linux/net/9p/trans_fd.c
<<
>>
Prefs
   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 <net/9p/9p.h>
  45#include <net/9p/client.h>
  46#include <net/9p/transport.h>
  47
  48#include <linux/syscalls.h> /* killme */
  49
  50#define P9_PORT 564
  51#define MAX_SOCK_BUF (64*1024)
  52#define MAXPOLLWADDR    2
  53
  54/**
  55 * struct p9_fd_opts - per-transport options
  56 * @rfd: file descriptor for reading (trans=fd)
  57 * @wfd: file descriptor for writing (trans=fd)
  58 * @port: port to connect to (trans=tcp)
  59 *
  60 */
  61
  62struct p9_fd_opts {
  63        int rfd;
  64        int wfd;
  65        u16 port;
  66        int privport;
  67};
  68
  69/**
  70 * struct p9_trans_fd - transport state
  71 * @rd: reference to file to read from
  72 * @wr: reference of file to write to
  73 * @conn: connection state reference
  74 *
  75 */
  76
  77struct p9_trans_fd {
  78        struct file *rd;
  79        struct file *wr;
  80        struct p9_conn *conn;
  81};
  82
  83/*
  84  * Option Parsing (code inspired by NFS code)
  85  *  - a little lazy - parse all fd-transport options
  86  */
  87
  88enum {
  89        /* Options that take integer arguments */
  90        Opt_port, Opt_rfdno, Opt_wfdno, Opt_err,
  91        /* Options that take no arguments */
  92        Opt_privport,
  93};
  94
  95static const match_table_t tokens = {
  96        {Opt_port, "port=%u"},
  97        {Opt_rfdno, "rfdno=%u"},
  98        {Opt_wfdno, "wfdno=%u"},
  99        {Opt_privport, "privport"},
 100        {Opt_err, NULL},
 101};
 102
 103enum {
 104        Rworksched = 1,         /* read work scheduled or running */
 105        Rpending = 2,           /* can read */
 106        Wworksched = 4,         /* write work scheduled or running */
 107        Wpending = 8,           /* can write */
 108};
 109
 110struct p9_poll_wait {
 111        struct p9_conn *conn;
 112        wait_queue_t wait;
 113        wait_queue_head_t *wait_addr;
 114};
 115
 116/**
 117 * struct p9_conn - fd mux connection state information
 118 * @mux_list: list link for mux to manage multiple connections (?)
 119 * @client: reference to client instance for this connection
 120 * @err: error state
 121 * @req_list: accounting for requests which have been sent
 122 * @unsent_req_list: accounting for requests that haven't been sent
 123 * @req: current request being processed (if any)
 124 * @tmp_buf: temporary buffer to read in header
 125 * @rsize: amount to read for current frame
 126 * @rpos: read position in current frame
 127 * @rbuf: current read buffer
 128 * @wpos: write position for current frame
 129 * @wsize: amount of data to write for current frame
 130 * @wbuf: current write buffer
 131 * @poll_pending_link: pending links to be polled per conn
 132 * @poll_wait: array of wait_q's for various worker threads
 133 * @pt: poll state
 134 * @rq: current read work
 135 * @wq: current write work
 136 * @wsched: ????
 137 *
 138 */
 139
 140struct p9_conn {
 141        struct list_head mux_list;
 142        struct p9_client *client;
 143        int err;
 144        struct list_head req_list;
 145        struct list_head unsent_req_list;
 146        struct p9_req_t *req;
 147        char tmp_buf[7];
 148        int rsize;
 149        int rpos;
 150        char *rbuf;
 151        int wpos;
 152        int wsize;
 153        char *wbuf;
 154        struct list_head poll_pending_link;
 155        struct p9_poll_wait poll_wait[MAXPOLLWADDR];
 156        poll_table pt;
 157        struct work_struct rq;
 158        struct work_struct wq;
 159        unsigned long wsched;
 160};
 161
 162static void p9_poll_workfn(struct work_struct *work);
 163
 164static DEFINE_SPINLOCK(p9_poll_lock);
 165static LIST_HEAD(p9_poll_pending_list);
 166static DECLARE_WORK(p9_poll_work, p9_poll_workfn);
 167
 168static unsigned int p9_ipport_resv_min = P9_DEF_MIN_RESVPORT;
 169static unsigned int p9_ipport_resv_max = P9_DEF_MAX_RESVPORT;
 170
 171static 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
 190/**
 191 * p9_conn_cancel - cancel all pending requests with error
 192 * @m: mux data
 193 * @err: error code
 194 *
 195 */
 196
 197static void p9_conn_cancel(struct p9_conn *m, int err)
 198{
 199        struct p9_req_t *req, *rtmp;
 200        unsigned long flags;
 201        LIST_HEAD(cancel_list);
 202
 203        p9_debug(P9_DEBUG_ERROR, "mux %p err %d\n", m, err);
 204
 205        spin_lock_irqsave(&m->client->lock, flags);
 206
 207        if (m->err) {
 208                spin_unlock_irqrestore(&m->client->lock, flags);
 209                return;
 210        }
 211
 212        m->err = err;
 213
 214        list_for_each_entry_safe(req, rtmp, &m->req_list, req_list) {
 215                req->status = REQ_STATUS_ERROR;
 216                if (!req->t_err)
 217                        req->t_err = err;
 218                list_move(&req->req_list, &cancel_list);
 219        }
 220        list_for_each_entry_safe(req, rtmp, &m->unsent_req_list, req_list) {
 221                req->status = REQ_STATUS_ERROR;
 222                if (!req->t_err)
 223                        req->t_err = err;
 224                list_move(&req->req_list, &cancel_list);
 225        }
 226        spin_unlock_irqrestore(&m->client->lock, flags);
 227
 228        list_for_each_entry_safe(req, rtmp, &cancel_list, req_list) {
 229                p9_debug(P9_DEBUG_ERROR, "call back req %p\n", req);
 230                list_del(&req->req_list);
 231                p9_client_cb(m->client, req);
 232        }
 233}
 234
 235static int
 236p9_fd_poll(struct p9_client *client, struct poll_table_struct *pt)
 237{
 238        int ret, n;
 239        struct p9_trans_fd *ts = NULL;
 240
 241        if (client && client->status == Connected)
 242                ts = client->trans;
 243
 244        if (!ts)
 245                return -EREMOTEIO;
 246
 247        if (!ts->rd->f_op->poll)
 248                return -EIO;
 249
 250        if (!ts->wr->f_op->poll)
 251                return -EIO;
 252
 253        ret = ts->rd->f_op->poll(ts->rd, pt);
 254        if (ret < 0)
 255                return ret;
 256
 257        if (ts->rd != ts->wr) {
 258                n = ts->wr->f_op->poll(ts->wr, pt);
 259                if (n < 0)
 260                        return n;
 261                ret = (ret & ~POLLOUT) | (n & ~POLLIN);
 262        }
 263
 264        return ret;
 265}
 266
 267/**
 268 * p9_fd_read- read from a fd
 269 * @client: client instance
 270 * @v: buffer to receive data into
 271 * @len: size of receive buffer
 272 *
 273 */
 274
 275static int p9_fd_read(struct p9_client *client, void *v, int len)
 276{
 277        int ret;
 278        struct p9_trans_fd *ts = NULL;
 279
 280        if (client && client->status != Disconnected)
 281                ts = client->trans;
 282
 283        if (!ts)
 284                return -EREMOTEIO;
 285
 286        if (!(ts->rd->f_flags & O_NONBLOCK))
 287                p9_debug(P9_DEBUG_ERROR, "blocking read ...\n");
 288
 289        ret = kernel_read(ts->rd, ts->rd->f_pos, v, len);
 290        if (ret <= 0 && ret != -ERESTARTSYS && ret != -EAGAIN)
 291                client->status = Disconnected;
 292        return ret;
 293}
 294
 295/**
 296 * p9_read_work - called when there is some data to be read from a transport
 297 * @work: container of work to be done
 298 *
 299 */
 300
 301static void p9_read_work(struct work_struct *work)
 302{
 303        int n, err;
 304        struct p9_conn *m;
 305
 306        m = container_of(work, struct p9_conn, rq);
 307
 308        if (m->err < 0)
 309                return;
 310
 311        p9_debug(P9_DEBUG_TRANS, "start mux %p pos %d\n", m, m->rpos);
 312
 313        if (!m->rbuf) {
 314                m->rbuf = m->tmp_buf;
 315                m->rpos = 0;
 316                m->rsize = 7; /* start by reading header */
 317        }
 318
 319        clear_bit(Rpending, &m->wsched);
 320        p9_debug(P9_DEBUG_TRANS, "read mux %p pos %d size: %d = %d\n",
 321                 m, m->rpos, m->rsize, m->rsize-m->rpos);
 322        err = p9_fd_read(m->client, m->rbuf + m->rpos,
 323                                                m->rsize - m->rpos);
 324        p9_debug(P9_DEBUG_TRANS, "mux %p got %d bytes\n", m, err);
 325        if (err == -EAGAIN) {
 326                goto end_clear;
 327        }
 328
 329        if (err <= 0)
 330                goto error;
 331
 332        m->rpos += err;
 333
 334        if ((!m->req) && (m->rpos == m->rsize)) { /* header read in */
 335                u16 tag;
 336                p9_debug(P9_DEBUG_TRANS, "got new header\n");
 337
 338                n = le32_to_cpu(*(__le32 *) m->rbuf); /* read packet size */
 339                if (n >= m->client->msize) {
 340                        p9_debug(P9_DEBUG_ERROR,
 341                                 "requested packet size too big: %d\n", n);
 342                        err = -EIO;
 343                        goto error;
 344                }
 345
 346                tag = le16_to_cpu(*(__le16 *) (m->rbuf+5)); /* read tag */
 347                p9_debug(P9_DEBUG_TRANS,
 348                         "mux %p pkt: size: %d bytes tag: %d\n", m, n, tag);
 349
 350                m->req = p9_tag_lookup(m->client, tag);
 351                if (!m->req || (m->req->status != REQ_STATUS_SENT &&
 352                                        m->req->status != REQ_STATUS_FLSH)) {
 353                        p9_debug(P9_DEBUG_ERROR, "Unexpected packet tag %d\n",
 354                                 tag);
 355                        err = -EIO;
 356                        goto error;
 357                }
 358
 359                if (m->req->rc == NULL) {
 360                        m->req->rc = kmalloc(sizeof(struct p9_fcall) +
 361                                                m->client->msize, GFP_NOFS);
 362                        if (!m->req->rc) {
 363                                m->req = NULL;
 364                                err = -ENOMEM;
 365                                goto error;
 366                        }
 367                }
 368                m->rbuf = (char *)m->req->rc + sizeof(struct p9_fcall);
 369                memcpy(m->rbuf, m->tmp_buf, m->rsize);
 370                m->rsize = n;
 371        }
 372
 373        /* not an else because some packets (like clunk) have no payload */
 374        if ((m->req) && (m->rpos == m->rsize)) { /* packet is read in */
 375                p9_debug(P9_DEBUG_TRANS, "got new packet\n");
 376                spin_lock(&m->client->lock);
 377                if (m->req->status != REQ_STATUS_ERROR)
 378                        m->req->status = REQ_STATUS_RCVD;
 379                list_del(&m->req->req_list);
 380                spin_unlock(&m->client->lock);
 381                p9_client_cb(m->client, m->req);
 382                m->rbuf = NULL;
 383                m->rpos = 0;
 384                m->rsize = 0;
 385                m->req = NULL;
 386        }
 387
 388end_clear:
 389        clear_bit(Rworksched, &m->wsched);
 390
 391        if (!list_empty(&m->req_list)) {
 392                if (test_and_clear_bit(Rpending, &m->wsched))
 393                        n = POLLIN;
 394                else
 395                        n = p9_fd_poll(m->client, NULL);
 396
 397                if ((n & POLLIN) && !test_and_set_bit(Rworksched, &m->wsched)) {
 398                        p9_debug(P9_DEBUG_TRANS, "sched read work %p\n", m);
 399                        schedule_work(&m->rq);
 400                }
 401        }
 402
 403        return;
 404error:
 405        p9_conn_cancel(m, err);
 406        clear_bit(Rworksched, &m->wsched);
 407}
 408
 409/**
 410 * p9_fd_write - write to a socket
 411 * @client: client instance
 412 * @v: buffer to send data from
 413 * @len: size of send buffer
 414 *
 415 */
 416
 417static int p9_fd_write(struct p9_client *client, void *v, int len)
 418{
 419        int ret;
 420        mm_segment_t oldfs;
 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        oldfs = get_fs();
 433        set_fs(get_ds());
 434        /* The cast to a user pointer is valid due to the set_fs() */
 435        ret = vfs_write(ts->wr, (__force void __user *)v, len, &ts->wr->f_pos);
 436        set_fs(oldfs);
 437
 438        if (ret <= 0 && ret != -ERESTARTSYS && ret != -EAGAIN)
 439                client->status = Disconnected;
 440        return ret;
 441}
 442
 443/**
 444 * p9_write_work - called when a transport can send some data
 445 * @work: container for work to be done
 446 *
 447 */
 448
 449static void p9_write_work(struct work_struct *work)
 450{
 451        int n, err;
 452        struct p9_conn *m;
 453        struct p9_req_t *req;
 454
 455        m = container_of(work, struct p9_conn, wq);
 456
 457        if (m->err < 0) {
 458                clear_bit(Wworksched, &m->wsched);
 459                return;
 460        }
 461
 462        if (!m->wsize) {
 463                spin_lock(&m->client->lock);
 464                if (list_empty(&m->unsent_req_list)) {
 465                        clear_bit(Wworksched, &m->wsched);
 466                        spin_unlock(&m->client->lock);
 467                        return;
 468                }
 469
 470                req = list_entry(m->unsent_req_list.next, struct p9_req_t,
 471                               req_list);
 472                req->status = REQ_STATUS_SENT;
 473                p9_debug(P9_DEBUG_TRANS, "move req %p\n", req);
 474                list_move_tail(&req->req_list, &m->req_list);
 475
 476                m->wbuf = req->tc->sdata;
 477                m->wsize = req->tc->size;
 478                m->wpos = 0;
 479                spin_unlock(&m->client->lock);
 480        }
 481
 482        p9_debug(P9_DEBUG_TRANS, "mux %p pos %d size %d\n",
 483                 m, m->wpos, m->wsize);
 484        clear_bit(Wpending, &m->wsched);
 485        err = p9_fd_write(m->client, m->wbuf + m->wpos, m->wsize - m->wpos);
 486        p9_debug(P9_DEBUG_TRANS, "mux %p sent %d bytes\n", m, err);
 487        if (err == -EAGAIN)
 488                goto end_clear;
 489
 490
 491        if (err < 0)
 492                goto error;
 493        else if (err == 0) {
 494                err = -EREMOTEIO;
 495                goto error;
 496        }
 497
 498        m->wpos += err;
 499        if (m->wpos == m->wsize)
 500                m->wpos = m->wsize = 0;
 501
 502end_clear:
 503        clear_bit(Wworksched, &m->wsched);
 504
 505        if (m->wsize || !list_empty(&m->unsent_req_list)) {
 506                if (test_and_clear_bit(Wpending, &m->wsched))
 507                        n = POLLOUT;
 508                else
 509                        n = p9_fd_poll(m->client, NULL);
 510
 511                if ((n & POLLOUT) &&
 512                   !test_and_set_bit(Wworksched, &m->wsched)) {
 513                        p9_debug(P9_DEBUG_TRANS, "sched write work %p\n", m);
 514                        schedule_work(&m->wq);
 515                }
 516        }
 517
 518        return;
 519
 520error:
 521        p9_conn_cancel(m, err);
 522        clear_bit(Wworksched, &m->wsched);
 523}
 524
 525static int p9_pollwake(wait_queue_t *wait, unsigned int mode, int sync, void *key)
 526{
 527        struct p9_poll_wait *pwait =
 528                container_of(wait, struct p9_poll_wait, wait);
 529        struct p9_conn *m = pwait->conn;
 530        unsigned long flags;
 531
 532        spin_lock_irqsave(&p9_poll_lock, flags);
 533        if (list_empty(&m->poll_pending_link))
 534                list_add_tail(&m->poll_pending_link, &p9_poll_pending_list);
 535        spin_unlock_irqrestore(&p9_poll_lock, flags);
 536
 537        schedule_work(&p9_poll_work);
 538        return 1;
 539}
 540
 541/**
 542 * p9_pollwait - add poll task to the wait queue
 543 * @filp: file pointer being polled
 544 * @wait_address: wait_q to block on
 545 * @p: poll state
 546 *
 547 * called by files poll operation to add v9fs-poll task to files wait queue
 548 */
 549
 550static void
 551p9_pollwait(struct file *filp, wait_queue_head_t *wait_address, poll_table *p)
 552{
 553        struct p9_conn *m = container_of(p, struct p9_conn, pt);
 554        struct p9_poll_wait *pwait = NULL;
 555        int i;
 556
 557        for (i = 0; i < ARRAY_SIZE(m->poll_wait); i++) {
 558                if (m->poll_wait[i].wait_addr == NULL) {
 559                        pwait = &m->poll_wait[i];
 560                        break;
 561                }
 562        }
 563
 564        if (!pwait) {
 565                p9_debug(P9_DEBUG_ERROR, "not enough wait_address slots\n");
 566                return;
 567        }
 568
 569        pwait->conn = m;
 570        pwait->wait_addr = wait_address;
 571        init_waitqueue_func_entry(&pwait->wait, p9_pollwake);
 572        add_wait_queue(wait_address, &pwait->wait);
 573}
 574
 575/**
 576 * p9_conn_create - allocate and initialize the per-session mux data
 577 * @client: client instance
 578 *
 579 * Note: Creates the polling task if this is the first session.
 580 */
 581
 582static struct p9_conn *p9_conn_create(struct p9_client *client)
 583{
 584        int n;
 585        struct p9_conn *m;
 586
 587        p9_debug(P9_DEBUG_TRANS, "client %p msize %d\n", client, client->msize);
 588        m = kzalloc(sizeof(struct p9_conn), GFP_KERNEL);
 589        if (!m)
 590                return ERR_PTR(-ENOMEM);
 591
 592        INIT_LIST_HEAD(&m->mux_list);
 593        m->client = client;
 594
 595        INIT_LIST_HEAD(&m->req_list);
 596        INIT_LIST_HEAD(&m->unsent_req_list);
 597        INIT_WORK(&m->rq, p9_read_work);
 598        INIT_WORK(&m->wq, p9_write_work);
 599        INIT_LIST_HEAD(&m->poll_pending_link);
 600        init_poll_funcptr(&m->pt, p9_pollwait);
 601
 602        n = p9_fd_poll(client, &m->pt);
 603        if (n & POLLIN) {
 604                p9_debug(P9_DEBUG_TRANS, "mux %p can read\n", m);
 605                set_bit(Rpending, &m->wsched);
 606        }
 607
 608        if (n & POLLOUT) {
 609                p9_debug(P9_DEBUG_TRANS, "mux %p can write\n", m);
 610                set_bit(Wpending, &m->wsched);
 611        }
 612
 613        return m;
 614}
 615
 616/**
 617 * p9_poll_mux - polls a mux and schedules read or write works if necessary
 618 * @m: connection to poll
 619 *
 620 */
 621
 622static void p9_poll_mux(struct p9_conn *m)
 623{
 624        int n;
 625
 626        if (m->err < 0)
 627                return;
 628
 629        n = p9_fd_poll(m->client, NULL);
 630        if (n < 0 || n & (POLLERR | POLLHUP | POLLNVAL)) {
 631                p9_debug(P9_DEBUG_TRANS, "error mux %p err %d\n", m, n);
 632                if (n >= 0)
 633                        n = -ECONNRESET;
 634                p9_conn_cancel(m, n);
 635        }
 636
 637        if (n & POLLIN) {
 638                set_bit(Rpending, &m->wsched);
 639                p9_debug(P9_DEBUG_TRANS, "mux %p can read\n", m);
 640                if (!test_and_set_bit(Rworksched, &m->wsched)) {
 641                        p9_debug(P9_DEBUG_TRANS, "sched read work %p\n", m);
 642                        schedule_work(&m->rq);
 643                }
 644        }
 645
 646        if (n & POLLOUT) {
 647                set_bit(Wpending, &m->wsched);
 648                p9_debug(P9_DEBUG_TRANS, "mux %p can write\n", m);
 649                if ((m->wsize || !list_empty(&m->unsent_req_list)) &&
 650                    !test_and_set_bit(Wworksched, &m->wsched)) {
 651                        p9_debug(P9_DEBUG_TRANS, "sched write work %p\n", m);
 652                        schedule_work(&m->wq);
 653                }
 654        }
 655}
 656
 657/**
 658 * p9_fd_request - send 9P request
 659 * The function can sleep until the request is scheduled for sending.
 660 * The function can be interrupted. Return from the function is not
 661 * a guarantee that the request is sent successfully.
 662 *
 663 * @client: client instance
 664 * @req: request to be sent
 665 *
 666 */
 667
 668static int p9_fd_request(struct p9_client *client, struct p9_req_t *req)
 669{
 670        int n;
 671        struct p9_trans_fd *ts = client->trans;
 672        struct p9_conn *m = ts->conn;
 673
 674        p9_debug(P9_DEBUG_TRANS, "mux %p task %p tcall %p id %d\n",
 675                 m, current, req->tc, req->tc->id);
 676        if (m->err < 0)
 677                return m->err;
 678
 679        spin_lock(&client->lock);
 680        req->status = REQ_STATUS_UNSENT;
 681        list_add_tail(&req->req_list, &m->unsent_req_list);
 682        spin_unlock(&client->lock);
 683
 684        if (test_and_clear_bit(Wpending, &m->wsched))
 685                n = POLLOUT;
 686        else
 687                n = p9_fd_poll(m->client, NULL);
 688
 689        if (n & POLLOUT && !test_and_set_bit(Wworksched, &m->wsched))
 690                schedule_work(&m->wq);
 691
 692        return 0;
 693}
 694
 695static int p9_fd_cancel(struct p9_client *client, struct p9_req_t *req)
 696{
 697        int ret = 1;
 698
 699        p9_debug(P9_DEBUG_TRANS, "client %p req %p\n", client, req);
 700
 701        spin_lock(&client->lock);
 702
 703        if (req->status == REQ_STATUS_UNSENT) {
 704                list_del(&req->req_list);
 705                req->status = REQ_STATUS_FLSHD;
 706                ret = 0;
 707        } else if (req->status == REQ_STATUS_SENT)
 708                req->status = REQ_STATUS_FLSH;
 709
 710        spin_unlock(&client->lock);
 711
 712        return ret;
 713}
 714
 715/**
 716 * parse_opts - parse mount options into p9_fd_opts structure
 717 * @params: options string passed from mount
 718 * @opts: fd transport-specific structure to parse options into
 719 *
 720 * Returns 0 upon success, -ERRNO upon failure
 721 */
 722
 723static int parse_opts(char *params, struct p9_fd_opts *opts)
 724{
 725        char *p;
 726        substring_t args[MAX_OPT_ARGS];
 727        int option;
 728        char *options, *tmp_options;
 729
 730        opts->port = P9_PORT;
 731        opts->rfd = ~0;
 732        opts->wfd = ~0;
 733
 734        if (!params)
 735                return 0;
 736
 737        tmp_options = kstrdup(params, GFP_KERNEL);
 738        if (!tmp_options) {
 739                p9_debug(P9_DEBUG_ERROR,
 740                         "failed to allocate copy of option string\n");
 741                return -ENOMEM;
 742        }
 743        options = tmp_options;
 744
 745        while ((p = strsep(&options, ",")) != NULL) {
 746                int token;
 747                int r;
 748                if (!*p)
 749                        continue;
 750                token = match_token(p, tokens, args);
 751                if ((token != Opt_err) && (token != Opt_privport)) {
 752                        r = match_int(&args[0], &option);
 753                        if (r < 0) {
 754                                p9_debug(P9_DEBUG_ERROR,
 755                                         "integer field, but no integer?\n");
 756                                continue;
 757                        }
 758                }
 759                switch (token) {
 760                case Opt_port:
 761                        opts->port = option;
 762                        break;
 763                case Opt_rfdno:
 764                        opts->rfd = option;
 765                        break;
 766                case Opt_wfdno:
 767                        opts->wfd = option;
 768                        break;
 769                case Opt_privport:
 770                        opts->privport = 1;
 771                        break;
 772                default:
 773                        continue;
 774                }
 775        }
 776
 777        kfree(tmp_options);
 778        return 0;
 779}
 780
 781static int p9_fd_open(struct p9_client *client, int rfd, int wfd)
 782{
 783        struct p9_trans_fd *ts = kmalloc(sizeof(struct p9_trans_fd),
 784                                           GFP_KERNEL);
 785        if (!ts)
 786                return -ENOMEM;
 787
 788        ts->rd = fget(rfd);
 789        ts->wr = fget(wfd);
 790        if (!ts->rd || !ts->wr) {
 791                if (ts->rd)
 792                        fput(ts->rd);
 793                if (ts->wr)
 794                        fput(ts->wr);
 795                kfree(ts);
 796                return -EIO;
 797        }
 798
 799        client->trans = ts;
 800        client->status = Connected;
 801
 802        return 0;
 803}
 804
 805static int p9_socket_open(struct p9_client *client, struct socket *csocket)
 806{
 807        struct p9_trans_fd *p;
 808        struct file *file;
 809        int ret;
 810
 811        p = kmalloc(sizeof(struct p9_trans_fd), GFP_KERNEL);
 812        if (!p)
 813                return -ENOMEM;
 814
 815        csocket->sk->sk_allocation = GFP_NOIO;
 816        file = sock_alloc_file(csocket, 0, NULL);
 817        if (IS_ERR(file)) {
 818                pr_err("%s (%d): failed to map fd\n",
 819                       __func__, task_pid_nr(current));
 820                sock_release(csocket);
 821                kfree(p);
 822                return PTR_ERR(file);
 823        }
 824
 825        get_file(file);
 826        p->wr = p->rd = file;
 827        client->trans = p;
 828        client->status = Connected;
 829
 830        p->rd->f_flags |= O_NONBLOCK;
 831
 832        p->conn = p9_conn_create(client);
 833        if (IS_ERR(p->conn)) {
 834                ret = PTR_ERR(p->conn);
 835                p->conn = NULL;
 836                kfree(p);
 837                sockfd_put(csocket);
 838                sockfd_put(csocket);
 839                return ret;
 840        }
 841        return 0;
 842}
 843
 844/**
 845 * p9_mux_destroy - cancels all pending requests and frees mux resources
 846 * @m: mux to destroy
 847 *
 848 */
 849
 850static void p9_conn_destroy(struct p9_conn *m)
 851{
 852        p9_debug(P9_DEBUG_TRANS, "mux %p prev %p next %p\n",
 853                 m, m->mux_list.prev, m->mux_list.next);
 854
 855        p9_mux_poll_stop(m);
 856        cancel_work_sync(&m->rq);
 857        cancel_work_sync(&m->wq);
 858
 859        p9_conn_cancel(m, -ECONNRESET);
 860
 861        m->client = NULL;
 862        kfree(m);
 863}
 864
 865/**
 866 * p9_fd_close - shutdown file descriptor transport
 867 * @client: client instance
 868 *
 869 */
 870
 871static void p9_fd_close(struct p9_client *client)
 872{
 873        struct p9_trans_fd *ts;
 874
 875        if (!client)
 876                return;
 877
 878        ts = client->trans;
 879        if (!ts)
 880                return;
 881
 882        client->status = Disconnected;
 883
 884        p9_conn_destroy(ts->conn);
 885
 886        if (ts->rd)
 887                fput(ts->rd);
 888        if (ts->wr)
 889                fput(ts->wr);
 890
 891        kfree(ts);
 892}
 893
 894/*
 895 * stolen from NFS - maybe should be made a generic function?
 896 */
 897static inline int valid_ipaddr4(const char *buf)
 898{
 899        int rc, count, in[4];
 900
 901        rc = sscanf(buf, "%d.%d.%d.%d", &in[0], &in[1], &in[2], &in[3]);
 902        if (rc != 4)
 903                return -EINVAL;
 904        for (count = 0; count < 4; count++) {
 905                if (in[count] > 255)
 906                        return -EINVAL;
 907        }
 908        return 0;
 909}
 910
 911static int p9_bind_privport(struct socket *sock)
 912{
 913        struct sockaddr_in cl;
 914        int port, err = -EINVAL;
 915
 916        memset(&cl, 0, sizeof(cl));
 917        cl.sin_family = AF_INET;
 918        cl.sin_addr.s_addr = INADDR_ANY;
 919        for (port = p9_ipport_resv_max; port >= p9_ipport_resv_min; port--) {
 920                cl.sin_port = htons((ushort)port);
 921                err = kernel_bind(sock, (struct sockaddr *)&cl, sizeof(cl));
 922                if (err != -EADDRINUSE)
 923                        break;
 924        }
 925        return err;
 926}
 927
 928
 929static int
 930p9_fd_create_tcp(struct p9_client *client, const char *addr, char *args)
 931{
 932        int err;
 933        struct socket *csocket;
 934        struct sockaddr_in sin_server;
 935        struct p9_fd_opts opts;
 936
 937        err = parse_opts(args, &opts);
 938        if (err < 0)
 939                return err;
 940
 941        if (valid_ipaddr4(addr) < 0)
 942                return -EINVAL;
 943
 944        csocket = NULL;
 945
 946        sin_server.sin_family = AF_INET;
 947        sin_server.sin_addr.s_addr = in_aton(addr);
 948        sin_server.sin_port = htons(opts.port);
 949        err = __sock_create(read_pnet(&current->nsproxy->net_ns), PF_INET,
 950                            SOCK_STREAM, IPPROTO_TCP, &csocket, 1);
 951        if (err) {
 952                pr_err("%s (%d): problem creating socket\n",
 953                       __func__, task_pid_nr(current));
 954                return err;
 955        }
 956
 957        if (opts.privport) {
 958                err = p9_bind_privport(csocket);
 959                if (err < 0) {
 960                        pr_err("%s (%d): problem binding to privport\n",
 961                               __func__, task_pid_nr(current));
 962                        sock_release(csocket);
 963                        return err;
 964                }
 965        }
 966
 967        err = csocket->ops->connect(csocket,
 968                                    (struct sockaddr *)&sin_server,
 969                                    sizeof(struct sockaddr_in), 0);
 970        if (err < 0) {
 971                pr_err("%s (%d): problem connecting socket to %s\n",
 972                       __func__, task_pid_nr(current), addr);
 973                sock_release(csocket);
 974                return err;
 975        }
 976
 977        return p9_socket_open(client, csocket);
 978}
 979
 980static int
 981p9_fd_create_unix(struct p9_client *client, const char *addr, char *args)
 982{
 983        int err;
 984        struct socket *csocket;
 985        struct sockaddr_un sun_server;
 986
 987        csocket = NULL;
 988
 989        if (strlen(addr) >= UNIX_PATH_MAX) {
 990                pr_err("%s (%d): address too long: %s\n",
 991                       __func__, task_pid_nr(current), addr);
 992                return -ENAMETOOLONG;
 993        }
 994
 995        sun_server.sun_family = PF_UNIX;
 996        strcpy(sun_server.sun_path, addr);
 997        err = __sock_create(read_pnet(&current->nsproxy->net_ns), PF_UNIX,
 998                            SOCK_STREAM, 0, &csocket, 1);
 999        if (err < 0) {
1000                pr_err("%s (%d): problem creating socket\n",
1001                       __func__, task_pid_nr(current));
1002
1003                return err;
1004        }
1005        err = csocket->ops->connect(csocket, (struct sockaddr *)&sun_server,
1006                        sizeof(struct sockaddr_un) - 1, 0);
1007        if (err < 0) {
1008                pr_err("%s (%d): problem connecting socket: %s: %d\n",
1009                       __func__, task_pid_nr(current), addr, err);
1010                sock_release(csocket);
1011                return err;
1012        }
1013
1014        return p9_socket_open(client, csocket);
1015}
1016
1017static int
1018p9_fd_create(struct p9_client *client, const char *addr, char *args)
1019{
1020        int err;
1021        struct p9_fd_opts opts;
1022        struct p9_trans_fd *p;
1023
1024        parse_opts(args, &opts);
1025
1026        if (opts.rfd == ~0 || opts.wfd == ~0) {
1027                pr_err("Insufficient options for proto=fd\n");
1028                return -ENOPROTOOPT;
1029        }
1030
1031        err = p9_fd_open(client, opts.rfd, opts.wfd);
1032        if (err < 0)
1033                return err;
1034
1035        p = (struct p9_trans_fd *) client->trans;
1036        p->conn = p9_conn_create(client);
1037        if (IS_ERR(p->conn)) {
1038                err = PTR_ERR(p->conn);
1039                p->conn = NULL;
1040                fput(p->rd);
1041                fput(p->wr);
1042                return err;
1043        }
1044
1045        return 0;
1046}
1047
1048static struct p9_trans_module p9_tcp_trans = {
1049        .name = "tcp",
1050        .maxsize = MAX_SOCK_BUF,
1051        .def = 0,
1052        .create = p9_fd_create_tcp,
1053        .close = p9_fd_close,
1054        .request = p9_fd_request,
1055        .cancel = p9_fd_cancel,
1056        .owner = THIS_MODULE,
1057};
1058
1059static struct p9_trans_module p9_unix_trans = {
1060        .name = "unix",
1061        .maxsize = MAX_SOCK_BUF,
1062        .def = 0,
1063        .create = p9_fd_create_unix,
1064        .close = p9_fd_close,
1065        .request = p9_fd_request,
1066        .cancel = p9_fd_cancel,
1067        .owner = THIS_MODULE,
1068};
1069
1070static struct p9_trans_module p9_fd_trans = {
1071        .name = "fd",
1072        .maxsize = MAX_SOCK_BUF,
1073        .def = 0,
1074        .create = p9_fd_create,
1075        .close = p9_fd_close,
1076        .request = p9_fd_request,
1077        .cancel = p9_fd_cancel,
1078        .owner = THIS_MODULE,
1079};
1080
1081/**
1082 * p9_poll_proc - poll worker thread
1083 * @a: thread state and arguments
1084 *
1085 * polls all v9fs transports for new events and queues the appropriate
1086 * work to the work queue
1087 *
1088 */
1089
1090static void p9_poll_workfn(struct work_struct *work)
1091{
1092        unsigned long flags;
1093
1094        p9_debug(P9_DEBUG_TRANS, "start %p\n", current);
1095
1096        spin_lock_irqsave(&p9_poll_lock, flags);
1097        while (!list_empty(&p9_poll_pending_list)) {
1098                struct p9_conn *conn = list_first_entry(&p9_poll_pending_list,
1099                                                        struct p9_conn,
1100                                                        poll_pending_link);
1101                list_del_init(&conn->poll_pending_link);
1102                spin_unlock_irqrestore(&p9_poll_lock, flags);
1103
1104                p9_poll_mux(conn);
1105
1106                spin_lock_irqsave(&p9_poll_lock, flags);
1107        }
1108        spin_unlock_irqrestore(&p9_poll_lock, flags);
1109
1110        p9_debug(P9_DEBUG_TRANS, "finish\n");
1111}
1112
1113int p9_trans_fd_init(void)
1114{
1115        v9fs_register_trans(&p9_tcp_trans);
1116        v9fs_register_trans(&p9_unix_trans);
1117        v9fs_register_trans(&p9_fd_trans);
1118
1119        return 0;
1120}
1121
1122void p9_trans_fd_exit(void)
1123{
1124        flush_work(&p9_poll_work);
1125        v9fs_unregister_trans(&p9_tcp_trans);
1126        v9fs_unregister_trans(&p9_unix_trans);
1127        v9fs_unregister_trans(&p9_fd_trans);
1128}
1129