1 /* $NetBSD: kern_descrip.c,v 1.218 2012/01/25 00:28:35 christos Exp $ */ 2 3 /*- 4 * Copyright (c) 2008, 2009 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Andrew Doran. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 29 * POSSIBILITY OF SUCH DAMAGE. 30 */ 31 32 /* 33 * Copyright (c) 1982, 1986, 1989, 1991, 1993 34 * The Regents of the University of California. All rights reserved. 35 * (c) UNIX System Laboratories, Inc. 36 * All or some portions of this file are derived from material licensed 37 * to the University of California by American Telephone and Telegraph 38 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 39 * the permission of UNIX System Laboratories, Inc. 40 * 41 * Redistribution and use in source and binary forms, with or without 42 * modification, are permitted provided that the following conditions 43 * are met: 44 * 1. Redistributions of source code must retain the above copyright 45 * notice, this list of conditions and the following disclaimer. 46 * 2. Redistributions in binary form must reproduce the above copyright 47 * notice, this list of conditions and the following disclaimer in the 48 * documentation and/or other materials provided with the distribution. 49 * 3. Neither the name of the University nor the names of its contributors 50 * may be used to endorse or promote products derived from this software 51 * without specific prior written permission. 52 * 53 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 54 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 55 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 56 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 57 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 58 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 59 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 60 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 61 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 62 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 63 * SUCH DAMAGE. 64 * 65 * @(#)kern_descrip.c 8.8 (Berkeley) 2/14/95 66 */ 67 68 /* 69 * File descriptor management. 70 */ 71 72 #include <sys/cdefs.h> 73 __KERNEL_RCSID(0, "$NetBSD: kern_descrip.c,v 1.218 2012/01/25 00:28:35 christos Exp $"); 74 75 #include <sys/param.h> 76 #include <sys/systm.h> 77 #include <sys/filedesc.h> 78 #include <sys/kernel.h> 79 #include <sys/proc.h> 80 #include <sys/file.h> 81 #include <sys/socket.h> 82 #include <sys/socketvar.h> 83 #include <sys/stat.h> 84 #include <sys/ioctl.h> 85 #include <sys/fcntl.h> 86 #include <sys/pool.h> 87 #include <sys/unistd.h> 88 #include <sys/resourcevar.h> 89 #include <sys/conf.h> 90 #include <sys/event.h> 91 #include <sys/kauth.h> 92 #include <sys/atomic.h> 93 #include <sys/syscallargs.h> 94 #include <sys/cpu.h> 95 #include <sys/kmem.h> 96 #include <sys/vnode.h> 97 #include <sys/sysctl.h> 98 #include <sys/ktrace.h> 99 100 /* 101 * A list (head) of open files, counter, and lock protecting them. 102 */ 103 struct filelist filehead __cacheline_aligned; 104 static u_int nfiles __cacheline_aligned; 105 kmutex_t filelist_lock __cacheline_aligned; 106 107 static pool_cache_t filedesc_cache __read_mostly; 108 static pool_cache_t file_cache __read_mostly; 109 static pool_cache_t fdfile_cache __read_mostly; 110 111 static int file_ctor(void *, void *, int); 112 static void file_dtor(void *, void *); 113 static int fdfile_ctor(void *, void *, int); 114 static void fdfile_dtor(void *, void *); 115 static int filedesc_ctor(void *, void *, int); 116 static void filedesc_dtor(void *, void *); 117 static int filedescopen(dev_t, int, int, lwp_t *); 118 119 static int sysctl_kern_file(SYSCTLFN_PROTO); 120 static int sysctl_kern_file2(SYSCTLFN_PROTO); 121 static void fill_file(struct kinfo_file *, const file_t *, const fdfile_t *, 122 int, pid_t); 123 124 const struct cdevsw filedesc_cdevsw = { 125 filedescopen, noclose, noread, nowrite, noioctl, 126 nostop, notty, nopoll, nommap, nokqfilter, D_OTHER | D_MPSAFE, 127 }; 128 129 /* For ease of reading. */ 130 __strong_alias(fd_putvnode,fd_putfile) 131 __strong_alias(fd_putsock,fd_putfile) 132 133 /* 134 * Initialize the descriptor system. 135 */ 136 void 137 fd_sys_init(void) 138 { 139 static struct sysctllog *clog; 140 141 mutex_init(&filelist_lock, MUTEX_DEFAULT, IPL_NONE); 142 143 file_cache = pool_cache_init(sizeof(file_t), coherency_unit, 0, 144 0, "file", NULL, IPL_NONE, file_ctor, file_dtor, NULL); 145 KASSERT(file_cache != NULL); 146 147 fdfile_cache = pool_cache_init(sizeof(fdfile_t), coherency_unit, 0, 148 PR_LARGECACHE, "fdfile", NULL, IPL_NONE, fdfile_ctor, fdfile_dtor, 149 NULL); 150 KASSERT(fdfile_cache != NULL); 151 152 filedesc_cache = pool_cache_init(sizeof(filedesc_t), coherency_unit, 153 0, 0, "filedesc", NULL, IPL_NONE, filedesc_ctor, filedesc_dtor, 154 NULL); 155 KASSERT(filedesc_cache != NULL); 156 157 sysctl_createv(&clog, 0, NULL, NULL, 158 CTLFLAG_PERMANENT, CTLTYPE_NODE, "kern", NULL, 159 NULL, 0, NULL, 0, CTL_KERN, CTL_EOL); 160 sysctl_createv(&clog, 0, NULL, NULL, 161 CTLFLAG_PERMANENT, 162 CTLTYPE_STRUCT, "file", 163 SYSCTL_DESCR("System open file table"), 164 sysctl_kern_file, 0, NULL, 0, 165 CTL_KERN, KERN_FILE, CTL_EOL); 166 sysctl_createv(&clog, 0, NULL, NULL, 167 CTLFLAG_PERMANENT, 168 CTLTYPE_STRUCT, "file2", 169 SYSCTL_DESCR("System open file table"), 170 sysctl_kern_file2, 0, NULL, 0, 171 CTL_KERN, KERN_FILE2, CTL_EOL); 172 } 173 174 static bool 175 fd_isused(filedesc_t *fdp, unsigned fd) 176 { 177 u_int off = fd >> NDENTRYSHIFT; 178 179 KASSERT(fd < fdp->fd_dt->dt_nfiles); 180 181 return (fdp->fd_lomap[off] & (1 << (fd & NDENTRYMASK))) != 0; 182 } 183 184 /* 185 * Verify that the bitmaps match the descriptor table. 186 */ 187 static inline void 188 fd_checkmaps(filedesc_t *fdp) 189 { 190 #ifdef DEBUG 191 fdtab_t *dt; 192 u_int fd; 193 194 dt = fdp->fd_dt; 195 if (fdp->fd_refcnt == -1) { 196 /* 197 * fd_free tears down the table without maintaining its bitmap. 198 */ 199 return; 200 } 201 for (fd = 0; fd < dt->dt_nfiles; fd++) { 202 if (fd < NDFDFILE) { 203 KASSERT(dt->dt_ff[fd] == 204 (fdfile_t *)fdp->fd_dfdfile[fd]); 205 } 206 if (dt->dt_ff[fd] == NULL) { 207 KASSERT(!fd_isused(fdp, fd)); 208 } else if (dt->dt_ff[fd]->ff_file != NULL) { 209 KASSERT(fd_isused(fdp, fd)); 210 } 211 } 212 #endif 213 } 214 215 static int 216 fd_next_zero(filedesc_t *fdp, uint32_t *bitmap, int want, u_int bits) 217 { 218 int i, off, maxoff; 219 uint32_t sub; 220 221 KASSERT(mutex_owned(&fdp->fd_lock)); 222 223 fd_checkmaps(fdp); 224 225 if (want > bits) 226 return -1; 227 228 off = want >> NDENTRYSHIFT; 229 i = want & NDENTRYMASK; 230 if (i) { 231 sub = bitmap[off] | ((u_int)~0 >> (NDENTRIES - i)); 232 if (sub != ~0) 233 goto found; 234 off++; 235 } 236 237 maxoff = NDLOSLOTS(bits); 238 while (off < maxoff) { 239 if ((sub = bitmap[off]) != ~0) 240 goto found; 241 off++; 242 } 243 244 return -1; 245 246 found: 247 return (off << NDENTRYSHIFT) + ffs(~sub) - 1; 248 } 249 250 static int 251 fd_last_set(filedesc_t *fd, int last) 252 { 253 int off, i; 254 fdfile_t **ff = fd->fd_dt->dt_ff; 255 uint32_t *bitmap = fd->fd_lomap; 256 257 KASSERT(mutex_owned(&fd->fd_lock)); 258 259 fd_checkmaps(fd); 260 261 off = (last - 1) >> NDENTRYSHIFT; 262 263 while (off >= 0 && !bitmap[off]) 264 off--; 265 266 if (off < 0) 267 return -1; 268 269 i = ((off + 1) << NDENTRYSHIFT) - 1; 270 if (i >= last) 271 i = last - 1; 272 273 /* XXX should use bitmap */ 274 while (i > 0 && (ff[i] == NULL || !ff[i]->ff_allocated)) 275 i--; 276 277 return i; 278 } 279 280 static inline void 281 fd_used(filedesc_t *fdp, unsigned fd) 282 { 283 u_int off = fd >> NDENTRYSHIFT; 284 fdfile_t *ff; 285 286 ff = fdp->fd_dt->dt_ff[fd]; 287 288 KASSERT(mutex_owned(&fdp->fd_lock)); 289 KASSERT((fdp->fd_lomap[off] & (1 << (fd & NDENTRYMASK))) == 0); 290 KASSERT(ff != NULL); 291 KASSERT(ff->ff_file == NULL); 292 KASSERT(!ff->ff_allocated); 293 294 ff->ff_allocated = true; 295 fdp->fd_lomap[off] |= 1 << (fd & NDENTRYMASK); 296 if (__predict_false(fdp->fd_lomap[off] == ~0)) { 297 KASSERT((fdp->fd_himap[off >> NDENTRYSHIFT] & 298 (1 << (off & NDENTRYMASK))) == 0); 299 fdp->fd_himap[off >> NDENTRYSHIFT] |= 1 << (off & NDENTRYMASK); 300 } 301 302 if ((int)fd > fdp->fd_lastfile) { 303 fdp->fd_lastfile = fd; 304 } 305 306 fd_checkmaps(fdp); 307 } 308 309 static inline void 310 fd_unused(filedesc_t *fdp, unsigned fd) 311 { 312 u_int off = fd >> NDENTRYSHIFT; 313 fdfile_t *ff; 314 315 ff = fdp->fd_dt->dt_ff[fd]; 316 317 /* 318 * Don't assert the lock is held here, as we may be copying 319 * the table during exec() and it is not needed there. 320 * procfs and sysctl are locked out by proc::p_reflock. 321 * 322 * KASSERT(mutex_owned(&fdp->fd_lock)); 323 */ 324 KASSERT(ff != NULL); 325 KASSERT(ff->ff_file == NULL); 326 KASSERT(ff->ff_allocated); 327 328 if (fd < fdp->fd_freefile) { 329 fdp->fd_freefile = fd; 330 } 331 332 if (fdp->fd_lomap[off] == ~0) { 333 KASSERT((fdp->fd_himap[off >> NDENTRYSHIFT] & 334 (1 << (off & NDENTRYMASK))) != 0); 335 fdp->fd_himap[off >> NDENTRYSHIFT] &= 336 ~(1 << (off & NDENTRYMASK)); 337 } 338 KASSERT((fdp->fd_lomap[off] & (1 << (fd & NDENTRYMASK))) != 0); 339 fdp->fd_lomap[off] &= ~(1 << (fd & NDENTRYMASK)); 340 ff->ff_allocated = false; 341 342 KASSERT(fd <= fdp->fd_lastfile); 343 if (fd == fdp->fd_lastfile) { 344 fdp->fd_lastfile = fd_last_set(fdp, fd); 345 } 346 fd_checkmaps(fdp); 347 } 348 349 /* 350 * Look up the file structure corresponding to a file descriptor 351 * and return the file, holding a reference on the descriptor. 352 */ 353 file_t * 354 fd_getfile(unsigned fd) 355 { 356 filedesc_t *fdp; 357 fdfile_t *ff; 358 file_t *fp; 359 fdtab_t *dt; 360 361 /* 362 * Look up the fdfile structure representing this descriptor. 363 * We are doing this unlocked. See fd_tryexpand(). 364 */ 365 fdp = curlwp->l_fd; 366 dt = fdp->fd_dt; 367 if (__predict_false(fd >= dt->dt_nfiles)) { 368 return NULL; 369 } 370 ff = dt->dt_ff[fd]; 371 KASSERT(fd >= NDFDFILE || ff == (fdfile_t *)fdp->fd_dfdfile[fd]); 372 if (__predict_false(ff == NULL)) { 373 return NULL; 374 } 375 376 /* Now get a reference to the descriptor. */ 377 if (fdp->fd_refcnt == 1) { 378 /* 379 * Single threaded: don't need to worry about concurrent 380 * access (other than earlier calls to kqueue, which may 381 * hold a reference to the descriptor). 382 */ 383 ff->ff_refcnt++; 384 } else { 385 /* 386 * Multi threaded: issue a memory barrier to ensure that we 387 * acquire the file pointer _after_ adding a reference. If 388 * no memory barrier, we could fetch a stale pointer. 389 */ 390 atomic_inc_uint(&ff->ff_refcnt); 391 #ifndef __HAVE_ATOMIC_AS_MEMBAR 392 membar_enter(); 393 #endif 394 } 395 396 /* 397 * If the file is not open or is being closed then put the 398 * reference back. 399 */ 400 fp = ff->ff_file; 401 if (__predict_true(fp != NULL)) { 402 return fp; 403 } 404 fd_putfile(fd); 405 return NULL; 406 } 407 408 /* 409 * Release a reference to a file descriptor acquired with fd_getfile(). 410 */ 411 void 412 fd_putfile(unsigned fd) 413 { 414 filedesc_t *fdp; 415 fdfile_t *ff; 416 u_int u, v; 417 418 fdp = curlwp->l_fd; 419 ff = fdp->fd_dt->dt_ff[fd]; 420 421 KASSERT(fd < fdp->fd_dt->dt_nfiles); 422 KASSERT(ff != NULL); 423 KASSERT((ff->ff_refcnt & FR_MASK) > 0); 424 KASSERT(fd >= NDFDFILE || ff == (fdfile_t *)fdp->fd_dfdfile[fd]); 425 426 if (fdp->fd_refcnt == 1) { 427 /* 428 * Single threaded: don't need to worry about concurrent 429 * access (other than earlier calls to kqueue, which may 430 * hold a reference to the descriptor). 431 */ 432 if (__predict_false((ff->ff_refcnt & FR_CLOSING) != 0)) { 433 fd_close(fd); 434 return; 435 } 436 ff->ff_refcnt--; 437 return; 438 } 439 440 /* 441 * Ensure that any use of the file is complete and globally 442 * visible before dropping the final reference. If no membar, 443 * the current CPU could still access memory associated with 444 * the file after it has been freed or recycled by another 445 * CPU. 446 */ 447 #ifndef __HAVE_ATOMIC_AS_MEMBAR 448 membar_exit(); 449 #endif 450 451 /* 452 * Be optimistic and start out with the assumption that no other 453 * threads are trying to close the descriptor. If the CAS fails, 454 * we lost a race and/or it's being closed. 455 */ 456 for (u = ff->ff_refcnt & FR_MASK;; u = v) { 457 v = atomic_cas_uint(&ff->ff_refcnt, u, u - 1); 458 if (__predict_true(u == v)) { 459 return; 460 } 461 if (__predict_false((v & FR_CLOSING) != 0)) { 462 break; 463 } 464 } 465 466 /* Another thread is waiting to close the file: join it. */ 467 (void)fd_close(fd); 468 } 469 470 /* 471 * Convenience wrapper around fd_getfile() that returns reference 472 * to a vnode. 473 */ 474 int 475 fd_getvnode(unsigned fd, file_t **fpp) 476 { 477 vnode_t *vp; 478 file_t *fp; 479 480 fp = fd_getfile(fd); 481 if (__predict_false(fp == NULL)) { 482 return EBADF; 483 } 484 if (__predict_false(fp->f_type != DTYPE_VNODE)) { 485 fd_putfile(fd); 486 return EINVAL; 487 } 488 vp = fp->f_data; 489 if (__predict_false(vp->v_type == VBAD)) { 490 /* XXX Is this case really necessary? */ 491 fd_putfile(fd); 492 return EBADF; 493 } 494 *fpp = fp; 495 return 0; 496 } 497 498 /* 499 * Convenience wrapper around fd_getfile() that returns reference 500 * to a socket. 501 */ 502 int 503 fd_getsock1(unsigned fd, struct socket **sop, file_t **fp) 504 { 505 *fp = fd_getfile(fd); 506 if (__predict_false(*fp == NULL)) { 507 return EBADF; 508 } 509 if (__predict_false((*fp)->f_type != DTYPE_SOCKET)) { 510 fd_putfile(fd); 511 return ENOTSOCK; 512 } 513 *sop = (*fp)->f_data; 514 return 0; 515 } 516 517 int 518 fd_getsock(unsigned fd, struct socket **sop) 519 { 520 file_t *fp; 521 return fd_getsock1(fd, sop, &fp); 522 } 523 524 /* 525 * Look up the file structure corresponding to a file descriptor 526 * and return it with a reference held on the file, not the 527 * descriptor. 528 * 529 * This is heavyweight and only used when accessing descriptors 530 * from a foreign process. The caller must ensure that `p' does 531 * not exit or fork across this call. 532 * 533 * To release the file (not descriptor) reference, use closef(). 534 */ 535 file_t * 536 fd_getfile2(proc_t *p, unsigned fd) 537 { 538 filedesc_t *fdp; 539 fdfile_t *ff; 540 file_t *fp; 541 fdtab_t *dt; 542 543 fdp = p->p_fd; 544 mutex_enter(&fdp->fd_lock); 545 dt = fdp->fd_dt; 546 if (fd >= dt->dt_nfiles) { 547 mutex_exit(&fdp->fd_lock); 548 return NULL; 549 } 550 if ((ff = dt->dt_ff[fd]) == NULL) { 551 mutex_exit(&fdp->fd_lock); 552 return NULL; 553 } 554 if ((fp = ff->ff_file) == NULL) { 555 mutex_exit(&fdp->fd_lock); 556 return NULL; 557 } 558 mutex_enter(&fp->f_lock); 559 fp->f_count++; 560 mutex_exit(&fp->f_lock); 561 mutex_exit(&fdp->fd_lock); 562 563 return fp; 564 } 565 566 /* 567 * Internal form of close. Must be called with a reference to the 568 * descriptor, and will drop the reference. When all descriptor 569 * references are dropped, releases the descriptor slot and a single 570 * reference to the file structure. 571 */ 572 int 573 fd_close(unsigned fd) 574 { 575 struct flock lf; 576 filedesc_t *fdp; 577 fdfile_t *ff; 578 file_t *fp; 579 proc_t *p; 580 lwp_t *l; 581 u_int refcnt; 582 583 l = curlwp; 584 p = l->l_proc; 585 fdp = l->l_fd; 586 ff = fdp->fd_dt->dt_ff[fd]; 587 588 KASSERT(fd >= NDFDFILE || ff == (fdfile_t *)fdp->fd_dfdfile[fd]); 589 590 mutex_enter(&fdp->fd_lock); 591 KASSERT((ff->ff_refcnt & FR_MASK) > 0); 592 if (__predict_false(ff->ff_file == NULL)) { 593 /* 594 * Another user of the file is already closing, and is 595 * waiting for other users of the file to drain. Release 596 * our reference, and wake up the closer. 597 */ 598 atomic_dec_uint(&ff->ff_refcnt); 599 cv_broadcast(&ff->ff_closing); 600 mutex_exit(&fdp->fd_lock); 601 602 /* 603 * An application error, so pretend that the descriptor 604 * was already closed. We can't safely wait for it to 605 * be closed without potentially deadlocking. 606 */ 607 return (EBADF); 608 } 609 KASSERT((ff->ff_refcnt & FR_CLOSING) == 0); 610 611 /* 612 * There may be multiple users of this file within the process. 613 * Notify existing and new users that the file is closing. This 614 * will prevent them from adding additional uses to this file 615 * while we are closing it. 616 */ 617 fp = ff->ff_file; 618 ff->ff_file = NULL; 619 ff->ff_exclose = false; 620 621 /* 622 * We expect the caller to hold a descriptor reference - drop it. 623 * The reference count may increase beyond zero at this point due 624 * to an erroneous descriptor reference by an application, but 625 * fd_getfile() will notice that the file is being closed and drop 626 * the reference again. 627 */ 628 if (fdp->fd_refcnt == 1) { 629 /* Single threaded. */ 630 refcnt = --(ff->ff_refcnt); 631 } else { 632 /* Multi threaded. */ 633 #ifndef __HAVE_ATOMIC_AS_MEMBAR 634 membar_producer(); 635 #endif 636 refcnt = atomic_dec_uint_nv(&ff->ff_refcnt); 637 } 638 if (__predict_false(refcnt != 0)) { 639 /* 640 * Wait for other references to drain. This is typically 641 * an application error - the descriptor is being closed 642 * while still in use. 643 * (Or just a threaded application trying to unblock its 644 * thread that sleeps in (say) accept()). 645 */ 646 atomic_or_uint(&ff->ff_refcnt, FR_CLOSING); 647 648 /* 649 * Remove any knotes attached to the file. A knote 650 * attached to the descriptor can hold references on it. 651 */ 652 mutex_exit(&fdp->fd_lock); 653 if (!SLIST_EMPTY(&ff->ff_knlist)) { 654 knote_fdclose(fd); 655 } 656 657 /* 658 * Since the file system code doesn't know which fd 659 * each request came from (think dup()), we have to 660 * ask it to return ERESTART for any long-term blocks. 661 * The re-entry through read/write/etc will detect the 662 * closed fd and return EBAFD. 663 * Blocked partial writes may return a short length. 664 */ 665 (*fp->f_ops->fo_restart)(fp); 666 mutex_enter(&fdp->fd_lock); 667 668 /* 669 * We need to see the count drop to zero at least once, 670 * in order to ensure that all pre-existing references 671 * have been drained. New references past this point are 672 * of no interest. 673 * XXX (dsl) this may need to call fo_restart() after a 674 * timeout to guarantee that all the system calls exit. 675 */ 676 while ((ff->ff_refcnt & FR_MASK) != 0) { 677 cv_wait(&ff->ff_closing, &fdp->fd_lock); 678 } 679 atomic_and_uint(&ff->ff_refcnt, ~FR_CLOSING); 680 } else { 681 /* If no references, there must be no knotes. */ 682 KASSERT(SLIST_EMPTY(&ff->ff_knlist)); 683 } 684 685 /* 686 * POSIX record locking dictates that any close releases ALL 687 * locks owned by this process. This is handled by setting 688 * a flag in the unlock to free ONLY locks obeying POSIX 689 * semantics, and not to free BSD-style file locks. 690 * If the descriptor was in a message, POSIX-style locks 691 * aren't passed with the descriptor. 692 */ 693 if (__predict_false((p->p_flag & PK_ADVLOCK) != 0 && 694 fp->f_type == DTYPE_VNODE)) { 695 lf.l_whence = SEEK_SET; 696 lf.l_start = 0; 697 lf.l_len = 0; 698 lf.l_type = F_UNLCK; 699 mutex_exit(&fdp->fd_lock); 700 (void)VOP_ADVLOCK(fp->f_data, p, F_UNLCK, &lf, F_POSIX); 701 mutex_enter(&fdp->fd_lock); 702 } 703 704 /* Free descriptor slot. */ 705 fd_unused(fdp, fd); 706 mutex_exit(&fdp->fd_lock); 707 708 /* Now drop reference to the file itself. */ 709 return closef(fp); 710 } 711 712 /* 713 * Duplicate a file descriptor. 714 */ 715 int 716 fd_dup(file_t *fp, int minfd, int *newp, bool exclose) 717 { 718 proc_t *p = curproc; 719 int error; 720 721 while ((error = fd_alloc(p, minfd, newp)) != 0) { 722 if (error != ENOSPC) { 723 return error; 724 } 725 fd_tryexpand(p); 726 } 727 728 curlwp->l_fd->fd_dt->dt_ff[*newp]->ff_exclose = exclose; 729 fd_affix(p, fp, *newp); 730 return 0; 731 } 732 733 /* 734 * dup2 operation. 735 */ 736 int 737 fd_dup2(file_t *fp, unsigned new, int flags) 738 { 739 filedesc_t *fdp = curlwp->l_fd; 740 fdfile_t *ff; 741 fdtab_t *dt; 742 743 if (flags & ~(O_CLOEXEC|O_NONBLOCK)) 744 return EINVAL; 745 /* 746 * Ensure there are enough slots in the descriptor table, 747 * and allocate an fdfile_t up front in case we need it. 748 */ 749 while (new >= fdp->fd_dt->dt_nfiles) { 750 fd_tryexpand(curproc); 751 } 752 ff = pool_cache_get(fdfile_cache, PR_WAITOK); 753 754 /* 755 * If there is already a file open, close it. If the file is 756 * half open, wait for it to be constructed before closing it. 757 * XXX Potential for deadlock here? 758 */ 759 mutex_enter(&fdp->fd_lock); 760 while (fd_isused(fdp, new)) { 761 mutex_exit(&fdp->fd_lock); 762 if (fd_getfile(new) != NULL) { 763 (void)fd_close(new); 764 } else { 765 /* 766 * Crummy, but unlikely to happen. 767 * Can occur if we interrupt another 768 * thread while it is opening a file. 769 */ 770 kpause("dup2", false, 1, NULL); 771 } 772 mutex_enter(&fdp->fd_lock); 773 } 774 dt = fdp->fd_dt; 775 if (dt->dt_ff[new] == NULL) { 776 KASSERT(new >= NDFDFILE); 777 dt->dt_ff[new] = ff; 778 ff = NULL; 779 } 780 fd_used(fdp, new); 781 mutex_exit(&fdp->fd_lock); 782 783 dt->dt_ff[new]->ff_exclose = (flags & O_CLOEXEC) != 0; 784 fp->f_flag |= flags & FNONBLOCK; 785 /* Slot is now allocated. Insert copy of the file. */ 786 fd_affix(curproc, fp, new); 787 if (ff != NULL) { 788 pool_cache_put(fdfile_cache, ff); 789 } 790 return 0; 791 } 792 793 /* 794 * Drop reference to a file structure. 795 */ 796 int 797 closef(file_t *fp) 798 { 799 struct flock lf; 800 int error; 801 802 /* 803 * Drop reference. If referenced elsewhere it's still open 804 * and we have nothing more to do. 805 */ 806 mutex_enter(&fp->f_lock); 807 KASSERT(fp->f_count > 0); 808 if (--fp->f_count > 0) { 809 mutex_exit(&fp->f_lock); 810 return 0; 811 } 812 KASSERT(fp->f_count == 0); 813 mutex_exit(&fp->f_lock); 814 815 /* We held the last reference - release locks, close and free. */ 816 if ((fp->f_flag & FHASLOCK) && fp->f_type == DTYPE_VNODE) { 817 lf.l_whence = SEEK_SET; 818 lf.l_start = 0; 819 lf.l_len = 0; 820 lf.l_type = F_UNLCK; 821 (void)VOP_ADVLOCK(fp->f_data, fp, F_UNLCK, &lf, F_FLOCK); 822 } 823 if (fp->f_ops != NULL) { 824 error = (*fp->f_ops->fo_close)(fp); 825 } else { 826 error = 0; 827 } 828 KASSERT(fp->f_count == 0); 829 KASSERT(fp->f_cred != NULL); 830 pool_cache_put(file_cache, fp); 831 832 return error; 833 } 834 835 /* 836 * Allocate a file descriptor for the process. 837 */ 838 int 839 fd_alloc(proc_t *p, int want, int *result) 840 { 841 filedesc_t *fdp = p->p_fd; 842 int i, lim, last, error; 843 u_int off, new; 844 fdtab_t *dt; 845 846 KASSERT(p == curproc || p == &proc0); 847 848 /* 849 * Search for a free descriptor starting at the higher 850 * of want or fd_freefile. 851 */ 852 mutex_enter(&fdp->fd_lock); 853 fd_checkmaps(fdp); 854 dt = fdp->fd_dt; 855 KASSERT(dt->dt_ff[0] == (fdfile_t *)fdp->fd_dfdfile[0]); 856 lim = min((int)p->p_rlimit[RLIMIT_NOFILE].rlim_cur, maxfiles); 857 last = min(dt->dt_nfiles, lim); 858 for (;;) { 859 if ((i = want) < fdp->fd_freefile) 860 i = fdp->fd_freefile; 861 off = i >> NDENTRYSHIFT; 862 new = fd_next_zero(fdp, fdp->fd_himap, off, 863 (last + NDENTRIES - 1) >> NDENTRYSHIFT); 864 if (new == -1) 865 break; 866 i = fd_next_zero(fdp, &fdp->fd_lomap[new], 867 new > off ? 0 : i & NDENTRYMASK, NDENTRIES); 868 if (i == -1) { 869 /* 870 * Free file descriptor in this block was 871 * below want, try again with higher want. 872 */ 873 want = (new + 1) << NDENTRYSHIFT; 874 continue; 875 } 876 i += (new << NDENTRYSHIFT); 877 if (i >= last) { 878 break; 879 } 880 if (dt->dt_ff[i] == NULL) { 881 KASSERT(i >= NDFDFILE); 882 dt->dt_ff[i] = pool_cache_get(fdfile_cache, PR_WAITOK); 883 } 884 KASSERT(dt->dt_ff[i]->ff_file == NULL); 885 fd_used(fdp, i); 886 if (want <= fdp->fd_freefile) { 887 fdp->fd_freefile = i; 888 } 889 *result = i; 890 KASSERT(i >= NDFDFILE || 891 dt->dt_ff[i] == (fdfile_t *)fdp->fd_dfdfile[i]); 892 fd_checkmaps(fdp); 893 mutex_exit(&fdp->fd_lock); 894 return 0; 895 } 896 897 /* No space in current array. Let the caller expand and retry. */ 898 error = (dt->dt_nfiles >= lim) ? EMFILE : ENOSPC; 899 mutex_exit(&fdp->fd_lock); 900 return error; 901 } 902 903 /* 904 * Allocate memory for a descriptor table. 905 */ 906 static fdtab_t * 907 fd_dtab_alloc(int n) 908 { 909 fdtab_t *dt; 910 size_t sz; 911 912 KASSERT(n > NDFILE); 913 914 sz = sizeof(*dt) + (n - NDFILE) * sizeof(dt->dt_ff[0]); 915 dt = kmem_alloc(sz, KM_SLEEP); 916 #ifdef DIAGNOSTIC 917 memset(dt, 0xff, sz); 918 #endif 919 dt->dt_nfiles = n; 920 dt->dt_link = NULL; 921 return dt; 922 } 923 924 /* 925 * Free a descriptor table, and all tables linked for deferred free. 926 */ 927 static void 928 fd_dtab_free(fdtab_t *dt) 929 { 930 fdtab_t *next; 931 size_t sz; 932 933 do { 934 next = dt->dt_link; 935 KASSERT(dt->dt_nfiles > NDFILE); 936 sz = sizeof(*dt) + 937 (dt->dt_nfiles - NDFILE) * sizeof(dt->dt_ff[0]); 938 #ifdef DIAGNOSTIC 939 memset(dt, 0xff, sz); 940 #endif 941 kmem_free(dt, sz); 942 dt = next; 943 } while (dt != NULL); 944 } 945 946 /* 947 * Allocate descriptor bitmap. 948 */ 949 static void 950 fd_map_alloc(int n, uint32_t **lo, uint32_t **hi) 951 { 952 uint8_t *ptr; 953 size_t szlo, szhi; 954 955 KASSERT(n > NDENTRIES); 956 957 szlo = NDLOSLOTS(n) * sizeof(uint32_t); 958 szhi = NDHISLOTS(n) * sizeof(uint32_t); 959 ptr = kmem_alloc(szlo + szhi, KM_SLEEP); 960 *lo = (uint32_t *)ptr; 961 *hi = (uint32_t *)(ptr + szlo); 962 } 963 964 /* 965 * Free descriptor bitmap. 966 */ 967 static void 968 fd_map_free(int n, uint32_t *lo, uint32_t *hi) 969 { 970 size_t szlo, szhi; 971 972 KASSERT(n > NDENTRIES); 973 974 szlo = NDLOSLOTS(n) * sizeof(uint32_t); 975 szhi = NDHISLOTS(n) * sizeof(uint32_t); 976 KASSERT(hi == (uint32_t *)((uint8_t *)lo + szlo)); 977 kmem_free(lo, szlo + szhi); 978 } 979 980 /* 981 * Expand a process' descriptor table. 982 */ 983 void 984 fd_tryexpand(proc_t *p) 985 { 986 filedesc_t *fdp; 987 int i, numfiles, oldnfiles; 988 fdtab_t *newdt, *dt; 989 uint32_t *newhimap, *newlomap; 990 991 KASSERT(p == curproc || p == &proc0); 992 993 fdp = p->p_fd; 994 newhimap = NULL; 995 newlomap = NULL; 996 oldnfiles = fdp->fd_dt->dt_nfiles; 997 998 if (oldnfiles < NDEXTENT) 999 numfiles = NDEXTENT; 1000 else 1001 numfiles = 2 * oldnfiles; 1002 1003 newdt = fd_dtab_alloc(numfiles); 1004 if (NDHISLOTS(numfiles) > NDHISLOTS(oldnfiles)) { 1005 fd_map_alloc(numfiles, &newlomap, &newhimap); 1006 } 1007 1008 mutex_enter(&fdp->fd_lock); 1009 dt = fdp->fd_dt; 1010 KASSERT(dt->dt_ff[0] == (fdfile_t *)fdp->fd_dfdfile[0]); 1011 if (dt->dt_nfiles != oldnfiles) { 1012 /* fdp changed; caller must retry */ 1013 mutex_exit(&fdp->fd_lock); 1014 fd_dtab_free(newdt); 1015 if (NDHISLOTS(numfiles) > NDHISLOTS(oldnfiles)) { 1016 fd_map_free(numfiles, newlomap, newhimap); 1017 } 1018 return; 1019 } 1020 1021 /* Copy the existing descriptor table and zero the new portion. */ 1022 i = sizeof(fdfile_t *) * oldnfiles; 1023 memcpy(newdt->dt_ff, dt->dt_ff, i); 1024 memset((uint8_t *)newdt->dt_ff + i, 0, 1025 numfiles * sizeof(fdfile_t *) - i); 1026 1027 /* 1028 * Link old descriptor array into list to be discarded. We defer 1029 * freeing until the last reference to the descriptor table goes 1030 * away (usually process exit). This allows us to do lockless 1031 * lookups in fd_getfile(). 1032 */ 1033 if (oldnfiles > NDFILE) { 1034 if (fdp->fd_refcnt > 1) { 1035 newdt->dt_link = dt; 1036 } else { 1037 fd_dtab_free(dt); 1038 } 1039 } 1040 1041 if (NDHISLOTS(numfiles) > NDHISLOTS(oldnfiles)) { 1042 i = NDHISLOTS(oldnfiles) * sizeof(uint32_t); 1043 memcpy(newhimap, fdp->fd_himap, i); 1044 memset((uint8_t *)newhimap + i, 0, 1045 NDHISLOTS(numfiles) * sizeof(uint32_t) - i); 1046 1047 i = NDLOSLOTS(oldnfiles) * sizeof(uint32_t); 1048 memcpy(newlomap, fdp->fd_lomap, i); 1049 memset((uint8_t *)newlomap + i, 0, 1050 NDLOSLOTS(numfiles) * sizeof(uint32_t) - i); 1051 1052 if (NDHISLOTS(oldnfiles) > NDHISLOTS(NDFILE)) { 1053 fd_map_free(oldnfiles, fdp->fd_lomap, fdp->fd_himap); 1054 } 1055 fdp->fd_himap = newhimap; 1056 fdp->fd_lomap = newlomap; 1057 } 1058 1059 /* 1060 * All other modifications must become globally visible before 1061 * the change to fd_dt. See fd_getfile(). 1062 */ 1063 membar_producer(); 1064 fdp->fd_dt = newdt; 1065 KASSERT(newdt->dt_ff[0] == (fdfile_t *)fdp->fd_dfdfile[0]); 1066 fd_checkmaps(fdp); 1067 mutex_exit(&fdp->fd_lock); 1068 } 1069 1070 /* 1071 * Create a new open file structure and allocate a file descriptor 1072 * for the current process. 1073 */ 1074 int 1075 fd_allocfile(file_t **resultfp, int *resultfd) 1076 { 1077 proc_t *p = curproc; 1078 kauth_cred_t cred; 1079 file_t *fp; 1080 int error; 1081 1082 while ((error = fd_alloc(p, 0, resultfd)) != 0) { 1083 if (error != ENOSPC) { 1084 return error; 1085 } 1086 fd_tryexpand(p); 1087 } 1088 1089 fp = pool_cache_get(file_cache, PR_WAITOK); 1090 if (fp == NULL) { 1091 fd_abort(p, NULL, *resultfd); 1092 return ENFILE; 1093 } 1094 KASSERT(fp->f_count == 0); 1095 KASSERT(fp->f_msgcount == 0); 1096 KASSERT(fp->f_unpcount == 0); 1097 1098 /* Replace cached credentials if not what we need. */ 1099 cred = curlwp->l_cred; 1100 if (__predict_false(cred != fp->f_cred)) { 1101 kauth_cred_free(fp->f_cred); 1102 kauth_cred_hold(cred); 1103 fp->f_cred = cred; 1104 } 1105 1106 /* 1107 * Don't allow recycled files to be scanned. 1108 * See uipc_usrreq.c. 1109 */ 1110 if (__predict_false((fp->f_flag & FSCAN) != 0)) { 1111 mutex_enter(&fp->f_lock); 1112 atomic_and_uint(&fp->f_flag, ~FSCAN); 1113 mutex_exit(&fp->f_lock); 1114 } 1115 1116 fp->f_advice = 0; 1117 fp->f_offset = 0; 1118 *resultfp = fp; 1119 1120 return 0; 1121 } 1122 1123 /* 1124 * Successful creation of a new descriptor: make visible to the process. 1125 */ 1126 void 1127 fd_affix(proc_t *p, file_t *fp, unsigned fd) 1128 { 1129 fdfile_t *ff; 1130 filedesc_t *fdp; 1131 1132 KASSERT(p == curproc || p == &proc0); 1133 1134 /* Add a reference to the file structure. */ 1135 mutex_enter(&fp->f_lock); 1136 fp->f_count++; 1137 mutex_exit(&fp->f_lock); 1138 1139 /* 1140 * Insert the new file into the descriptor slot. 1141 * 1142 * The memory barriers provided by lock activity in this routine 1143 * ensure that any updates to the file structure become globally 1144 * visible before the file becomes visible to other LWPs in the 1145 * current process. 1146 */ 1147 fdp = p->p_fd; 1148 ff = fdp->fd_dt->dt_ff[fd]; 1149 1150 KASSERT(ff != NULL); 1151 KASSERT(ff->ff_file == NULL); 1152 KASSERT(ff->ff_allocated); 1153 KASSERT(fd_isused(fdp, fd)); 1154 KASSERT(fd >= NDFDFILE || ff == (fdfile_t *)fdp->fd_dfdfile[fd]); 1155 1156 /* No need to lock in order to make file initially visible. */ 1157 ff->ff_file = fp; 1158 } 1159 1160 /* 1161 * Abort creation of a new descriptor: free descriptor slot and file. 1162 */ 1163 void 1164 fd_abort(proc_t *p, file_t *fp, unsigned fd) 1165 { 1166 filedesc_t *fdp; 1167 fdfile_t *ff; 1168 1169 KASSERT(p == curproc || p == &proc0); 1170 1171 fdp = p->p_fd; 1172 ff = fdp->fd_dt->dt_ff[fd]; 1173 1174 KASSERT(fd >= NDFDFILE || ff == (fdfile_t *)fdp->fd_dfdfile[fd]); 1175 1176 mutex_enter(&fdp->fd_lock); 1177 KASSERT(fd_isused(fdp, fd)); 1178 fd_unused(fdp, fd); 1179 mutex_exit(&fdp->fd_lock); 1180 1181 if (fp != NULL) { 1182 KASSERT(fp->f_count == 0); 1183 KASSERT(fp->f_cred != NULL); 1184 pool_cache_put(file_cache, fp); 1185 } 1186 } 1187 1188 static int 1189 file_ctor(void *arg, void *obj, int flags) 1190 { 1191 file_t *fp = obj; 1192 1193 memset(fp, 0, sizeof(*fp)); 1194 1195 mutex_enter(&filelist_lock); 1196 if (__predict_false(nfiles >= maxfiles)) { 1197 mutex_exit(&filelist_lock); 1198 tablefull("file", "increase kern.maxfiles or MAXFILES"); 1199 return ENFILE; 1200 } 1201 nfiles++; 1202 LIST_INSERT_HEAD(&filehead, fp, f_list); 1203 mutex_init(&fp->f_lock, MUTEX_DEFAULT, IPL_NONE); 1204 fp->f_cred = curlwp->l_cred; 1205 kauth_cred_hold(fp->f_cred); 1206 mutex_exit(&filelist_lock); 1207 1208 return 0; 1209 } 1210 1211 static void 1212 file_dtor(void *arg, void *obj) 1213 { 1214 file_t *fp = obj; 1215 1216 mutex_enter(&filelist_lock); 1217 nfiles--; 1218 LIST_REMOVE(fp, f_list); 1219 mutex_exit(&filelist_lock); 1220 1221 kauth_cred_free(fp->f_cred); 1222 mutex_destroy(&fp->f_lock); 1223 } 1224 1225 static int 1226 fdfile_ctor(void *arg, void *obj, int flags) 1227 { 1228 fdfile_t *ff = obj; 1229 1230 memset(ff, 0, sizeof(*ff)); 1231 cv_init(&ff->ff_closing, "fdclose"); 1232 1233 return 0; 1234 } 1235 1236 static void 1237 fdfile_dtor(void *arg, void *obj) 1238 { 1239 fdfile_t *ff = obj; 1240 1241 cv_destroy(&ff->ff_closing); 1242 } 1243 1244 file_t * 1245 fgetdummy(void) 1246 { 1247 file_t *fp; 1248 1249 fp = kmem_zalloc(sizeof(*fp), KM_SLEEP); 1250 if (fp != NULL) { 1251 mutex_init(&fp->f_lock, MUTEX_DEFAULT, IPL_NONE); 1252 } 1253 return fp; 1254 } 1255 1256 void 1257 fputdummy(file_t *fp) 1258 { 1259 1260 mutex_destroy(&fp->f_lock); 1261 kmem_free(fp, sizeof(*fp)); 1262 } 1263 1264 /* 1265 * Create an initial filedesc structure. 1266 */ 1267 filedesc_t * 1268 fd_init(filedesc_t *fdp) 1269 { 1270 #ifdef DIAGNOSTIC 1271 unsigned fd; 1272 #endif 1273 1274 if (__predict_true(fdp == NULL)) { 1275 fdp = pool_cache_get(filedesc_cache, PR_WAITOK); 1276 } else { 1277 KASSERT(fdp == &filedesc0); 1278 filedesc_ctor(NULL, fdp, PR_WAITOK); 1279 } 1280 1281 #ifdef DIAGNOSTIC 1282 KASSERT(fdp->fd_lastfile == -1); 1283 KASSERT(fdp->fd_lastkqfile == -1); 1284 KASSERT(fdp->fd_knhash == NULL); 1285 KASSERT(fdp->fd_freefile == 0); 1286 KASSERT(fdp->fd_exclose == false); 1287 KASSERT(fdp->fd_dt == &fdp->fd_dtbuiltin); 1288 KASSERT(fdp->fd_dtbuiltin.dt_nfiles == NDFILE); 1289 for (fd = 0; fd < NDFDFILE; fd++) { 1290 KASSERT(fdp->fd_dtbuiltin.dt_ff[fd] == 1291 (fdfile_t *)fdp->fd_dfdfile[fd]); 1292 } 1293 for (fd = NDFDFILE; fd < NDFILE; fd++) { 1294 KASSERT(fdp->fd_dtbuiltin.dt_ff[fd] == NULL); 1295 } 1296 KASSERT(fdp->fd_himap == fdp->fd_dhimap); 1297 KASSERT(fdp->fd_lomap == fdp->fd_dlomap); 1298 #endif /* DIAGNOSTIC */ 1299 1300 fdp->fd_refcnt = 1; 1301 fd_checkmaps(fdp); 1302 1303 return fdp; 1304 } 1305 1306 /* 1307 * Initialize a file descriptor table. 1308 */ 1309 static int 1310 filedesc_ctor(void *arg, void *obj, int flag) 1311 { 1312 filedesc_t *fdp = obj; 1313 fdfile_t **ffp; 1314 int i; 1315 1316 memset(fdp, 0, sizeof(*fdp)); 1317 mutex_init(&fdp->fd_lock, MUTEX_DEFAULT, IPL_NONE); 1318 fdp->fd_lastfile = -1; 1319 fdp->fd_lastkqfile = -1; 1320 fdp->fd_dt = &fdp->fd_dtbuiltin; 1321 fdp->fd_dtbuiltin.dt_nfiles = NDFILE; 1322 fdp->fd_himap = fdp->fd_dhimap; 1323 fdp->fd_lomap = fdp->fd_dlomap; 1324 1325 CTASSERT(sizeof(fdp->fd_dfdfile[0]) >= sizeof(fdfile_t)); 1326 for (i = 0, ffp = fdp->fd_dt->dt_ff; i < NDFDFILE; i++, ffp++) { 1327 *ffp = (fdfile_t *)fdp->fd_dfdfile[i]; 1328 (void)fdfile_ctor(NULL, fdp->fd_dfdfile[i], PR_WAITOK); 1329 } 1330 1331 return 0; 1332 } 1333 1334 static void 1335 filedesc_dtor(void *arg, void *obj) 1336 { 1337 filedesc_t *fdp = obj; 1338 int i; 1339 1340 for (i = 0; i < NDFDFILE; i++) { 1341 fdfile_dtor(NULL, fdp->fd_dfdfile[i]); 1342 } 1343 1344 mutex_destroy(&fdp->fd_lock); 1345 } 1346 1347 /* 1348 * Make p share curproc's filedesc structure. 1349 */ 1350 void 1351 fd_share(struct proc *p) 1352 { 1353 filedesc_t *fdp; 1354 1355 fdp = curlwp->l_fd; 1356 p->p_fd = fdp; 1357 atomic_inc_uint(&fdp->fd_refcnt); 1358 } 1359 1360 /* 1361 * Acquire a hold on a filedesc structure. 1362 */ 1363 void 1364 fd_hold(lwp_t *l) 1365 { 1366 filedesc_t *fdp = l->l_fd; 1367 1368 atomic_inc_uint(&fdp->fd_refcnt); 1369 } 1370 1371 /* 1372 * Copy a filedesc structure. 1373 */ 1374 filedesc_t * 1375 fd_copy(void) 1376 { 1377 filedesc_t *newfdp, *fdp; 1378 fdfile_t *ff, **ffp, **nffp, *ff2; 1379 int i, j, numfiles, lastfile, newlast; 1380 file_t *fp; 1381 fdtab_t *newdt; 1382 1383 fdp = curproc->p_fd; 1384 newfdp = pool_cache_get(filedesc_cache, PR_WAITOK); 1385 newfdp->fd_refcnt = 1; 1386 1387 #ifdef DIAGNOSTIC 1388 KASSERT(newfdp->fd_lastfile == -1); 1389 KASSERT(newfdp->fd_lastkqfile == -1); 1390 KASSERT(newfdp->fd_knhash == NULL); 1391 KASSERT(newfdp->fd_freefile == 0); 1392 KASSERT(newfdp->fd_exclose == false); 1393 KASSERT(newfdp->fd_dt == &newfdp->fd_dtbuiltin); 1394 KASSERT(newfdp->fd_dtbuiltin.dt_nfiles == NDFILE); 1395 for (i = 0; i < NDFDFILE; i++) { 1396 KASSERT(newfdp->fd_dtbuiltin.dt_ff[i] == 1397 (fdfile_t *)&newfdp->fd_dfdfile[i]); 1398 } 1399 for (i = NDFDFILE; i < NDFILE; i++) { 1400 KASSERT(newfdp->fd_dtbuiltin.dt_ff[i] == NULL); 1401 } 1402 #endif /* DIAGNOSTIC */ 1403 1404 mutex_enter(&fdp->fd_lock); 1405 fd_checkmaps(fdp); 1406 numfiles = fdp->fd_dt->dt_nfiles; 1407 lastfile = fdp->fd_lastfile; 1408 1409 /* 1410 * If the number of open files fits in the internal arrays 1411 * of the open file structure, use them, otherwise allocate 1412 * additional memory for the number of descriptors currently 1413 * in use. 1414 */ 1415 if (lastfile < NDFILE) { 1416 i = NDFILE; 1417 newdt = newfdp->fd_dt; 1418 KASSERT(newfdp->fd_dt == &newfdp->fd_dtbuiltin); 1419 } else { 1420 /* 1421 * Compute the smallest multiple of NDEXTENT needed 1422 * for the file descriptors currently in use, 1423 * allowing the table to shrink. 1424 */ 1425 i = numfiles; 1426 while (i >= 2 * NDEXTENT && i > lastfile * 2) { 1427 i /= 2; 1428 } 1429 KASSERT(i > NDFILE); 1430 newdt = fd_dtab_alloc(i); 1431 newfdp->fd_dt = newdt; 1432 memcpy(newdt->dt_ff, newfdp->fd_dtbuiltin.dt_ff, 1433 NDFDFILE * sizeof(fdfile_t **)); 1434 memset(newdt->dt_ff + NDFDFILE, 0, 1435 (i - NDFDFILE) * sizeof(fdfile_t **)); 1436 } 1437 if (NDHISLOTS(i) <= NDHISLOTS(NDFILE)) { 1438 newfdp->fd_himap = newfdp->fd_dhimap; 1439 newfdp->fd_lomap = newfdp->fd_dlomap; 1440 } else { 1441 fd_map_alloc(i, &newfdp->fd_lomap, &newfdp->fd_himap); 1442 KASSERT(i >= NDENTRIES * NDENTRIES); 1443 memset(newfdp->fd_himap, 0, NDHISLOTS(i)*sizeof(uint32_t)); 1444 memset(newfdp->fd_lomap, 0, NDLOSLOTS(i)*sizeof(uint32_t)); 1445 } 1446 newfdp->fd_freefile = fdp->fd_freefile; 1447 newfdp->fd_exclose = fdp->fd_exclose; 1448 1449 ffp = fdp->fd_dt->dt_ff; 1450 nffp = newdt->dt_ff; 1451 newlast = -1; 1452 for (i = 0; i <= (int)lastfile; i++, ffp++, nffp++) { 1453 KASSERT(i >= NDFDFILE || 1454 *nffp == (fdfile_t *)newfdp->fd_dfdfile[i]); 1455 ff = *ffp; 1456 if (ff == NULL || (fp = ff->ff_file) == NULL) { 1457 /* Descriptor unused, or descriptor half open. */ 1458 KASSERT(!fd_isused(newfdp, i)); 1459 continue; 1460 } 1461 if (__predict_false(fp->f_type == DTYPE_KQUEUE)) { 1462 /* kqueue descriptors cannot be copied. */ 1463 if (i < newfdp->fd_freefile) { 1464 newfdp->fd_freefile = i; 1465 } 1466 continue; 1467 } 1468 /* It's active: add a reference to the file. */ 1469 mutex_enter(&fp->f_lock); 1470 fp->f_count++; 1471 mutex_exit(&fp->f_lock); 1472 1473 /* Allocate an fdfile_t to represent it. */ 1474 if (i >= NDFDFILE) { 1475 ff2 = pool_cache_get(fdfile_cache, PR_WAITOK); 1476 *nffp = ff2; 1477 } else { 1478 ff2 = newdt->dt_ff[i]; 1479 } 1480 ff2->ff_file = fp; 1481 ff2->ff_exclose = ff->ff_exclose; 1482 ff2->ff_allocated = true; 1483 1484 /* Fix up bitmaps. */ 1485 j = i >> NDENTRYSHIFT; 1486 KASSERT((newfdp->fd_lomap[j] & (1 << (i & NDENTRYMASK))) == 0); 1487 newfdp->fd_lomap[j] |= 1 << (i & NDENTRYMASK); 1488 if (__predict_false(newfdp->fd_lomap[j] == ~0)) { 1489 KASSERT((newfdp->fd_himap[j >> NDENTRYSHIFT] & 1490 (1 << (j & NDENTRYMASK))) == 0); 1491 newfdp->fd_himap[j >> NDENTRYSHIFT] |= 1492 1 << (j & NDENTRYMASK); 1493 } 1494 newlast = i; 1495 } 1496 KASSERT(newdt->dt_ff[0] == (fdfile_t *)newfdp->fd_dfdfile[0]); 1497 newfdp->fd_lastfile = newlast; 1498 fd_checkmaps(newfdp); 1499 mutex_exit(&fdp->fd_lock); 1500 1501 return newfdp; 1502 } 1503 1504 /* 1505 * Release a filedesc structure. 1506 */ 1507 void 1508 fd_free(void) 1509 { 1510 fdfile_t *ff; 1511 file_t *fp; 1512 int fd, nf; 1513 fdtab_t *dt; 1514 lwp_t * const l = curlwp; 1515 filedesc_t * const fdp = l->l_fd; 1516 const bool noadvlock = (l->l_proc->p_flag & PK_ADVLOCK) == 0; 1517 1518 KASSERT(fdp->fd_dt->dt_ff[0] == (fdfile_t *)fdp->fd_dfdfile[0]); 1519 KASSERT(fdp->fd_dtbuiltin.dt_nfiles == NDFILE); 1520 KASSERT(fdp->fd_dtbuiltin.dt_link == NULL); 1521 1522 #ifndef __HAVE_ATOMIC_AS_MEMBAR 1523 membar_exit(); 1524 #endif 1525 if (atomic_dec_uint_nv(&fdp->fd_refcnt) > 0) 1526 return; 1527 1528 /* 1529 * Close any files that the process holds open. 1530 */ 1531 dt = fdp->fd_dt; 1532 fd_checkmaps(fdp); 1533 #ifdef DEBUG 1534 fdp->fd_refcnt = -1; /* see fd_checkmaps */ 1535 #endif 1536 for (fd = 0, nf = dt->dt_nfiles; fd < nf; fd++) { 1537 ff = dt->dt_ff[fd]; 1538 KASSERT(fd >= NDFDFILE || 1539 ff == (fdfile_t *)fdp->fd_dfdfile[fd]); 1540 if (ff == NULL) 1541 continue; 1542 if ((fp = ff->ff_file) != NULL) { 1543 /* 1544 * Must use fd_close() here if there is 1545 * a reference from kqueue or we might have posix 1546 * advisory locks. 1547 */ 1548 if (__predict_true(ff->ff_refcnt == 0) && 1549 (noadvlock || fp->f_type != DTYPE_VNODE)) { 1550 ff->ff_file = NULL; 1551 ff->ff_exclose = false; 1552 ff->ff_allocated = false; 1553 closef(fp); 1554 } else { 1555 ff->ff_refcnt++; 1556 fd_close(fd); 1557 } 1558 } 1559 KASSERT(ff->ff_refcnt == 0); 1560 KASSERT(ff->ff_file == NULL); 1561 KASSERT(!ff->ff_exclose); 1562 KASSERT(!ff->ff_allocated); 1563 if (fd >= NDFDFILE) { 1564 pool_cache_put(fdfile_cache, ff); 1565 dt->dt_ff[fd] = NULL; 1566 } 1567 } 1568 1569 /* 1570 * Clean out the descriptor table for the next user and return 1571 * to the cache. 1572 */ 1573 if (__predict_false(dt != &fdp->fd_dtbuiltin)) { 1574 fd_dtab_free(fdp->fd_dt); 1575 /* Otherwise, done above. */ 1576 memset(&fdp->fd_dtbuiltin.dt_ff[NDFDFILE], 0, 1577 (NDFILE - NDFDFILE) * sizeof(fdp->fd_dtbuiltin.dt_ff[0])); 1578 fdp->fd_dt = &fdp->fd_dtbuiltin; 1579 } 1580 if (__predict_false(NDHISLOTS(nf) > NDHISLOTS(NDFILE))) { 1581 KASSERT(fdp->fd_himap != fdp->fd_dhimap); 1582 KASSERT(fdp->fd_lomap != fdp->fd_dlomap); 1583 fd_map_free(nf, fdp->fd_lomap, fdp->fd_himap); 1584 } 1585 if (__predict_false(fdp->fd_knhash != NULL)) { 1586 hashdone(fdp->fd_knhash, HASH_LIST, fdp->fd_knhashmask); 1587 fdp->fd_knhash = NULL; 1588 fdp->fd_knhashmask = 0; 1589 } else { 1590 KASSERT(fdp->fd_knhashmask == 0); 1591 } 1592 fdp->fd_dt = &fdp->fd_dtbuiltin; 1593 fdp->fd_lastkqfile = -1; 1594 fdp->fd_lastfile = -1; 1595 fdp->fd_freefile = 0; 1596 fdp->fd_exclose = false; 1597 memset(&fdp->fd_startzero, 0, sizeof(*fdp) - 1598 offsetof(filedesc_t, fd_startzero)); 1599 fdp->fd_himap = fdp->fd_dhimap; 1600 fdp->fd_lomap = fdp->fd_dlomap; 1601 KASSERT(fdp->fd_dtbuiltin.dt_nfiles == NDFILE); 1602 KASSERT(fdp->fd_dtbuiltin.dt_link == NULL); 1603 KASSERT(fdp->fd_dt == &fdp->fd_dtbuiltin); 1604 #ifdef DEBUG 1605 fdp->fd_refcnt = 0; /* see fd_checkmaps */ 1606 #endif 1607 fd_checkmaps(fdp); 1608 pool_cache_put(filedesc_cache, fdp); 1609 } 1610 1611 /* 1612 * File Descriptor pseudo-device driver (/dev/fd/). 1613 * 1614 * Opening minor device N dup()s the file (if any) connected to file 1615 * descriptor N belonging to the calling process. Note that this driver 1616 * consists of only the ``open()'' routine, because all subsequent 1617 * references to this file will be direct to the other driver. 1618 */ 1619 static int 1620 filedescopen(dev_t dev, int mode, int type, lwp_t *l) 1621 { 1622 1623 /* 1624 * XXX Kludge: set dupfd to contain the value of the 1625 * the file descriptor being sought for duplication. The error 1626 * return ensures that the vnode for this device will be released 1627 * by vn_open. Open will detect this special error and take the 1628 * actions in fd_dupopen below. Other callers of vn_open or VOP_OPEN 1629 * will simply report the error. 1630 */ 1631 l->l_dupfd = minor(dev); /* XXX */ 1632 return EDUPFD; 1633 } 1634 1635 /* 1636 * Duplicate the specified descriptor to a free descriptor. 1637 */ 1638 int 1639 fd_dupopen(int old, int *new, int mode, int error) 1640 { 1641 filedesc_t *fdp; 1642 fdfile_t *ff; 1643 file_t *fp; 1644 fdtab_t *dt; 1645 1646 if ((fp = fd_getfile(old)) == NULL) { 1647 return EBADF; 1648 } 1649 fdp = curlwp->l_fd; 1650 dt = fdp->fd_dt; 1651 ff = dt->dt_ff[old]; 1652 1653 /* 1654 * There are two cases of interest here. 1655 * 1656 * For EDUPFD simply dup (old) to file descriptor 1657 * (new) and return. 1658 * 1659 * For EMOVEFD steal away the file structure from (old) and 1660 * store it in (new). (old) is effectively closed by 1661 * this operation. 1662 * 1663 * Any other error code is just returned. 1664 */ 1665 switch (error) { 1666 case EDUPFD: 1667 /* 1668 * Check that the mode the file is being opened for is a 1669 * subset of the mode of the existing descriptor. 1670 */ 1671 if (((mode & (FREAD|FWRITE)) | fp->f_flag) != fp->f_flag) { 1672 error = EACCES; 1673 break; 1674 } 1675 1676 /* Copy it. */ 1677 error = fd_dup(fp, 0, new, ff->ff_exclose); 1678 break; 1679 1680 case EMOVEFD: 1681 /* Copy it. */ 1682 error = fd_dup(fp, 0, new, ff->ff_exclose); 1683 if (error != 0) { 1684 break; 1685 } 1686 1687 /* Steal away the file pointer from 'old'. */ 1688 (void)fd_close(old); 1689 return 0; 1690 } 1691 1692 fd_putfile(old); 1693 return error; 1694 } 1695 1696 /* 1697 * Close open files on exec. 1698 */ 1699 void 1700 fd_closeexec(void) 1701 { 1702 proc_t *p; 1703 filedesc_t *fdp; 1704 fdfile_t *ff; 1705 lwp_t *l; 1706 fdtab_t *dt; 1707 int fd; 1708 1709 l = curlwp; 1710 p = l->l_proc; 1711 fdp = p->p_fd; 1712 1713 if (fdp->fd_refcnt > 1) { 1714 fdp = fd_copy(); 1715 fd_free(); 1716 p->p_fd = fdp; 1717 l->l_fd = fdp; 1718 } 1719 if (!fdp->fd_exclose) { 1720 return; 1721 } 1722 fdp->fd_exclose = false; 1723 dt = fdp->fd_dt; 1724 1725 for (fd = 0; fd <= fdp->fd_lastfile; fd++) { 1726 if ((ff = dt->dt_ff[fd]) == NULL) { 1727 KASSERT(fd >= NDFDFILE); 1728 continue; 1729 } 1730 KASSERT(fd >= NDFDFILE || 1731 ff == (fdfile_t *)fdp->fd_dfdfile[fd]); 1732 if (ff->ff_file == NULL) 1733 continue; 1734 if (ff->ff_exclose) { 1735 /* 1736 * We need a reference to close the file. 1737 * No other threads can see the fdfile_t at 1738 * this point, so don't bother locking. 1739 */ 1740 KASSERT((ff->ff_refcnt & FR_CLOSING) == 0); 1741 ff->ff_refcnt++; 1742 fd_close(fd); 1743 } 1744 } 1745 } 1746 1747 /* 1748 * Sets descriptor owner. If the owner is a process, 'pgid' 1749 * is set to positive value, process ID. If the owner is process group, 1750 * 'pgid' is set to -pg_id. 1751 */ 1752 int 1753 fsetown(pid_t *pgid, u_long cmd, const void *data) 1754 { 1755 pid_t id = *(const pid_t *)data; 1756 int error; 1757 1758 switch (cmd) { 1759 case TIOCSPGRP: 1760 if (id < 0) 1761 return EINVAL; 1762 id = -id; 1763 break; 1764 default: 1765 break; 1766 } 1767 if (id > 0) { 1768 mutex_enter(proc_lock); 1769 error = proc_find(id) ? 0 : ESRCH; 1770 mutex_exit(proc_lock); 1771 } else if (id < 0) { 1772 error = pgid_in_session(curproc, -id); 1773 } else { 1774 error = 0; 1775 } 1776 if (!error) { 1777 *pgid = id; 1778 } 1779 return error; 1780 } 1781 1782 void 1783 fd_set_exclose(struct lwp *l, int fd, bool exclose) 1784 { 1785 filedesc_t *fdp = l->l_fd; 1786 fdfile_t *ff = fdp->fd_dt->dt_ff[fd]; 1787 1788 ff->ff_exclose = exclose; 1789 if (exclose) 1790 fdp->fd_exclose = true; 1791 } 1792 1793 /* 1794 * Return descriptor owner information. If the value is positive, 1795 * it's process ID. If it's negative, it's process group ID and 1796 * needs the sign removed before use. 1797 */ 1798 int 1799 fgetown(pid_t pgid, u_long cmd, void *data) 1800 { 1801 1802 switch (cmd) { 1803 case TIOCGPGRP: 1804 *(int *)data = -pgid; 1805 break; 1806 default: 1807 *(int *)data = pgid; 1808 break; 1809 } 1810 return 0; 1811 } 1812 1813 /* 1814 * Send signal to descriptor owner, either process or process group. 1815 */ 1816 void 1817 fownsignal(pid_t pgid, int signo, int code, int band, void *fdescdata) 1818 { 1819 ksiginfo_t ksi; 1820 1821 KASSERT(!cpu_intr_p()); 1822 1823 if (pgid == 0) { 1824 return; 1825 } 1826 1827 KSI_INIT(&ksi); 1828 ksi.ksi_signo = signo; 1829 ksi.ksi_code = code; 1830 ksi.ksi_band = band; 1831 1832 mutex_enter(proc_lock); 1833 if (pgid > 0) { 1834 struct proc *p1; 1835 1836 p1 = proc_find(pgid); 1837 if (p1 != NULL) { 1838 kpsignal(p1, &ksi, fdescdata); 1839 } 1840 } else { 1841 struct pgrp *pgrp; 1842 1843 KASSERT(pgid < 0); 1844 pgrp = pgrp_find(-pgid); 1845 if (pgrp != NULL) { 1846 kpgsignal(pgrp, &ksi, fdescdata, 0); 1847 } 1848 } 1849 mutex_exit(proc_lock); 1850 } 1851 1852 int 1853 fd_clone(file_t *fp, unsigned fd, int flag, const struct fileops *fops, 1854 void *data) 1855 { 1856 1857 fp->f_flag = flag; 1858 fp->f_type = DTYPE_MISC; 1859 fp->f_ops = fops; 1860 fp->f_data = data; 1861 curlwp->l_dupfd = fd; 1862 fd_affix(curproc, fp, fd); 1863 1864 return EMOVEFD; 1865 } 1866 1867 int 1868 fnullop_fcntl(file_t *fp, u_int cmd, void *data) 1869 { 1870 1871 if (cmd == F_SETFL) 1872 return 0; 1873 1874 return EOPNOTSUPP; 1875 } 1876 1877 int 1878 fnullop_poll(file_t *fp, int which) 1879 { 1880 1881 return 0; 1882 } 1883 1884 int 1885 fnullop_kqfilter(file_t *fp, struct knote *kn) 1886 { 1887 1888 return 0; 1889 } 1890 1891 void 1892 fnullop_restart(file_t *fp) 1893 { 1894 1895 } 1896 1897 int 1898 fbadop_read(file_t *fp, off_t *offset, struct uio *uio, 1899 kauth_cred_t cred, int flags) 1900 { 1901 1902 return EOPNOTSUPP; 1903 } 1904 1905 int 1906 fbadop_write(file_t *fp, off_t *offset, struct uio *uio, 1907 kauth_cred_t cred, int flags) 1908 { 1909 1910 return EOPNOTSUPP; 1911 } 1912 1913 int 1914 fbadop_ioctl(file_t *fp, u_long com, void *data) 1915 { 1916 1917 return EOPNOTSUPP; 1918 } 1919 1920 int 1921 fbadop_stat(file_t *fp, struct stat *sb) 1922 { 1923 1924 return EOPNOTSUPP; 1925 } 1926 1927 int 1928 fbadop_close(file_t *fp) 1929 { 1930 1931 return EOPNOTSUPP; 1932 } 1933 1934 /* 1935 * sysctl routines pertaining to file descriptors 1936 */ 1937 1938 /* Initialized in sysctl_init() for now... */ 1939 extern kmutex_t sysctl_file_marker_lock; 1940 static u_int sysctl_file_marker = 1; 1941 1942 /* 1943 * Expects to be called with proc_lock and sysctl_file_marker_lock locked. 1944 */ 1945 static void 1946 sysctl_file_marker_reset(void) 1947 { 1948 struct proc *p; 1949 1950 PROCLIST_FOREACH(p, &allproc) { 1951 struct filedesc *fd = p->p_fd; 1952 fdtab_t *dt; 1953 u_int i; 1954 1955 mutex_enter(&fd->fd_lock); 1956 dt = fd->fd_dt; 1957 for (i = 0; i < dt->dt_nfiles; i++) { 1958 struct file *fp; 1959 fdfile_t *ff; 1960 1961 if ((ff = dt->dt_ff[i]) == NULL) { 1962 continue; 1963 } 1964 if ((fp = ff->ff_file) == NULL) { 1965 continue; 1966 } 1967 fp->f_marker = 0; 1968 } 1969 mutex_exit(&fd->fd_lock); 1970 } 1971 } 1972 1973 /* 1974 * sysctl helper routine for kern.file pseudo-subtree. 1975 */ 1976 static int 1977 sysctl_kern_file(SYSCTLFN_ARGS) 1978 { 1979 int error; 1980 size_t buflen; 1981 struct file *fp, fbuf; 1982 char *start, *where; 1983 struct proc *p; 1984 1985 start = where = oldp; 1986 buflen = *oldlenp; 1987 1988 if (where == NULL) { 1989 /* 1990 * overestimate by 10 files 1991 */ 1992 *oldlenp = sizeof(filehead) + (nfiles + 10) * 1993 sizeof(struct file); 1994 return 0; 1995 } 1996 1997 /* 1998 * first sysctl_copyout filehead 1999 */ 2000 if (buflen < sizeof(filehead)) { 2001 *oldlenp = 0; 2002 return 0; 2003 } 2004 sysctl_unlock(); 2005 error = sysctl_copyout(l, &filehead, where, sizeof(filehead)); 2006 if (error) { 2007 sysctl_relock(); 2008 return error; 2009 } 2010 buflen -= sizeof(filehead); 2011 where += sizeof(filehead); 2012 2013 /* 2014 * followed by an array of file structures 2015 */ 2016 mutex_enter(&sysctl_file_marker_lock); 2017 mutex_enter(proc_lock); 2018 PROCLIST_FOREACH(p, &allproc) { 2019 struct filedesc *fd; 2020 fdtab_t *dt; 2021 u_int i; 2022 2023 if (p->p_stat == SIDL) { 2024 /* skip embryonic processes */ 2025 continue; 2026 } 2027 mutex_enter(p->p_lock); 2028 error = kauth_authorize_process(l->l_cred, 2029 KAUTH_PROCESS_CANSEE, p, 2030 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_OPENFILES), 2031 NULL, NULL); 2032 mutex_exit(p->p_lock); 2033 if (error != 0) { 2034 /* 2035 * Don't leak kauth retval if we're silently 2036 * skipping this entry. 2037 */ 2038 error = 0; 2039 continue; 2040 } 2041 2042 /* 2043 * Grab a hold on the process. 2044 */ 2045 if (!rw_tryenter(&p->p_reflock, RW_READER)) { 2046 continue; 2047 } 2048 mutex_exit(proc_lock); 2049 2050 fd = p->p_fd; 2051 mutex_enter(&fd->fd_lock); 2052 dt = fd->fd_dt; 2053 for (i = 0; i < dt->dt_nfiles; i++) { 2054 fdfile_t *ff; 2055 2056 if ((ff = dt->dt_ff[i]) == NULL) { 2057 continue; 2058 } 2059 if ((fp = ff->ff_file) == NULL) { 2060 continue; 2061 } 2062 2063 mutex_enter(&fp->f_lock); 2064 2065 if ((fp->f_count == 0) || 2066 (fp->f_marker == sysctl_file_marker)) { 2067 mutex_exit(&fp->f_lock); 2068 continue; 2069 } 2070 2071 /* Check that we have enough space. */ 2072 if (buflen < sizeof(struct file)) { 2073 *oldlenp = where - start; 2074 mutex_exit(&fp->f_lock); 2075 error = ENOMEM; 2076 break; 2077 } 2078 2079 memcpy(&fbuf, fp, sizeof(fbuf)); 2080 mutex_exit(&fp->f_lock); 2081 error = sysctl_copyout(l, &fbuf, where, sizeof(fbuf)); 2082 if (error) { 2083 break; 2084 } 2085 buflen -= sizeof(struct file); 2086 where += sizeof(struct file); 2087 2088 fp->f_marker = sysctl_file_marker; 2089 } 2090 mutex_exit(&fd->fd_lock); 2091 2092 /* 2093 * Release reference to process. 2094 */ 2095 mutex_enter(proc_lock); 2096 rw_exit(&p->p_reflock); 2097 2098 if (error) 2099 break; 2100 } 2101 2102 sysctl_file_marker++; 2103 /* Reset all markers if wrapped. */ 2104 if (sysctl_file_marker == 0) { 2105 sysctl_file_marker_reset(); 2106 sysctl_file_marker++; 2107 } 2108 2109 mutex_exit(proc_lock); 2110 mutex_exit(&sysctl_file_marker_lock); 2111 2112 *oldlenp = where - start; 2113 sysctl_relock(); 2114 return error; 2115 } 2116 2117 /* 2118 * sysctl helper function for kern.file2 2119 */ 2120 static int 2121 sysctl_kern_file2(SYSCTLFN_ARGS) 2122 { 2123 struct proc *p; 2124 struct file *fp; 2125 struct filedesc *fd; 2126 struct kinfo_file kf; 2127 char *dp; 2128 u_int i, op; 2129 size_t len, needed, elem_size, out_size; 2130 int error, arg, elem_count; 2131 fdfile_t *ff; 2132 fdtab_t *dt; 2133 2134 if (namelen == 1 && name[0] == CTL_QUERY) 2135 return sysctl_query(SYSCTLFN_CALL(rnode)); 2136 2137 if (namelen != 4) 2138 return EINVAL; 2139 2140 error = 0; 2141 dp = oldp; 2142 len = (oldp != NULL) ? *oldlenp : 0; 2143 op = name[0]; 2144 arg = name[1]; 2145 elem_size = name[2]; 2146 elem_count = name[3]; 2147 out_size = MIN(sizeof(kf), elem_size); 2148 needed = 0; 2149 2150 if (elem_size < 1 || elem_count < 0) 2151 return EINVAL; 2152 2153 switch (op) { 2154 case KERN_FILE_BYFILE: 2155 case KERN_FILE_BYPID: 2156 /* 2157 * We're traversing the process list in both cases; the BYFILE 2158 * case does additional work of keeping track of files already 2159 * looked at. 2160 */ 2161 2162 /* doesn't use arg so it must be zero */ 2163 if ((op == KERN_FILE_BYFILE) && (arg != 0)) 2164 return EINVAL; 2165 2166 if ((op == KERN_FILE_BYPID) && (arg < -1)) 2167 /* -1 means all processes */ 2168 return EINVAL; 2169 2170 sysctl_unlock(); 2171 if (op == KERN_FILE_BYFILE) 2172 mutex_enter(&sysctl_file_marker_lock); 2173 mutex_enter(proc_lock); 2174 PROCLIST_FOREACH(p, &allproc) { 2175 if (p->p_stat == SIDL) { 2176 /* skip embryonic processes */ 2177 continue; 2178 } 2179 if (arg > 0 && p->p_pid != arg) { 2180 /* pick only the one we want */ 2181 /* XXX want 0 to mean "kernel files" */ 2182 continue; 2183 } 2184 mutex_enter(p->p_lock); 2185 error = kauth_authorize_process(l->l_cred, 2186 KAUTH_PROCESS_CANSEE, p, 2187 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_OPENFILES), 2188 NULL, NULL); 2189 mutex_exit(p->p_lock); 2190 if (error != 0) { 2191 /* 2192 * Don't leak kauth retval if we're silently 2193 * skipping this entry. 2194 */ 2195 error = 0; 2196 continue; 2197 } 2198 2199 /* 2200 * Grab a hold on the process. 2201 */ 2202 if (!rw_tryenter(&p->p_reflock, RW_READER)) { 2203 continue; 2204 } 2205 mutex_exit(proc_lock); 2206 2207 fd = p->p_fd; 2208 mutex_enter(&fd->fd_lock); 2209 dt = fd->fd_dt; 2210 for (i = 0; i < dt->dt_nfiles; i++) { 2211 if ((ff = dt->dt_ff[i]) == NULL) { 2212 continue; 2213 } 2214 if ((fp = ff->ff_file) == NULL) { 2215 continue; 2216 } 2217 2218 if ((op == KERN_FILE_BYFILE) && 2219 (fp->f_marker == sysctl_file_marker)) { 2220 continue; 2221 } 2222 if (len >= elem_size && elem_count > 0) { 2223 mutex_enter(&fp->f_lock); 2224 fill_file(&kf, fp, ff, i, p->p_pid); 2225 mutex_exit(&fp->f_lock); 2226 mutex_exit(&fd->fd_lock); 2227 error = sysctl_copyout(l, 2228 &kf, dp, out_size); 2229 mutex_enter(&fd->fd_lock); 2230 if (error) 2231 break; 2232 dp += elem_size; 2233 len -= elem_size; 2234 } 2235 if (op == KERN_FILE_BYFILE) 2236 fp->f_marker = sysctl_file_marker; 2237 needed += elem_size; 2238 if (elem_count > 0 && elem_count != INT_MAX) 2239 elem_count--; 2240 } 2241 mutex_exit(&fd->fd_lock); 2242 2243 /* 2244 * Release reference to process. 2245 */ 2246 mutex_enter(proc_lock); 2247 rw_exit(&p->p_reflock); 2248 } 2249 if (op == KERN_FILE_BYFILE) { 2250 sysctl_file_marker++; 2251 2252 /* Reset all markers if wrapped. */ 2253 if (sysctl_file_marker == 0) { 2254 sysctl_file_marker_reset(); 2255 sysctl_file_marker++; 2256 } 2257 } 2258 mutex_exit(proc_lock); 2259 if (op == KERN_FILE_BYFILE) 2260 mutex_exit(&sysctl_file_marker_lock); 2261 sysctl_relock(); 2262 break; 2263 default: 2264 return EINVAL; 2265 } 2266 2267 if (oldp == NULL) 2268 needed += KERN_FILESLOP * elem_size; 2269 *oldlenp = needed; 2270 2271 return error; 2272 } 2273 2274 static void 2275 fill_file(struct kinfo_file *kp, const file_t *fp, const fdfile_t *ff, 2276 int i, pid_t pid) 2277 { 2278 2279 memset(kp, 0, sizeof(*kp)); 2280 2281 kp->ki_fileaddr = PTRTOUINT64(fp); 2282 kp->ki_flag = fp->f_flag; 2283 kp->ki_iflags = 0; 2284 kp->ki_ftype = fp->f_type; 2285 kp->ki_count = fp->f_count; 2286 kp->ki_msgcount = fp->f_msgcount; 2287 kp->ki_fucred = PTRTOUINT64(fp->f_cred); 2288 kp->ki_fuid = kauth_cred_geteuid(fp->f_cred); 2289 kp->ki_fgid = kauth_cred_getegid(fp->f_cred); 2290 kp->ki_fops = PTRTOUINT64(fp->f_ops); 2291 kp->ki_foffset = fp->f_offset; 2292 kp->ki_fdata = PTRTOUINT64(fp->f_data); 2293 2294 /* vnode information to glue this file to something */ 2295 if (fp->f_type == DTYPE_VNODE) { 2296 struct vnode *vp = (struct vnode *)fp->f_data; 2297 2298 kp->ki_vun = PTRTOUINT64(vp->v_un.vu_socket); 2299 kp->ki_vsize = vp->v_size; 2300 kp->ki_vtype = vp->v_type; 2301 kp->ki_vtag = vp->v_tag; 2302 kp->ki_vdata = PTRTOUINT64(vp->v_data); 2303 } 2304 2305 /* process information when retrieved via KERN_FILE_BYPID */ 2306 if (ff != NULL) { 2307 kp->ki_pid = pid; 2308 kp->ki_fd = i; 2309 kp->ki_ofileflags = ff->ff_exclose; 2310 kp->ki_usecount = ff->ff_refcnt; 2311 } 2312 } 2313