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 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
29
30 #include <linux/in.h>
31 #include <linux/module.h>
32 #include <linux/net.h>
33 #include <linux/ipv6.h>
34 #include <linux/kthread.h>
35 #include <linux/errno.h>
36 #include <linux/kernel.h>
37 #include <linux/un.h>
38 #include <linux/uaccess.h>
39 #include <linux/inet.h>
40 #include <linux/idr.h>
41 #include <linux/file.h>
42 #include <linux/parser.h>
43 #include <linux/slab.h>
44 #include <linux/seq_file.h>
45 #include <net/9p/9p.h>
46 #include <net/9p/client.h>
47 #include <net/9p/transport.h>
48
49 #include <linux/syscalls.h> /* killme */
50
51 #define P9_PORT 564
52 #define MAX_SOCK_BUF (64*1024)
53 #define MAXPOLLWADDR 2
54
55 static struct p9_trans_module p9_tcp_trans;
56 static struct p9_trans_module p9_fd_trans;
57
58 /**
59 * struct p9_fd_opts - per-transport options
60 * @rfd: file descriptor for reading (trans=fd)
61 * @wfd: file descriptor for writing (trans=fd)
62 * @port: port to connect to (trans=tcp)
63 *
64 */
65
66 struct p9_fd_opts {
67 int rfd;
68 int wfd;
69 u16 port;
70 bool privport;
71 };
72
73 /*
74 * Option Parsing (code inspired by NFS code)
75 * - a little lazy - parse all fd-transport options
76 */
77
78 enum {
79 /* Options that take integer arguments */
80 Opt_port, Opt_rfdno, Opt_wfdno, Opt_err,
81 /* Options that take no arguments */
82 Opt_privport,
83 };
84
85 static const match_table_t tokens = {
86 {Opt_port, "port=%u"},
87 {Opt_rfdno, "rfdno=%u"},
88 {Opt_wfdno, "wfdno=%u"},
89 {Opt_privport, "privport"},
90 {Opt_err, NULL},
91 };
92
93 enum {
94 Rworksched = 1, /* read work scheduled or running */
95 Rpending = 2, /* can read */
96 Wworksched = 4, /* write work scheduled or running */
97 Wpending = 8, /* can write */
98 };
99
100 struct p9_poll_wait {
101 struct p9_conn *conn;
102 wait_queue_entry_t wait;
103 wait_queue_head_t *wait_addr;
104 };
105
106 /**
107 * struct p9_conn - fd mux connection state information
108 * @mux_list: list link for mux to manage multiple connections (?)
109 * @client: reference to client instance for this connection
110 * @err: error state
111 * @req_list: accounting for requests which have been sent
112 * @unsent_req_list: accounting for requests that haven't been sent
113 * @req: current request being processed (if any)
114 * @tmp_buf: temporary buffer to read in header
115 * @rc: temporary fcall for reading current frame
116 * @wpos: write position for current frame
117 * @wsize: amount of data to write for current frame
118 * @wbuf: current write buffer
119 * @poll_pending_link: pending links to be polled per conn
120 * @poll_wait: array of wait_q's for various worker threads
121 * @pt: poll state
122 * @rq: current read work
123 * @wq: current write work
124 * @wsched: ????
125 *
126 */
127
128 struct p9_conn {
129 struct list_head mux_list;
130 struct p9_client *client;
131 int err;
132 struct list_head req_list;
133 struct list_head unsent_req_list;
134 struct p9_req_t *rreq;
135 struct p9_req_t *wreq;
136 char tmp_buf[7];
137 struct p9_fcall rc;
138 int wpos;
139 int wsize;
140 char *wbuf;
141 struct list_head poll_pending_link;
142 struct p9_poll_wait poll_wait[MAXPOLLWADDR];
143 poll_table pt;
144 struct work_struct rq;
145 struct work_struct wq;
146 unsigned long wsched;
147 };
148
149 /**
150 * struct p9_trans_fd - transport state
151 * @rd: reference to file to read from
152 * @wr: reference of file to write to
153 * @conn: connection state reference
154 *
155 */
156
157 struct p9_trans_fd {
158 struct file *rd;
159 struct file *wr;
160 struct p9_conn conn;
161 };
162
163 static void p9_poll_workfn(struct work_struct *work);
164
165 static DEFINE_SPINLOCK(p9_poll_lock);
166 static LIST_HEAD(p9_poll_pending_list);
167 static DECLARE_WORK(p9_poll_work, p9_poll_workfn);
168
169 static unsigned int p9_ipport_resv_min = P9_DEF_MIN_RESVPORT;
170 static unsigned int p9_ipport_resv_max = P9_DEF_MAX_RESVPORT;
171
p9_mux_poll_stop(struct p9_conn * m)172 static void p9_mux_poll_stop(struct p9_conn *m)
173 {
174 unsigned long flags;
175 int i;
176
177 for (i = 0; i < ARRAY_SIZE(m->poll_wait); i++) {
178 struct p9_poll_wait *pwait = &m->poll_wait[i];
179
180 if (pwait->wait_addr) {
181 remove_wait_queue(pwait->wait_addr, &pwait->wait);
182 pwait->wait_addr = NULL;
183 }
184 }
185
186 spin_lock_irqsave(&p9_poll_lock, flags);
187 list_del_init(&m->poll_pending_link);
188 spin_unlock_irqrestore(&p9_poll_lock, flags);
189
190 flush_work(&p9_poll_work);
191 }
192
193 /**
194 * p9_conn_cancel - cancel all pending requests with error
195 * @m: mux data
196 * @err: error code
197 *
198 */
199
p9_conn_cancel(struct p9_conn * m,int err)200 static void p9_conn_cancel(struct p9_conn *m, int err)
201 {
202 struct p9_req_t *req, *rtmp;
203 LIST_HEAD(cancel_list);
204
205 p9_debug(P9_DEBUG_ERROR, "mux %p err %d\n", m, err);
206
207 spin_lock(&m->client->lock);
208
209 if (m->err) {
210 spin_unlock(&m->client->lock);
211 return;
212 }
213
214 m->err = err;
215
216 list_for_each_entry_safe(req, rtmp, &m->req_list, req_list) {
217 list_move(&req->req_list, &cancel_list);
218 }
219 list_for_each_entry_safe(req, rtmp, &m->unsent_req_list, req_list) {
220 list_move(&req->req_list, &cancel_list);
221 }
222
223 list_for_each_entry_safe(req, rtmp, &cancel_list, req_list) {
224 p9_debug(P9_DEBUG_ERROR, "call back req %p\n", req);
225 list_del(&req->req_list);
226 if (!req->t_err)
227 req->t_err = err;
228 p9_client_cb(m->client, req, REQ_STATUS_ERROR);
229 }
230 spin_unlock(&m->client->lock);
231 }
232
233 static __poll_t
p9_fd_poll(struct p9_client * client,struct poll_table_struct * pt,int * err)234 p9_fd_poll(struct p9_client *client, struct poll_table_struct *pt, int *err)
235 {
236 __poll_t ret;
237 struct p9_trans_fd *ts = NULL;
238
239 if (client && client->status == Connected)
240 ts = client->trans;
241
242 if (!ts) {
243 if (err)
244 *err = -EREMOTEIO;
245 return EPOLLERR;
246 }
247
248 ret = vfs_poll(ts->rd, pt);
249 if (ts->rd != ts->wr)
250 ret = (ret & ~EPOLLOUT) | (vfs_poll(ts->wr, pt) & ~EPOLLIN);
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
p9_fd_read(struct p9_client * client,void * v,int len)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 loff_t pos;
267
268 if (client && client->status != Disconnected)
269 ts = client->trans;
270
271 if (!ts)
272 return -EREMOTEIO;
273
274 if (!(ts->rd->f_flags & O_NONBLOCK))
275 p9_debug(P9_DEBUG_ERROR, "blocking read ...\n");
276
277 pos = ts->rd->f_pos;
278 ret = kernel_read(ts->rd, v, len, &pos);
279 if (ret <= 0 && ret != -ERESTARTSYS && ret != -EAGAIN)
280 client->status = Disconnected;
281 return ret;
282 }
283
284 /**
285 * p9_read_work - called when there is some data to be read from a transport
286 * @work: container of work to be done
287 *
288 */
289
p9_read_work(struct work_struct * work)290 static void p9_read_work(struct work_struct *work)
291 {
292 __poll_t n;
293 int err;
294 struct p9_conn *m;
295
296 m = container_of(work, struct p9_conn, rq);
297
298 if (m->err < 0)
299 return;
300
301 p9_debug(P9_DEBUG_TRANS, "start mux %p pos %zd\n", m, m->rc.offset);
302
303 if (!m->rc.sdata) {
304 m->rc.sdata = m->tmp_buf;
305 m->rc.offset = 0;
306 m->rc.capacity = 7; /* start by reading header */
307 }
308
309 clear_bit(Rpending, &m->wsched);
310 p9_debug(P9_DEBUG_TRANS, "read mux %p pos %zd size: %zd = %zd\n",
311 m, m->rc.offset, m->rc.capacity,
312 m->rc.capacity - m->rc.offset);
313 err = p9_fd_read(m->client, m->rc.sdata + m->rc.offset,
314 m->rc.capacity - m->rc.offset);
315 p9_debug(P9_DEBUG_TRANS, "mux %p got %d bytes\n", m, err);
316 if (err == -EAGAIN)
317 goto end_clear;
318
319 if (err <= 0)
320 goto error;
321
322 m->rc.offset += err;
323
324 /* header read in */
325 if ((!m->rreq) && (m->rc.offset == m->rc.capacity)) {
326 p9_debug(P9_DEBUG_TRANS, "got new header\n");
327
328 /* Header size */
329 m->rc.size = 7;
330 err = p9_parse_header(&m->rc, &m->rc.size, NULL, NULL, 0);
331 if (err) {
332 p9_debug(P9_DEBUG_ERROR,
333 "error parsing header: %d\n", err);
334 goto error;
335 }
336
337 if (m->rc.size >= m->client->msize) {
338 p9_debug(P9_DEBUG_ERROR,
339 "requested packet size too big: %d\n",
340 m->rc.size);
341 err = -EIO;
342 goto error;
343 }
344
345 p9_debug(P9_DEBUG_TRANS,
346 "mux %p pkt: size: %d bytes tag: %d\n",
347 m, m->rc.size, m->rc.tag);
348
349 m->rreq = p9_tag_lookup(m->client, m->rc.tag);
350 if (!m->rreq || (m->rreq->status != REQ_STATUS_SENT)) {
351 p9_debug(P9_DEBUG_ERROR, "Unexpected packet tag %d\n",
352 m->rc.tag);
353 err = -EIO;
354 goto error;
355 }
356
357 if (!m->rreq->rc.sdata) {
358 p9_debug(P9_DEBUG_ERROR,
359 "No recv fcall for tag %d (req %p), disconnecting!\n",
360 m->rc.tag, m->rreq);
361 m->rreq = NULL;
362 err = -EIO;
363 goto error;
364 }
365 m->rc.sdata = m->rreq->rc.sdata;
366 memcpy(m->rc.sdata, m->tmp_buf, m->rc.capacity);
367 m->rc.capacity = m->rc.size;
368 }
369
370 /* packet is read in
371 * not an else because some packets (like clunk) have no payload
372 */
373 if ((m->rreq) && (m->rc.offset == m->rc.capacity)) {
374 p9_debug(P9_DEBUG_TRANS, "got new packet\n");
375 m->rreq->rc.size = m->rc.offset;
376 spin_lock(&m->client->lock);
377 if (m->rreq->status == REQ_STATUS_SENT) {
378 list_del(&m->rreq->req_list);
379 p9_client_cb(m->client, m->rreq, REQ_STATUS_RCVD);
380 } else if (m->rreq->status == REQ_STATUS_FLSHD) {
381 /* Ignore replies associated with a cancelled request. */
382 p9_debug(P9_DEBUG_TRANS,
383 "Ignore replies associated with a cancelled request\n");
384 } else {
385 spin_unlock(&m->client->lock);
386 p9_debug(P9_DEBUG_ERROR,
387 "Request tag %d errored out while we were reading the reply\n",
388 m->rc.tag);
389 err = -EIO;
390 goto error;
391 }
392 spin_unlock(&m->client->lock);
393 m->rc.sdata = NULL;
394 m->rc.offset = 0;
395 m->rc.capacity = 0;
396 p9_req_put(m->rreq);
397 m->rreq = NULL;
398 }
399
400 end_clear:
401 clear_bit(Rworksched, &m->wsched);
402
403 if (!list_empty(&m->req_list)) {
404 if (test_and_clear_bit(Rpending, &m->wsched))
405 n = EPOLLIN;
406 else
407 n = p9_fd_poll(m->client, NULL, NULL);
408
409 if ((n & EPOLLIN) && !test_and_set_bit(Rworksched, &m->wsched)) {
410 p9_debug(P9_DEBUG_TRANS, "sched read work %p\n", m);
411 schedule_work(&m->rq);
412 }
413 }
414
415 return;
416 error:
417 p9_conn_cancel(m, err);
418 clear_bit(Rworksched, &m->wsched);
419 }
420
421 /**
422 * p9_fd_write - write to a socket
423 * @client: client instance
424 * @v: buffer to send data from
425 * @len: size of send buffer
426 *
427 */
428
p9_fd_write(struct p9_client * client,void * v,int len)429 static int p9_fd_write(struct p9_client *client, void *v, int len)
430 {
431 ssize_t ret;
432 struct p9_trans_fd *ts = NULL;
433
434 if (client && client->status != Disconnected)
435 ts = client->trans;
436
437 if (!ts)
438 return -EREMOTEIO;
439
440 if (!(ts->wr->f_flags & O_NONBLOCK))
441 p9_debug(P9_DEBUG_ERROR, "blocking write ...\n");
442
443 ret = kernel_write(ts->wr, v, len, &ts->wr->f_pos);
444 if (ret <= 0 && ret != -ERESTARTSYS && ret != -EAGAIN)
445 client->status = Disconnected;
446 return ret;
447 }
448
449 /**
450 * p9_write_work - called when a transport can send some data
451 * @work: container for work to be done
452 *
453 */
454
p9_write_work(struct work_struct * work)455 static void p9_write_work(struct work_struct *work)
456 {
457 __poll_t n;
458 int err;
459 struct p9_conn *m;
460 struct p9_req_t *req;
461
462 m = container_of(work, struct p9_conn, wq);
463
464 if (m->err < 0) {
465 clear_bit(Wworksched, &m->wsched);
466 return;
467 }
468
469 if (!m->wsize) {
470 spin_lock(&m->client->lock);
471 if (list_empty(&m->unsent_req_list)) {
472 clear_bit(Wworksched, &m->wsched);
473 spin_unlock(&m->client->lock);
474 return;
475 }
476
477 req = list_entry(m->unsent_req_list.next, struct p9_req_t,
478 req_list);
479 req->status = REQ_STATUS_SENT;
480 p9_debug(P9_DEBUG_TRANS, "move req %p\n", req);
481 list_move_tail(&req->req_list, &m->req_list);
482
483 m->wbuf = req->tc.sdata;
484 m->wsize = req->tc.size;
485 m->wpos = 0;
486 p9_req_get(req);
487 m->wreq = req;
488 spin_unlock(&m->client->lock);
489 }
490
491 p9_debug(P9_DEBUG_TRANS, "mux %p pos %d size %d\n",
492 m, m->wpos, m->wsize);
493 clear_bit(Wpending, &m->wsched);
494 err = p9_fd_write(m->client, m->wbuf + m->wpos, m->wsize - m->wpos);
495 p9_debug(P9_DEBUG_TRANS, "mux %p sent %d bytes\n", m, err);
496 if (err == -EAGAIN)
497 goto end_clear;
498
499
500 if (err < 0)
501 goto error;
502 else if (err == 0) {
503 err = -EREMOTEIO;
504 goto error;
505 }
506
507 m->wpos += err;
508 if (m->wpos == m->wsize) {
509 m->wpos = m->wsize = 0;
510 p9_req_put(m->wreq);
511 m->wreq = NULL;
512 }
513
514 end_clear:
515 clear_bit(Wworksched, &m->wsched);
516
517 if (m->wsize || !list_empty(&m->unsent_req_list)) {
518 if (test_and_clear_bit(Wpending, &m->wsched))
519 n = EPOLLOUT;
520 else
521 n = p9_fd_poll(m->client, NULL, NULL);
522
523 if ((n & EPOLLOUT) &&
524 !test_and_set_bit(Wworksched, &m->wsched)) {
525 p9_debug(P9_DEBUG_TRANS, "sched write work %p\n", m);
526 schedule_work(&m->wq);
527 }
528 }
529
530 return;
531
532 error:
533 p9_conn_cancel(m, err);
534 clear_bit(Wworksched, &m->wsched);
535 }
536
p9_pollwake(wait_queue_entry_t * wait,unsigned int mode,int sync,void * key)537 static int p9_pollwake(wait_queue_entry_t *wait, unsigned int mode, int sync, void *key)
538 {
539 struct p9_poll_wait *pwait =
540 container_of(wait, struct p9_poll_wait, wait);
541 struct p9_conn *m = pwait->conn;
542 unsigned long flags;
543
544 spin_lock_irqsave(&p9_poll_lock, flags);
545 if (list_empty(&m->poll_pending_link))
546 list_add_tail(&m->poll_pending_link, &p9_poll_pending_list);
547 spin_unlock_irqrestore(&p9_poll_lock, flags);
548
549 schedule_work(&p9_poll_work);
550 return 1;
551 }
552
553 /**
554 * p9_pollwait - add poll task to the wait queue
555 * @filp: file pointer being polled
556 * @wait_address: wait_q to block on
557 * @p: poll state
558 *
559 * called by files poll operation to add v9fs-poll task to files wait queue
560 */
561
562 static void
p9_pollwait(struct file * filp,wait_queue_head_t * wait_address,poll_table * p)563 p9_pollwait(struct file *filp, wait_queue_head_t *wait_address, poll_table *p)
564 {
565 struct p9_conn *m = container_of(p, struct p9_conn, pt);
566 struct p9_poll_wait *pwait = NULL;
567 int i;
568
569 for (i = 0; i < ARRAY_SIZE(m->poll_wait); i++) {
570 if (m->poll_wait[i].wait_addr == NULL) {
571 pwait = &m->poll_wait[i];
572 break;
573 }
574 }
575
576 if (!pwait) {
577 p9_debug(P9_DEBUG_ERROR, "not enough wait_address slots\n");
578 return;
579 }
580
581 pwait->conn = m;
582 pwait->wait_addr = wait_address;
583 init_waitqueue_func_entry(&pwait->wait, p9_pollwake);
584 add_wait_queue(wait_address, &pwait->wait);
585 }
586
587 /**
588 * p9_conn_create - initialize the per-session mux data
589 * @client: client instance
590 *
591 * Note: Creates the polling task if this is the first session.
592 */
593
p9_conn_create(struct p9_client * client)594 static void p9_conn_create(struct p9_client *client)
595 {
596 __poll_t n;
597 struct p9_trans_fd *ts = client->trans;
598 struct p9_conn *m = &ts->conn;
599
600 p9_debug(P9_DEBUG_TRANS, "client %p msize %d\n", client, client->msize);
601
602 INIT_LIST_HEAD(&m->mux_list);
603 m->client = client;
604
605 INIT_LIST_HEAD(&m->req_list);
606 INIT_LIST_HEAD(&m->unsent_req_list);
607 INIT_WORK(&m->rq, p9_read_work);
608 INIT_WORK(&m->wq, p9_write_work);
609 INIT_LIST_HEAD(&m->poll_pending_link);
610 init_poll_funcptr(&m->pt, p9_pollwait);
611
612 n = p9_fd_poll(client, &m->pt, NULL);
613 if (n & EPOLLIN) {
614 p9_debug(P9_DEBUG_TRANS, "mux %p can read\n", m);
615 set_bit(Rpending, &m->wsched);
616 }
617
618 if (n & EPOLLOUT) {
619 p9_debug(P9_DEBUG_TRANS, "mux %p can write\n", m);
620 set_bit(Wpending, &m->wsched);
621 }
622 }
623
624 /**
625 * p9_poll_mux - polls a mux and schedules read or write works if necessary
626 * @m: connection to poll
627 *
628 */
629
p9_poll_mux(struct p9_conn * m)630 static void p9_poll_mux(struct p9_conn *m)
631 {
632 __poll_t n;
633 int err = -ECONNRESET;
634
635 if (m->err < 0)
636 return;
637
638 n = p9_fd_poll(m->client, NULL, &err);
639 if (n & (EPOLLERR | EPOLLHUP | EPOLLNVAL)) {
640 p9_debug(P9_DEBUG_TRANS, "error mux %p err %d\n", m, n);
641 p9_conn_cancel(m, err);
642 }
643
644 if (n & EPOLLIN) {
645 set_bit(Rpending, &m->wsched);
646 p9_debug(P9_DEBUG_TRANS, "mux %p can read\n", m);
647 if (!test_and_set_bit(Rworksched, &m->wsched)) {
648 p9_debug(P9_DEBUG_TRANS, "sched read work %p\n", m);
649 schedule_work(&m->rq);
650 }
651 }
652
653 if (n & EPOLLOUT) {
654 set_bit(Wpending, &m->wsched);
655 p9_debug(P9_DEBUG_TRANS, "mux %p can write\n", m);
656 if ((m->wsize || !list_empty(&m->unsent_req_list)) &&
657 !test_and_set_bit(Wworksched, &m->wsched)) {
658 p9_debug(P9_DEBUG_TRANS, "sched write work %p\n", m);
659 schedule_work(&m->wq);
660 }
661 }
662 }
663
664 /**
665 * p9_fd_request - send 9P request
666 * The function can sleep until the request is scheduled for sending.
667 * The function can be interrupted. Return from the function is not
668 * a guarantee that the request is sent successfully.
669 *
670 * @client: client instance
671 * @req: request to be sent
672 *
673 */
674
p9_fd_request(struct p9_client * client,struct p9_req_t * req)675 static int p9_fd_request(struct p9_client *client, struct p9_req_t *req)
676 {
677 __poll_t n;
678 struct p9_trans_fd *ts = client->trans;
679 struct p9_conn *m = &ts->conn;
680
681 p9_debug(P9_DEBUG_TRANS, "mux %p task %p tcall %p id %d\n",
682 m, current, &req->tc, req->tc.id);
683 if (m->err < 0)
684 return m->err;
685
686 spin_lock(&client->lock);
687 req->status = REQ_STATUS_UNSENT;
688 list_add_tail(&req->req_list, &m->unsent_req_list);
689 spin_unlock(&client->lock);
690
691 if (test_and_clear_bit(Wpending, &m->wsched))
692 n = EPOLLOUT;
693 else
694 n = p9_fd_poll(m->client, NULL, NULL);
695
696 if (n & EPOLLOUT && !test_and_set_bit(Wworksched, &m->wsched))
697 schedule_work(&m->wq);
698
699 return 0;
700 }
701
p9_fd_cancel(struct p9_client * client,struct p9_req_t * req)702 static int p9_fd_cancel(struct p9_client *client, struct p9_req_t *req)
703 {
704 int ret = 1;
705
706 p9_debug(P9_DEBUG_TRANS, "client %p req %p\n", client, req);
707
708 spin_lock(&client->lock);
709
710 if (req->status == REQ_STATUS_UNSENT) {
711 list_del(&req->req_list);
712 req->status = REQ_STATUS_FLSHD;
713 p9_req_put(req);
714 ret = 0;
715 }
716 spin_unlock(&client->lock);
717
718 return ret;
719 }
720
p9_fd_cancelled(struct p9_client * client,struct p9_req_t * req)721 static int p9_fd_cancelled(struct p9_client *client, struct p9_req_t *req)
722 {
723 p9_debug(P9_DEBUG_TRANS, "client %p req %p\n", client, req);
724
725 spin_lock(&client->lock);
726 /* Ignore cancelled request if message has been received
727 * before lock.
728 */
729 if (req->status == REQ_STATUS_RCVD) {
730 spin_unlock(&client->lock);
731 return 0;
732 }
733
734 /* we haven't received a response for oldreq,
735 * remove it from the list.
736 */
737 list_del(&req->req_list);
738 req->status = REQ_STATUS_FLSHD;
739 spin_unlock(&client->lock);
740 p9_req_put(req);
741
742 return 0;
743 }
744
p9_fd_show_options(struct seq_file * m,struct p9_client * clnt)745 static int p9_fd_show_options(struct seq_file *m, struct p9_client *clnt)
746 {
747 if (clnt->trans_mod == &p9_tcp_trans) {
748 if (clnt->trans_opts.tcp.port != P9_PORT)
749 seq_printf(m, ",port=%u", clnt->trans_opts.tcp.port);
750 } else if (clnt->trans_mod == &p9_fd_trans) {
751 if (clnt->trans_opts.fd.rfd != ~0)
752 seq_printf(m, ",rfd=%u", clnt->trans_opts.fd.rfd);
753 if (clnt->trans_opts.fd.wfd != ~0)
754 seq_printf(m, ",wfd=%u", clnt->trans_opts.fd.wfd);
755 }
756 return 0;
757 }
758
759 /**
760 * parse_opts - parse mount options into p9_fd_opts structure
761 * @params: options string passed from mount
762 * @opts: fd transport-specific structure to parse options into
763 *
764 * Returns 0 upon success, -ERRNO upon failure
765 */
766
parse_opts(char * params,struct p9_fd_opts * opts)767 static int parse_opts(char *params, struct p9_fd_opts *opts)
768 {
769 char *p;
770 substring_t args[MAX_OPT_ARGS];
771 int option;
772 char *options, *tmp_options;
773
774 opts->port = P9_PORT;
775 opts->rfd = ~0;
776 opts->wfd = ~0;
777 opts->privport = false;
778
779 if (!params)
780 return 0;
781
782 tmp_options = kstrdup(params, GFP_KERNEL);
783 if (!tmp_options) {
784 p9_debug(P9_DEBUG_ERROR,
785 "failed to allocate copy of option string\n");
786 return -ENOMEM;
787 }
788 options = tmp_options;
789
790 while ((p = strsep(&options, ",")) != NULL) {
791 int token;
792 int r;
793 if (!*p)
794 continue;
795 token = match_token(p, tokens, args);
796 if ((token != Opt_err) && (token != Opt_privport)) {
797 r = match_int(&args[0], &option);
798 if (r < 0) {
799 p9_debug(P9_DEBUG_ERROR,
800 "integer field, but no integer?\n");
801 continue;
802 }
803 }
804 switch (token) {
805 case Opt_port:
806 opts->port = option;
807 break;
808 case Opt_rfdno:
809 opts->rfd = option;
810 break;
811 case Opt_wfdno:
812 opts->wfd = option;
813 break;
814 case Opt_privport:
815 opts->privport = true;
816 break;
817 default:
818 continue;
819 }
820 }
821
822 kfree(tmp_options);
823 return 0;
824 }
825
p9_fd_open(struct p9_client * client,int rfd,int wfd)826 static int p9_fd_open(struct p9_client *client, int rfd, int wfd)
827 {
828 struct p9_trans_fd *ts = kzalloc(sizeof(struct p9_trans_fd),
829 GFP_KERNEL);
830 if (!ts)
831 return -ENOMEM;
832
833 ts->rd = fget(rfd);
834 if (!ts->rd)
835 goto out_free_ts;
836 if (!(ts->rd->f_mode & FMODE_READ))
837 goto out_put_rd;
838 ts->wr = fget(wfd);
839 if (!ts->wr)
840 goto out_put_rd;
841 if (!(ts->wr->f_mode & FMODE_WRITE))
842 goto out_put_wr;
843
844 client->trans = ts;
845 client->status = Connected;
846
847 return 0;
848
849 out_put_wr:
850 fput(ts->wr);
851 out_put_rd:
852 fput(ts->rd);
853 out_free_ts:
854 kfree(ts);
855 return -EIO;
856 }
857
p9_socket_open(struct p9_client * client,struct socket * csocket)858 static int p9_socket_open(struct p9_client *client, struct socket *csocket)
859 {
860 struct p9_trans_fd *p;
861 struct file *file;
862
863 p = kzalloc(sizeof(struct p9_trans_fd), GFP_KERNEL);
864 if (!p)
865 return -ENOMEM;
866
867 csocket->sk->sk_allocation = GFP_NOIO;
868 file = sock_alloc_file(csocket, 0, NULL);
869 if (IS_ERR(file)) {
870 pr_err("%s (%d): failed to map fd\n",
871 __func__, task_pid_nr(current));
872 kfree(p);
873 return PTR_ERR(file);
874 }
875
876 get_file(file);
877 p->wr = p->rd = file;
878 client->trans = p;
879 client->status = Connected;
880
881 p->rd->f_flags |= O_NONBLOCK;
882
883 p9_conn_create(client);
884 return 0;
885 }
886
887 /**
888 * p9_mux_destroy - cancels all pending requests of mux
889 * @m: mux to destroy
890 *
891 */
892
p9_conn_destroy(struct p9_conn * m)893 static void p9_conn_destroy(struct p9_conn *m)
894 {
895 p9_debug(P9_DEBUG_TRANS, "mux %p prev %p next %p\n",
896 m, m->mux_list.prev, m->mux_list.next);
897
898 p9_mux_poll_stop(m);
899 cancel_work_sync(&m->rq);
900 if (m->rreq) {
901 p9_req_put(m->rreq);
902 m->rreq = NULL;
903 }
904 cancel_work_sync(&m->wq);
905 if (m->wreq) {
906 p9_req_put(m->wreq);
907 m->wreq = NULL;
908 }
909
910 p9_conn_cancel(m, -ECONNRESET);
911
912 m->client = NULL;
913 }
914
915 /**
916 * p9_fd_close - shutdown file descriptor transport
917 * @client: client instance
918 *
919 */
920
p9_fd_close(struct p9_client * client)921 static void p9_fd_close(struct p9_client *client)
922 {
923 struct p9_trans_fd *ts;
924
925 if (!client)
926 return;
927
928 ts = client->trans;
929 if (!ts)
930 return;
931
932 client->status = Disconnected;
933
934 p9_conn_destroy(&ts->conn);
935
936 if (ts->rd)
937 fput(ts->rd);
938 if (ts->wr)
939 fput(ts->wr);
940
941 kfree(ts);
942 }
943
944 /*
945 * stolen from NFS - maybe should be made a generic function?
946 */
valid_ipaddr4(const char * buf)947 static inline int valid_ipaddr4(const char *buf)
948 {
949 int rc, count, in[4];
950
951 rc = sscanf(buf, "%d.%d.%d.%d", &in[0], &in[1], &in[2], &in[3]);
952 if (rc != 4)
953 return -EINVAL;
954 for (count = 0; count < 4; count++) {
955 if (in[count] > 255)
956 return -EINVAL;
957 }
958 return 0;
959 }
960
p9_bind_privport(struct socket * sock)961 static int p9_bind_privport(struct socket *sock)
962 {
963 struct sockaddr_in cl;
964 int port, err = -EINVAL;
965
966 memset(&cl, 0, sizeof(cl));
967 cl.sin_family = AF_INET;
968 cl.sin_addr.s_addr = INADDR_ANY;
969 for (port = p9_ipport_resv_max; port >= p9_ipport_resv_min; port--) {
970 cl.sin_port = htons((ushort)port);
971 err = kernel_bind(sock, (struct sockaddr *)&cl, sizeof(cl));
972 if (err != -EADDRINUSE)
973 break;
974 }
975 return err;
976 }
977
978
979 static int
p9_fd_create_tcp(struct p9_client * client,const char * addr,char * args)980 p9_fd_create_tcp(struct p9_client *client, const char *addr, char *args)
981 {
982 int err;
983 struct socket *csocket;
984 struct sockaddr_in sin_server;
985 struct p9_fd_opts opts;
986
987 err = parse_opts(args, &opts);
988 if (err < 0)
989 return err;
990
991 if (addr == NULL || valid_ipaddr4(addr) < 0)
992 return -EINVAL;
993
994 csocket = NULL;
995
996 client->trans_opts.tcp.port = opts.port;
997 client->trans_opts.tcp.privport = opts.privport;
998 sin_server.sin_family = AF_INET;
999 sin_server.sin_addr.s_addr = in_aton(addr);
1000 sin_server.sin_port = htons(opts.port);
1001 err = __sock_create(current->nsproxy->net_ns, PF_INET,
1002 SOCK_STREAM, IPPROTO_TCP, &csocket, 1);
1003 if (err) {
1004 pr_err("%s (%d): problem creating socket\n",
1005 __func__, task_pid_nr(current));
1006 return err;
1007 }
1008
1009 if (opts.privport) {
1010 err = p9_bind_privport(csocket);
1011 if (err < 0) {
1012 pr_err("%s (%d): problem binding to privport\n",
1013 __func__, task_pid_nr(current));
1014 sock_release(csocket);
1015 return err;
1016 }
1017 }
1018
1019 err = csocket->ops->connect(csocket,
1020 (struct sockaddr *)&sin_server,
1021 sizeof(struct sockaddr_in), 0);
1022 if (err < 0) {
1023 pr_err("%s (%d): problem connecting socket to %s\n",
1024 __func__, task_pid_nr(current), addr);
1025 sock_release(csocket);
1026 return err;
1027 }
1028
1029 return p9_socket_open(client, csocket);
1030 }
1031
1032 static int
p9_fd_create_unix(struct p9_client * client,const char * addr,char * args)1033 p9_fd_create_unix(struct p9_client *client, const char *addr, char *args)
1034 {
1035 int err;
1036 struct socket *csocket;
1037 struct sockaddr_un sun_server;
1038
1039 csocket = NULL;
1040
1041 if (!addr || !strlen(addr))
1042 return -EINVAL;
1043
1044 if (strlen(addr) >= UNIX_PATH_MAX) {
1045 pr_err("%s (%d): address too long: %s\n",
1046 __func__, task_pid_nr(current), addr);
1047 return -ENAMETOOLONG;
1048 }
1049
1050 sun_server.sun_family = PF_UNIX;
1051 strcpy(sun_server.sun_path, addr);
1052 err = __sock_create(current->nsproxy->net_ns, PF_UNIX,
1053 SOCK_STREAM, 0, &csocket, 1);
1054 if (err < 0) {
1055 pr_err("%s (%d): problem creating socket\n",
1056 __func__, task_pid_nr(current));
1057
1058 return err;
1059 }
1060 err = csocket->ops->connect(csocket, (struct sockaddr *)&sun_server,
1061 sizeof(struct sockaddr_un) - 1, 0);
1062 if (err < 0) {
1063 pr_err("%s (%d): problem connecting socket: %s: %d\n",
1064 __func__, task_pid_nr(current), addr, err);
1065 sock_release(csocket);
1066 return err;
1067 }
1068
1069 return p9_socket_open(client, csocket);
1070 }
1071
1072 static int
p9_fd_create(struct p9_client * client,const char * addr,char * args)1073 p9_fd_create(struct p9_client *client, const char *addr, char *args)
1074 {
1075 int err;
1076 struct p9_fd_opts opts;
1077
1078 parse_opts(args, &opts);
1079 client->trans_opts.fd.rfd = opts.rfd;
1080 client->trans_opts.fd.wfd = opts.wfd;
1081
1082 if (opts.rfd == ~0 || opts.wfd == ~0) {
1083 pr_err("Insufficient options for proto=fd\n");
1084 return -ENOPROTOOPT;
1085 }
1086
1087 err = p9_fd_open(client, opts.rfd, opts.wfd);
1088 if (err < 0)
1089 return err;
1090
1091 p9_conn_create(client);
1092
1093 return 0;
1094 }
1095
1096 static struct p9_trans_module p9_tcp_trans = {
1097 .name = "tcp",
1098 .maxsize = MAX_SOCK_BUF,
1099 .def = 0,
1100 .create = p9_fd_create_tcp,
1101 .close = p9_fd_close,
1102 .request = p9_fd_request,
1103 .cancel = p9_fd_cancel,
1104 .cancelled = p9_fd_cancelled,
1105 .show_options = p9_fd_show_options,
1106 .owner = THIS_MODULE,
1107 };
1108
1109 static struct p9_trans_module p9_unix_trans = {
1110 .name = "unix",
1111 .maxsize = MAX_SOCK_BUF,
1112 .def = 0,
1113 .create = p9_fd_create_unix,
1114 .close = p9_fd_close,
1115 .request = p9_fd_request,
1116 .cancel = p9_fd_cancel,
1117 .cancelled = p9_fd_cancelled,
1118 .show_options = p9_fd_show_options,
1119 .owner = THIS_MODULE,
1120 };
1121
1122 static struct p9_trans_module p9_fd_trans = {
1123 .name = "fd",
1124 .maxsize = MAX_SOCK_BUF,
1125 .def = 0,
1126 .create = p9_fd_create,
1127 .close = p9_fd_close,
1128 .request = p9_fd_request,
1129 .cancel = p9_fd_cancel,
1130 .cancelled = p9_fd_cancelled,
1131 .show_options = p9_fd_show_options,
1132 .owner = THIS_MODULE,
1133 };
1134
1135 /**
1136 * p9_poll_workfn - poll worker thread
1137 * @work: work queue
1138 *
1139 * polls all v9fs transports for new events and queues the appropriate
1140 * work to the work queue
1141 *
1142 */
1143
p9_poll_workfn(struct work_struct * work)1144 static void p9_poll_workfn(struct work_struct *work)
1145 {
1146 unsigned long flags;
1147
1148 p9_debug(P9_DEBUG_TRANS, "start %p\n", current);
1149
1150 spin_lock_irqsave(&p9_poll_lock, flags);
1151 while (!list_empty(&p9_poll_pending_list)) {
1152 struct p9_conn *conn = list_first_entry(&p9_poll_pending_list,
1153 struct p9_conn,
1154 poll_pending_link);
1155 list_del_init(&conn->poll_pending_link);
1156 spin_unlock_irqrestore(&p9_poll_lock, flags);
1157
1158 p9_poll_mux(conn);
1159
1160 spin_lock_irqsave(&p9_poll_lock, flags);
1161 }
1162 spin_unlock_irqrestore(&p9_poll_lock, flags);
1163
1164 p9_debug(P9_DEBUG_TRANS, "finish\n");
1165 }
1166
p9_trans_fd_init(void)1167 int p9_trans_fd_init(void)
1168 {
1169 v9fs_register_trans(&p9_tcp_trans);
1170 v9fs_register_trans(&p9_unix_trans);
1171 v9fs_register_trans(&p9_fd_trans);
1172
1173 return 0;
1174 }
1175
p9_trans_fd_exit(void)1176 void p9_trans_fd_exit(void)
1177 {
1178 flush_work(&p9_poll_work);
1179 v9fs_unregister_trans(&p9_tcp_trans);
1180 v9fs_unregister_trans(&p9_unix_trans);
1181 v9fs_unregister_trans(&p9_fd_trans);
1182 }
1183