	if (IS_ERR(p->conn)) {
971 		err = PTR_ERR(p->conn);
972 		p->conn = NULL;
973 		goto error;
974 	}
975 
976 	return 0;
977 
978 error:
979 	if (csocket)
980 		sock_release(csocket);
981 
982 	kfree(p);
983 	return err;
984 }
985 
986 static int
987 p9_fd_create(struct p9_client *client, const char *addr, char *args)
988 {
989 	int err;
990 	struct p9_fd_opts opts;
991 	struct p9_trans_fd *p = NULL; /* this get allocated in p9_fd_open */
992 
993 	parse_opts(args, &opts);
994 
995 	if (opts.rfd == ~0 || opts.wfd == ~0) {
996 		printk(KERN_ERR "v9fs: Insufficient options for proto=fd\n");
997 		return -ENOPROTOOPT;
998 	}
999 
1000 	err = p9_fd_open(client, opts.rfd, opts.wfd);
1001 	if (err < 0)
1002 		goto error;
1003 
1004 	p = (struct p9_trans_fd *) client->trans;
1005 	p->conn = p9_conn_create(client);
1006 	if (IS_ERR(p->conn)) {
1007 		err = PTR_ERR(p->conn);
1008 		p->conn = NULL;
1009 		goto error;
1010 	}
1011 
1012 	return 0;
1013 
1014 error:
1015 	kfree(p);
1016 	return err;
1017 }
1018 
1019 static struct p9_trans_module p9_tcp_trans = {
1020 	.name = "tcp",
1021 	.maxsize = MAX_SOCK_BUF,
1022 	.def = 1,
1023 	.create = p9_fd_create_tcp,
1024 	.close = p9_fd_close,
1025 	.request = p9_fd_request,
1026 	.cancel = p9_fd_cancel,
1027 	.owner = THIS_MODULE,
1028 };
1029 
1030 static struct p9_trans_module p9_unix_trans = {
1031 	.name = "unix",
1032 	.maxsize = MAX_SOCK_BUF,
1033 	.def = 0,
1034 	.create = p9_fd_create_unix,
1035 	.close = p9_fd_close,
1036 	.request = p9_fd_request,
1037 	.cancel = p9_fd_cancel,
1038 	.owner = THIS_MODULE,
1039 };
1040 
1041 static struct p9_trans_module p9_fd_trans = {
1042 	.name = "fd",
1043 	.maxsize = MAX_SOCK_BUF,
1044 	.def = 0,
1045 	.create = p9_fd_create,
1046 	.close = p9_fd_close,
1047 	.request = p9_fd_request,
1048 	.cancel = p9_fd_cancel,
1049 	.owner = THIS_MODULE,
1050 };
1051 
1052 /**
1053  * p9_poll_proc - poll worker thread
1054  * @a: thread state and arguments
1055  *
1056  * polls all v9fs transports for new events and queues the appropriate
1057  * work to the work queue
1058  *
1059  */
1060 
1061 static int p9_poll_proc(void *a)
1062 {
1063 	unsigned long flags;
1064 
1065 	P9_DPRINTK(P9_DEBUG_TRANS, "start %p\n", current);
1066  repeat:
1067 	spin_lock_irqsave(&p9_poll_lock, flags);
1068 	while (!list_empty(&p9_poll_pending_list)) {
1069 		struct p9_conn *conn = list_first_entry(&p9_poll_pending_list,
1070 							struct p9_conn,
1071 							poll_pending_link);
1072 		list_del_init(&conn->poll_pending_link);
1073 		spin_unlock_irqrestore(&p9_poll_lock, flags);
1074 
1075 		p9_poll_mux(conn);
1076 
1077 		spin_lock_irqsave(&p9_poll_lock, flags);
1078 	}
1079 	spin_unlock_irqrestore(&p9_poll_lock, flags);
1080 
1081 	set_current_state(TASK_INTERRUPTIBLE);
1082 	if (list_empty(&p9_poll_pending_list)) {
1083 		P9_DPRINTK(P9_DEBUG_TRANS, "sleeping...\n");
1084 		schedule();
1085 	}
1086 	__set_current_state(TASK_RUNNING);
1087 
1088 	if (!kthread_should_stop())
1089 		goto repeat;
1090 
1091 	P9_DPRINTK(P9_DEBUG_TRANS, "finish\n");
1092 	return 0;
1093 }
1094 
1095 int p9_trans_fd_init(void)
1096 {
1097 	p9_mux_wq = create_workqueue("v9fs");
1098 	if (!p9_mux_wq) {
1099 		printk(KERN_WARNING "v9fs: mux: creating workqueue failed\n");
1100 		return -ENOMEM;
1101 	}
1102 
1103 	p9_poll_task = kthread_run(p9_poll_proc, NULL, "v9fs-poll");
1104 	if (IS_ERR(p9_poll_task)) {
1105 		destroy_workqueue(p9_mux_wq);
1106 		printk(KERN_WARNING "v9fs: mux: creating poll task failed\n");
1107 		return PTR_ERR(p9_poll_task);
1108 	}
1109 
1110 	v9fs_register_trans(&p9_tcp_trans);
1111 	v9fs_register_trans(&p9_unix_trans);
1112 	v9fs_register_trans(&p9_fd_trans);
1113 
1114 	return 0;
1115 }
1116 
1117 void p9_trans_fd_exit(void)
1118 {
1119 	kthread_stop(p9_poll_task);
1120 	v9fs_unregister_trans(&p9_tcp_trans);
1121 	v9fs_unregister_trans(&p9_unix_trans);
1122 	v9fs_unregister_trans(&p9_fd_trans);
1123 
1124 	destroy_workqueue(p9_mux_wq);
1125 }
1126