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 *req; 135 char tmp_buf[7]; 136 struct p9_fcall rc; 137 int wpos; 138 int wsize; 139 char *wbuf; 140 struct list_head poll_pending_link; 141 struct p9_poll_wait poll_wait[MAXPOLLWADDR]; 142 poll_table pt; 143 struct work_struct rq; 144 struct work_struct wq; 145 unsigned long wsched; 146 }; 147 148 /** 149 * struct p9_trans_fd - transport state 150 * @rd: reference to file to read from 151 * @wr: reference of file to write to 152 * @conn: connection state reference 153 * 154 */ 155 156 struct p9_trans_fd { 157 struct file *rd; 158 struct file *wr; 159 struct p9_conn conn; 160 }; 161 162 static void p9_poll_workfn(struct work_struct *work); 163 164 static DEFINE_SPINLOCK(p9_poll_lock); 165 static LIST_HEAD(p9_poll_pending_list); 166 static DECLARE_WORK(p9_poll_work, p9_poll_workfn); 167 168 static unsigned int p9_ipport_resv_min = P9_DEF_MIN_RESVPORT; 169 static unsigned int p9_ipport_resv_max = P9_DEF_MAX_RESVPORT; 170 171 static void p9_mux_poll_stop(struct p9_conn *m) 172 { 173 unsigned long flags; 174 int i; 175 176 for (i = 0; i < ARRAY_SIZE(m->poll_wait); i++) { 177 struct p9_poll_wait *pwait = &m->poll_wait[i]; 178 179 if (pwait->wait_addr) { 180 remove_wait_queue(pwait->wait_addr, &pwait->wait); 181 pwait->wait_addr = NULL; 182 } 183 } 184 185 spin_lock_irqsave(&p9_poll_lock, flags); 186 list_del_init(&m->poll_pending_link); 187 spin_unlock_irqrestore(&p9_poll_lock, flags); 188 189 flush_work(&p9_poll_work); 190 } 191 192 /** 193 * p9_conn_cancel - cancel all pending requests with error 194 * @m: mux data 195 * @err: error code 196 * 197 */ 198 199 static void p9_conn_cancel(struct p9_conn *m, int err) 200 { 201 struct p9_req_t *req, *rtmp; 202 LIST_HEAD(cancel_list); 203 204 p9_debug(P9_DEBUG_ERROR, "mux %p err %d\n", m, err); 205 206 spin_lock(&m->client->lock); 207 208 if (m->err) { 209 spin_unlock(&m->client->lock); 210 return; 211 } 212 213 m->err = err; 214 215 list_for_each_entry_safe(req, rtmp, &m->req_list, req_list) { 216 list_move(&req->req_list, &cancel_list); 217 } 218 list_for_each_entry_safe(req, rtmp, &m->unsent_req_list, req_list) { 219 list_move(&req->req_list, &cancel_list); 220 } 221 222 list_for_each_entry_safe(req, rtmp, &cancel_list, req_list) { 223 p9_debug(P9_DEBUG_ERROR, "call back req %p\n", req); 224 list_del(&req->req_list); 225 if (!req->t_err) 226 req->t_err = err; 227 p9_client_cb(m->client, req, REQ_STATUS_ERROR); 228 } 229 spin_unlock(&m->client->lock); 230 } 231 232 static __poll_t 233 p9_fd_poll(struct p9_client *client, struct poll_table_struct *pt, int *err) 234 { 235 __poll_t ret; 236 struct p9_trans_fd *ts = NULL; 237 238 if (client && client->status == Connected) 239 ts = client->trans; 240 241 if (!ts) { 242 if (err) 243 *err = -EREMOTEIO; 244 return EPOLLERR; 245 } 246 247 ret = vfs_poll(ts->rd, pt); 248 if (ts->rd != ts->wr) 249 ret = (ret & ~EPOLLOUT) | (vfs_poll(ts->wr, pt) & ~EPOLLIN); 250 return ret; 251 } 252 253 /** 254 * p9_fd_read- read from a fd 255 * @client: client instance 256 * @v: buffer to receive data into 257 * @len: size of receive buffer 258 * 259 */ 260 261 static int p9_fd_read(struct p9_client *client, void *v, int len) 262 { 263 int ret; 264 struct p9_trans_fd *ts = NULL; 265 loff_t pos; 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_debug(P9_DEBUG_ERROR, "blocking read ...\n"); 275 276 pos = ts->rd->f_pos; 277 ret = kernel_read(ts->rd, v, len, &pos); 278 if (ret <= 0 && ret != -ERESTARTSYS && ret != -EAGAIN) 279 client->status = Disconnected; 280 return ret; 281 } 282 283 /** 284 * p9_read_work - called when there is some data to be read from a transport 285 * @work: container of work to be done 286 * 287 */ 288 289 static void p9_read_work(struct work_struct *work) 290 { 291 __poll_t n; 292 int err; 293 struct p9_conn *m; 294 int status = REQ_STATUS_ERROR; 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->req) && (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->req = p9_tag_lookup(m->client, m->rc.tag); 350 if (!m->req || (m->req->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->req->rc == NULL) { 358 p9_debug(P9_DEBUG_ERROR, 359 "No recv fcall for tag %d (req %p), disconnecting!\n", 360 m->rc.tag, m->req); 361 m->req = NULL; 362 err = -EIO; 363 goto error; 364 } 365 m->rc.sdata = (char *)m->req->rc + sizeof(struct p9_fcall); 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->req) && (m->rc.offset == m->rc.capacity)) { 374 p9_debug(P9_DEBUG_TRANS, "got new packet\n"); 375 m->req->rc->size = m->rc.offset; 376 spin_lock(&m->client->lock); 377 if (m->req->status != REQ_STATUS_ERROR) 378 status = REQ_STATUS_RCVD; 379 list_del(&m->req->req_list); 380 /* update req->status while holding client->lock */ 381 p9_client_cb(m->client, m->req, status); 382 spin_unlock(&m->client->lock); 383 m->rc.sdata = NULL; 384 m->rc.offset = 0; 385 m->rc.capacity = 0; 386 m->req = NULL; 387 } 388 389 end_clear: 390 clear_bit(Rworksched, &m->wsched); 391 392 if (!list_empty(&m->req_list)) { 393 if (test_and_clear_bit(Rpending, &m->wsched)) 394 n = EPOLLIN; 395 else 396 n = p9_fd_poll(m->client, NULL, NULL); 397 398 if ((n & EPOLLIN) && !test_and_set_bit(Rworksched, &m->wsched)) { 399 p9_debug(P9_DEBUG_TRANS, "sched read work %p\n", m); 400 schedule_work(&m->rq); 401 } 402 } 403 404 return; 405 error: 406 p9_conn_cancel(m, err); 407 clear_bit(Rworksched, &m->wsched); 408 } 409 410 /** 411 * p9_fd_write - write to a socket 412 * @client: client instance 413 * @v: buffer to send data from 414 * @len: size of send buffer 415 * 416 */ 417 418 static int p9_fd_write(struct p9_client *client, void *v, int len) 419 { 420 ssize_t ret; 421 struct p9_trans_fd *ts = NULL; 422 423 if (client && client->status != Disconnected) 424 ts = client->trans; 425 426 if (!ts) 427 return -EREMOTEIO; 428 429 if (!(ts->wr->f_flags & O_NONBLOCK)) 430 p9_debug(P9_DEBUG_ERROR, "blocking write ...\n"); 431 432 ret = kernel_write(ts->wr, v, len, &ts->wr->f_pos); 433 if (ret <= 0 && ret != -ERESTARTSYS && ret != -EAGAIN) 434 client->status = Disconnected; 435 return ret; 436 } 437 438 /** 439 * p9_write_work - called when a transport can send some data 440 * @work: container for work to be done 441 * 442 */ 443 444 static void p9_write_work(struct work_struct *work) 445 { 446 __poll_t n; 447 int err; 448 struct p9_conn *m; 449 struct p9_req_t *req; 450 451 m = container_of(work, struct p9_conn, wq); 452 453 if (m->err < 0) { 454 clear_bit(Wworksched, &m->wsched); 455 return; 456 } 457 458 if (!m->wsize) { 459 spin_lock(&m->client->lock); 460 if (list_empty(&m->unsent_req_list)) { 461 clear_bit(Wworksched, &m->wsched); 462 spin_unlock(&m->client->lock); 463 return; 464 } 465 466 req = list_entry(m->unsent_req_list.next, struct p9_req_t, 467 req_list); 468 req->status = REQ_STATUS_SENT; 469 p9_debug(P9_DEBUG_TRANS, "move req %p\n", req); 470 list_move_tail(&req->req_list, &m->req_list); 471 472 m->wbuf = req->tc->sdata; 473 m->wsize = req->tc->size; 474 m->wpos = 0; 475 spin_unlock(&m->client->lock); 476 } 477 478 p9_debug(P9_DEBUG_TRANS, "mux %p pos %d size %d\n", 479 m, m->wpos, m->wsize); 480 clear_bit(Wpending, &m->wsched); 481 err = p9_fd_write(m->client, m->wbuf + m->wpos, m->wsize - m->wpos); 482 p9_debug(P9_DEBUG_TRANS, "mux %p sent %d bytes\n", m, err); 483 if (err == -EAGAIN) 484 goto end_clear; 485 486 487 if (err < 0) 488 goto error; 489 else if (err == 0) { 490 err = -EREMOTEIO; 491 goto error; 492 } 493 494 m->wpos += err; 495 if (m->wpos == m->wsize) 496 m->wpos = m->wsize = 0; 497 498 end_clear: 499 clear_bit(Wworksched, &m->wsched); 500 501 if (m->wsize || !list_empty(&m->unsent_req_list)) { 502 if (test_and_clear_bit(Wpending, &m->wsched)) 503 n = EPOLLOUT; 504 else 505 n = p9_fd_poll(m->client, NULL, NULL); 506 507 if ((n & EPOLLOUT) && 508 !test_and_set_bit(Wworksched, &m->wsched)) { 509 p9_debug(P9_DEBUG_TRANS, "sched write work %p\n", m); 510 schedule_work(&m->wq); 511 } 512 } 513 514 return; 515 516 error: 517 p9_conn_cancel(m, err); 518 clear_bit(Wworksched, &m->wsched); 519 } 520 521 static int p9_pollwake(wait_queue_entry_t *wait, unsigned int mode, int sync, void *key) 522 { 523 struct p9_poll_wait *pwait = 524 container_of(wait, struct p9_poll_wait, wait); 525 struct p9_conn *m = pwait->conn; 526 unsigned long flags; 527 528 spin_lock_irqsave(&p9_poll_lock, flags); 529 if (list_empty(&m->poll_pending_link)) 530 list_add_tail(&m->poll_pending_link, &p9_poll_pending_list); 531 spin_unlock_irqrestore(&p9_poll_lock, flags); 532 533 schedule_work(&p9_poll_work); 534 return 1; 535 } 536 537 /** 538 * p9_pollwait - add poll task to the wait queue 539 * @filp: file pointer being polled 540 * @wait_address: wait_q to block on 541 * @p: poll state 542 * 543 * called by files poll operation to add v9fs-poll task to files wait queue 544 */ 545 546 static void 547 p9_pollwait(struct file *filp, wait_queue_head_t *wait_address, poll_table *p) 548 { 549 struct p9_conn *m = container_of(p, struct p9_conn, pt); 550 struct p9_poll_wait *pwait = NULL; 551 int i; 552 553 for (i = 0; i < ARRAY_SIZE(m->poll_wait); i++) { 554 if (m->poll_wait[i].wait_addr == NULL) { 555 pwait = &m->poll_wait[i]; 556 break; 557 } 558 } 559 560 if (!pwait) { 561 p9_debug(P9_DEBUG_ERROR, "not enough wait_address slots\n"); 562 return; 563 } 564 565 pwait->conn = m; 566 pwait->wait_addr = wait_address; 567 init_waitqueue_func_entry(&pwait->wait, p9_pollwake); 568 add_wait_queue(wait_address, &pwait->wait); 569 } 570 571 /** 572 * p9_conn_create - initialize the per-session mux data 573 * @client: client instance 574 * 575 * Note: Creates the polling task if this is the first session. 576 */ 577 578 static void p9_conn_create(struct p9_client *client) 579 { 580 __poll_t n; 581 struct p9_trans_fd *ts = client->trans; 582 struct p9_conn *m = &ts->conn; 583 584 p9_debug(P9_DEBUG_TRANS, "client %p msize %d\n", client, client->msize); 585 586 INIT_LIST_HEAD(&m->mux_list); 587 m->client = client; 588 589 INIT_LIST_HEAD(&m->req_list); 590 INIT_LIST_HEAD(&m->unsent_req_list); 591 INIT_WORK(&m->rq, p9_read_work); 592 INIT_WORK(&m->wq, p9_write_work); 593 INIT_LIST_HEAD(&m->poll_pending_link); 594 init_poll_funcptr(&m->pt, p9_pollwait); 595 596 n = p9_fd_poll(client, &m->pt, NULL); 597 if (n & EPOLLIN) { 598 p9_debug(P9_DEBUG_TRANS, "mux %p can read\n", m); 599 set_bit(Rpending, &m->wsched); 600 } 601 602 if (n & EPOLLOUT) { 603 p9_debug(P9_DEBUG_TRANS, "mux %p can write\n", m); 604 set_bit(Wpending, &m->wsched); 605 } 606 } 607 608 /** 609 * p9_poll_mux - polls a mux and schedules read or write works if necessary 610 * @m: connection to poll 611 * 612 */ 613 614 static void p9_poll_mux(struct p9_conn *m) 615 { 616 __poll_t n; 617 int err = -ECONNRESET; 618 619 if (m->err < 0) 620 return; 621 622 n = p9_fd_poll(m->client, NULL, &err); 623 if (n & (EPOLLERR | EPOLLHUP | EPOLLNVAL)) { 624 p9_debug(P9_DEBUG_TRANS, "error mux %p err %d\n", m, n); 625 p9_conn_cancel(m, err); 626 } 627 628 if (n & EPOLLIN) { 629 set_bit(Rpending, &m->wsched); 630 p9_debug(P9_DEBUG_TRANS, "mux %p can read\n", m); 631 if (!test_and_set_bit(Rworksched, &m->wsched)) { 632 p9_debug(P9_DEBUG_TRANS, "sched read work %p\n", m); 633 schedule_work(&m->rq); 634 } 635 } 636 637 if (n & EPOLLOUT) { 638 set_bit(Wpending, &m->wsched); 639 p9_debug(P9_DEBUG_TRANS, "mux %p can write\n", m); 640 if ((m->wsize || !list_empty(&m->unsent_req_list)) && 641 !test_and_set_bit(Wworksched, &m->wsched)) { 642 p9_debug(P9_DEBUG_TRANS, "sched write work %p\n", m); 643 schedule_work(&m->wq); 644 } 645 } 646 } 647 648 /** 649 * p9_fd_request - send 9P request 650 * The function can sleep until the request is scheduled for sending. 651 * The function can be interrupted. Return from the function is not 652 * a guarantee that the request is sent successfully. 653 * 654 * @client: client instance 655 * @req: request to be sent 656 * 657 */ 658 659 static int p9_fd_request(struct p9_client *client, struct p9_req_t *req) 660 { 661 __poll_t n; 662 struct p9_trans_fd *ts = client->trans; 663 struct p9_conn *m = &ts->conn; 664 665 p9_debug(P9_DEBUG_TRANS, "mux %p task %p tcall %p id %d\n", 666 m, current, req->tc, req->tc->id); 667 if (m->err < 0) 668 return m->err; 669 670 spin_lock(&client->lock); 671 req->status = REQ_STATUS_UNSENT; 672 list_add_tail(&req->req_list, &m->unsent_req_list); 673 spin_unlock(&client->lock); 674 675 if (test_and_clear_bit(Wpending, &m->wsched)) 676 n = EPOLLOUT; 677 else 678 n = p9_fd_poll(m->client, NULL, NULL); 679 680 if (n & EPOLLOUT && !test_and_set_bit(Wworksched, &m->wsched)) 681 schedule_work(&m->wq); 682 683 return 0; 684 } 685 686 static int p9_fd_cancel(struct p9_client *client, struct p9_req_t *req) 687 { 688 int ret = 1; 689 690 p9_debug(P9_DEBUG_TRANS, "client %p req %p\n", client, req); 691 692 spin_lock(&client->lock); 693 694 if (req->status == REQ_STATUS_UNSENT) { 695 list_del(&req->req_list); 696 req->status = REQ_STATUS_FLSHD; 697 ret = 0; 698 } 699 spin_unlock(&client->lock); 700 701 return ret; 702 } 703 704 static int p9_fd_cancelled(struct p9_client *client, struct p9_req_t *req) 705 { 706 p9_debug(P9_DEBUG_TRANS, "client %p req %p\n", client, req); 707 708 /* we haven't received a response for oldreq, 709 * remove it from the list. 710 */ 711 spin_lock(&client->lock); 712 list_del(&req->req_list); 713 spin_unlock(&client->lock); 714 715 return 0; 716 } 717 718 static int p9_fd_show_options(struct seq_file *m, struct p9_client *clnt) 719 { 720 if (clnt->trans_mod == &p9_tcp_trans) { 721 if (clnt->trans_opts.tcp.port != P9_PORT) 722 seq_printf(m, ",port=%u", clnt->trans_opts.tcp.port); 723 } else if (clnt->trans_mod == &p9_fd_trans) { 724 if (clnt->trans_opts.fd.rfd != ~0) 725 seq_printf(m, ",rfd=%u", clnt->trans_opts.fd.rfd); 726 if (clnt->trans_opts.fd.wfd != ~0) 727 seq_printf(m, ",wfd=%u", clnt->trans_opts.fd.wfd); 728 } 729 return 0; 730 } 731 732 /** 733 * parse_opts - parse mount options into p9_fd_opts structure 734 * @params: options string passed from mount 735 * @opts: fd transport-specific structure to parse options into 736 * 737 * Returns 0 upon success, -ERRNO upon failure 738 */ 739 740 static int parse_opts(char *params, struct p9_fd_opts *opts) 741 { 742 char *p; 743 substring_t args[MAX_OPT_ARGS]; 744 int option; 745 char *options, *tmp_options; 746 747 opts->port = P9_PORT; 748 opts->rfd = ~0; 749 opts->wfd = ~0; 750 opts->privport = false; 751 752 if (!params) 753 return 0; 754 755 tmp_options = kstrdup(params, GFP_KERNEL); 756 if (!tmp_options) { 757 p9_debug(P9_DEBUG_ERROR, 758 "failed to allocate copy of option string\n"); 759 return -ENOMEM; 760 } 761 options = tmp_options; 762 763 while ((p = strsep(&options, ",")) != NULL) { 764 int token; 765 int r; 766 if (!*p) 767 continue; 768 token = match_token(p, tokens, args); 769 if ((token != Opt_err) && (token != Opt_privport)) { 770 r = match_int(&args[0], &option); 771 if (r < 0) { 772 p9_debug(P9_DEBUG_ERROR, 773 "integer field, but no integer?\n"); 774 continue; 775 } 776 } 777 switch (token) { 778 case Opt_port: 779 opts->port = option; 780 break; 781 case Opt_rfdno: 782 opts->rfd = option; 783 break; 784 case Opt_wfdno: 785 opts->wfd = option; 786 break; 787 case Opt_privport: 788 opts->privport = true; 789 break; 790 default: 791 continue; 792 } 793 } 794 795 kfree(tmp_options); 796 return 0; 797 } 798 799 static int p9_fd_open(struct p9_client *client, int rfd, int wfd) 800 { 801 struct p9_trans_fd *ts = kzalloc(sizeof(struct p9_trans_fd), 802 GFP_KERNEL); 803 if (!ts) 804 return -ENOMEM; 805 806 ts->rd = fget(rfd); 807 ts->wr = fget(wfd); 808 if (!ts->rd || !ts->wr) { 809 if (ts->rd) 810 fput(ts->rd); 811 if (ts->wr) 812 fput(ts->wr); 813 kfree(ts); 814 return -EIO; 815 } 816 817 client->trans = ts; 818 client->status = Connected; 819 820 return 0; 821 } 822 823 static int p9_socket_open(struct p9_client *client, struct socket *csocket) 824 { 825 struct p9_trans_fd *p; 826 struct file *file; 827 828 p = kzalloc(sizeof(struct p9_trans_fd), GFP_KERNEL); 829 if (!p) 830 return -ENOMEM; 831 832 csocket->sk->sk_allocation = GFP_NOIO; 833 file = sock_alloc_file(csocket, 0, NULL); 834 if (IS_ERR(file)) { 835 pr_err("%s (%d): failed to map fd\n", 836 __func__, task_pid_nr(current)); 837 kfree(p); 838 return PTR_ERR(file); 839 } 840 841 get_file(file); 842 p->wr = p->rd = file; 843 client->trans = p; 844 client->status = Connected; 845 846 p->rd->f_flags |= O_NONBLOCK; 847 848 p9_conn_create(client); 849 return 0; 850 } 851 852 /** 853 * p9_mux_destroy - cancels all pending requests of mux 854 * @m: mux to destroy 855 * 856 */ 857 858 static void p9_conn_destroy(struct p9_conn *m) 859 { 860 p9_debug(P9_DEBUG_TRANS, "mux %p prev %p next %p\n", 861 m, m->mux_list.prev, m->mux_list.next); 862 863 p9_mux_poll_stop(m); 864 cancel_work_sync(&m->rq); 865 cancel_work_sync(&m->wq); 866 867 p9_conn_cancel(m, -ECONNRESET); 868 869 m->client = NULL; 870 } 871 872 /** 873 * p9_fd_close - shutdown file descriptor transport 874 * @client: client instance 875 * 876 */ 877 878 static void p9_fd_close(struct p9_client *client) 879 { 880 struct p9_trans_fd *ts; 881 882 if (!client) 883 return; 884 885 ts = client->trans; 886 if (!ts) 887 return; 888 889 client->status = Disconnected; 890 891 p9_conn_destroy(&ts->conn); 892 893 if (ts->rd) 894 fput(ts->rd); 895 if (ts->wr) 896 fput(ts->wr); 897 898 kfree(ts); 899 } 900 901 /* 902 * stolen from NFS - maybe should be made a generic function? 903 */ 904 static inline int valid_ipaddr4(const char *buf) 905 { 906 int rc, count, in[4]; 907 908 rc = sscanf(buf, "%d.%d.%d.%d", &in[0], &in[1], &in[2], &in[3]); 909 if (rc != 4) 910 return -EINVAL; 911 for (count = 0; count < 4; count++) { 912 if (in[count] > 255) 913 return -EINVAL; 914 } 915 return 0; 916 } 917 918 static int p9_bind_privport(struct socket *sock) 919 { 920 struct sockaddr_in cl; 921 int port, err = -EINVAL; 922 923 memset(&cl, 0, sizeof(cl)); 924 cl.sin_family = AF_INET; 925 cl.sin_addr.s_addr = INADDR_ANY; 926 for (port = p9_ipport_resv_max; port >= p9_ipport_resv_min; port--) { 927 cl.sin_port = htons((ushort)port); 928 err = kernel_bind(sock, (struct sockaddr *)&cl, sizeof(cl)); 929 if (err != -EADDRINUSE) 930 break; 931 } 932 return err; 933 } 934 935 936 static int 937 p9_fd_create_tcp(struct p9_client *client, const char *addr, char *args) 938 { 939 int err; 940 struct socket *csocket; 941 struct sockaddr_in sin_server; 942 struct p9_fd_opts opts; 943 944 err = parse_opts(args, &opts); 945 if (err < 0) 946 return err; 947 948 if (addr == NULL || valid_ipaddr4(addr) < 0) 949 return -EINVAL; 950 951 csocket = NULL; 952 953 client->trans_opts.tcp.port = opts.port; 954 client->trans_opts.tcp.privport = opts.privport; 955 sin_server.sin_family = AF_INET; 956 sin_server.sin_addr.s_addr = in_aton(addr); 957 sin_server.sin_port = htons(opts.port); 958 err = __sock_create(current->nsproxy->net_ns, PF_INET, 959 SOCK_STREAM, IPPROTO_TCP, &csocket, 1); 960 if (err) { 961 pr_err("%s (%d): problem creating socket\n", 962 __func__, task_pid_nr(current)); 963 return err; 964 } 965 966 if (opts.privport) { 967 err = p9_bind_privport(csocket); 968 if (err < 0) { 969 pr_err("%s (%d): problem binding to privport\n", 970 __func__, task_pid_nr(current)); 971 sock_release(csocket); 972 return err; 973 } 974 } 975 976 err = csocket->ops->connect(csocket, 977 (struct sockaddr *)&sin_server, 978 sizeof(struct sockaddr_in), 0); 979 if (err < 0) { 980 pr_err("%s (%d): problem connecting socket to %s\n", 981 __func__, task_pid_nr(current), addr); 982 sock_release(csocket); 983 return err; 984 } 985 986 return p9_socket_open(client, csocket); 987 } 988 989 static int 990 p9_fd_create_unix(struct p9_client *client, const char *addr, char *args) 991 { 992 int err; 993 struct socket *csocket; 994 struct sockaddr_un sun_server; 995 996 csocket = NULL; 997 998 if (addr == NULL) 999 return -EINVAL; 1000 1001 if (strlen(addr) >= UNIX_PATH_MAX) { 1002 pr_err("%s (%d): address too long: %s\n", 1003 __func__, task_pid_nr(current), addr); 1004 return -ENAMETOOLONG; 1005 } 1006 1007 sun_server.sun_family = PF_UNIX; 1008 strcpy(sun_server.sun_path, addr); 1009 err = __sock_create(current->nsproxy->net_ns, PF_UNIX, 1010 SOCK_STREAM, 0, &csocket, 1); 1011 if (err < 0) { 1012 pr_err("%s (%d): problem creating socket\n", 1013 __func__, task_pid_nr(current)); 1014 1015 return err; 1016 } 1017 err = csocket->ops->connect(csocket, (struct sockaddr *)&sun_server, 1018 sizeof(struct sockaddr_un) - 1, 0); 1019 if (err < 0) { 1020 pr_err("%s (%d): problem connecting socket: %s: %d\n", 1021 __func__, task_pid_nr(current), addr, err); 1022 sock_release(csocket); 1023 return err; 1024 } 1025 1026 return p9_socket_open(client, csocket); 1027 } 1028 1029 static int 1030 p9_fd_create(struct p9_client *client, const char *addr, char *args) 1031 { 1032 int err; 1033 struct p9_fd_opts opts; 1034 1035 parse_opts(args, &opts); 1036 client->trans_opts.fd.rfd = opts.rfd; 1037 client->trans_opts.fd.wfd = opts.wfd; 1038 1039 if (opts.rfd == ~0 || opts.wfd == ~0) { 1040 pr_err("Insufficient options for proto=fd\n"); 1041 return -ENOPROTOOPT; 1042 } 1043 1044 err = p9_fd_open(client, opts.rfd, opts.wfd); 1045 if (err < 0) 1046 return err; 1047 1048 p9_conn_create(client); 1049 1050 return 0; 1051 } 1052 1053 static struct p9_trans_module p9_tcp_trans = { 1054 .name = "tcp", 1055 .maxsize = MAX_SOCK_BUF, 1056 .def = 0, 1057 .create = p9_fd_create_tcp, 1058 .close = p9_fd_close, 1059 .request = p9_fd_request, 1060 .cancel = p9_fd_cancel, 1061 .cancelled = p9_fd_cancelled, 1062 .show_options = p9_fd_show_options, 1063 .owner = THIS_MODULE, 1064 }; 1065 1066 static struct p9_trans_module p9_unix_trans = { 1067 .name = "unix", 1068 .maxsize = MAX_SOCK_BUF, 1069 .def = 0, 1070 .create = p9_fd_create_unix, 1071 .close = p9_fd_close, 1072 .request = p9_fd_request, 1073 .cancel = p9_fd_cancel, 1074 .cancelled = p9_fd_cancelled, 1075 .show_options = p9_fd_show_options, 1076 .owner = THIS_MODULE, 1077 }; 1078 1079 static struct p9_trans_module p9_fd_trans = { 1080 .name = "fd", 1081 .maxsize = MAX_SOCK_BUF, 1082 .def = 0, 1083 .create = p9_fd_create, 1084 .close = p9_fd_close, 1085 .request = p9_fd_request, 1086 .cancel = p9_fd_cancel, 1087 .cancelled = p9_fd_cancelled, 1088 .show_options = p9_fd_show_options, 1089 .owner = THIS_MODULE, 1090 }; 1091 1092 /** 1093 * p9_poll_workfn - poll worker thread 1094 * @work: work queue 1095 * 1096 * polls all v9fs transports for new events and queues the appropriate 1097 * work to the work queue 1098 * 1099 */ 1100 1101 static void p9_poll_workfn(struct work_struct *work) 1102 { 1103 unsigned long flags; 1104 1105 p9_debug(P9_DEBUG_TRANS, "start %p\n", current); 1106 1107 spin_lock_irqsave(&p9_poll_lock, flags); 1108 while (!list_empty(&p9_poll_pending_list)) { 1109 struct p9_conn *conn = list_first_entry(&p9_poll_pending_list, 1110 struct p9_conn, 1111 poll_pending_link); 1112 list_del_init(&conn->poll_pending_link); 1113 spin_unlock_irqrestore(&p9_poll_lock, flags); 1114 1115 p9_poll_mux(conn); 1116 1117 spin_lock_irqsave(&p9_poll_lock, flags); 1118 } 1119 spin_unlock_irqrestore(&p9_poll_lock, flags); 1120 1121 p9_debug(P9_DEBUG_TRANS, "finish\n"); 1122 } 1123 1124 int p9_trans_fd_init(void) 1125 { 1126 v9fs_register_trans(&p9_tcp_trans); 1127 v9fs_register_trans(&p9_unix_trans); 1128 v9fs_register_trans(&p9_fd_trans); 1129 1130 return 0; 1131 } 1132 1133 void p9_trans_fd_exit(void) 1134 { 1135 flush_work(&p9_poll_work); 1136 v9fs_unregister_trans(&p9_tcp_trans); 1137 v9fs_unregister_trans(&p9_unix_trans); 1138 v9fs_unregister_trans(&p9_fd_trans); 1139 } 1140