kernel.c revision 318910
1/* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21/* 22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 23 * Copyright (c) 2012, 2015 by Delphix. All rights reserved. 24 * Copyright (c) 2013, Joyent, Inc. All rights reserved. 25 */ 26 27#include <assert.h> 28#include <fcntl.h> 29#include <poll.h> 30#include <stdio.h> 31#include <stdlib.h> 32#include <string.h> 33#include <zlib.h> 34#include <libgen.h> 35#include <sys/spa.h> 36#include <sys/stat.h> 37#include <sys/processor.h> 38#include <sys/zfs_context.h> 39#include <sys/rrwlock.h> 40#include <sys/zmod.h> 41#include <sys/utsname.h> 42#include <sys/systeminfo.h> 43 44/* 45 * Emulation of kernel services in userland. 46 */ 47 48#ifndef __FreeBSD__ 49int aok; 50#endif 51uint64_t physmem; 52vnode_t *rootdir = (vnode_t *)0xabcd1234; 53char hw_serial[HW_HOSTID_LEN]; 54#ifdef illumos 55kmutex_t cpu_lock; 56#endif 57 58/* If set, all blocks read will be copied to the specified directory. */ 59char *vn_dumpdir = NULL; 60 61struct utsname utsname = { 62 "userland", "libzpool", "1", "1", "na" 63}; 64 65/* this only exists to have its address taken */ 66struct proc p0; 67 68/* 69 * ========================================================================= 70 * threads 71 * ========================================================================= 72 */ 73/*ARGSUSED*/ 74kthread_t * 75zk_thread_create(void (*func)(), void *arg) 76{ 77 thread_t tid; 78 79 VERIFY(thr_create(0, 0, (void *(*)(void *))func, arg, THR_DETACHED, 80 &tid) == 0); 81 82 return ((void *)(uintptr_t)tid); 83} 84 85/* 86 * ========================================================================= 87 * kstats 88 * ========================================================================= 89 */ 90/*ARGSUSED*/ 91kstat_t * 92kstat_create(char *module, int instance, char *name, char *class, 93 uchar_t type, ulong_t ndata, uchar_t ks_flag) 94{ 95 return (NULL); 96} 97 98/*ARGSUSED*/ 99void 100kstat_install(kstat_t *ksp) 101{} 102 103/*ARGSUSED*/ 104void 105kstat_delete(kstat_t *ksp) 106{} 107 108/* 109 * ========================================================================= 110 * mutexes 111 * ========================================================================= 112 */ 113void 114zmutex_init(kmutex_t *mp) 115{ 116 mp->m_owner = NULL; 117 mp->initialized = B_TRUE; 118 (void) _mutex_init(&mp->m_lock, USYNC_THREAD, NULL); 119} 120 121void 122zmutex_destroy(kmutex_t *mp) 123{ 124 ASSERT(mp->initialized == B_TRUE); 125 ASSERT(mp->m_owner == NULL); 126 (void) _mutex_destroy(&(mp)->m_lock); 127 mp->m_owner = (void *)-1UL; 128 mp->initialized = B_FALSE; 129} 130 131int 132zmutex_owned(kmutex_t *mp) 133{ 134 ASSERT(mp->initialized == B_TRUE); 135 136 return (mp->m_owner == curthread); 137} 138 139void 140mutex_enter(kmutex_t *mp) 141{ 142 ASSERT(mp->initialized == B_TRUE); 143 ASSERT(mp->m_owner != (void *)-1UL); 144 ASSERT(mp->m_owner != curthread); 145 VERIFY(mutex_lock(&mp->m_lock) == 0); 146 ASSERT(mp->m_owner == NULL); 147 mp->m_owner = curthread; 148} 149 150int 151mutex_tryenter(kmutex_t *mp) 152{ 153 ASSERT(mp->initialized == B_TRUE); 154 ASSERT(mp->m_owner != (void *)-1UL); 155 if (0 == mutex_trylock(&mp->m_lock)) { 156 ASSERT(mp->m_owner == NULL); 157 mp->m_owner = curthread; 158 return (1); 159 } else { 160 return (0); 161 } 162} 163 164void 165mutex_exit(kmutex_t *mp) 166{ 167 ASSERT(mp->initialized == B_TRUE); 168 ASSERT(mutex_owner(mp) == curthread); 169 mp->m_owner = NULL; 170 VERIFY(mutex_unlock(&mp->m_lock) == 0); 171} 172 173void * 174mutex_owner(kmutex_t *mp) 175{ 176 ASSERT(mp->initialized == B_TRUE); 177 return (mp->m_owner); 178} 179 180/* 181 * ========================================================================= 182 * rwlocks 183 * ========================================================================= 184 */ 185/*ARGSUSED*/ 186void 187rw_init(krwlock_t *rwlp, char *name, int type, void *arg) 188{ 189 rwlock_init(&rwlp->rw_lock, USYNC_THREAD, NULL); 190 rwlp->rw_owner = NULL; 191 rwlp->initialized = B_TRUE; 192 rwlp->rw_count = 0; 193} 194 195void 196rw_destroy(krwlock_t *rwlp) 197{ 198 ASSERT(rwlp->rw_count == 0); 199 rwlock_destroy(&rwlp->rw_lock); 200 rwlp->rw_owner = (void *)-1UL; 201 rwlp->initialized = B_FALSE; 202} 203 204void 205rw_enter(krwlock_t *rwlp, krw_t rw) 206{ 207 //ASSERT(!RW_LOCK_HELD(rwlp)); 208 ASSERT(rwlp->initialized == B_TRUE); 209 ASSERT(rwlp->rw_owner != (void *)-1UL); 210 ASSERT(rwlp->rw_owner != curthread); 211 212 if (rw == RW_READER) { 213 VERIFY(rw_rdlock(&rwlp->rw_lock) == 0); 214 ASSERT(rwlp->rw_count >= 0); 215 atomic_add_int(&rwlp->rw_count, 1); 216 } else { 217 VERIFY(rw_wrlock(&rwlp->rw_lock) == 0); 218 ASSERT(rwlp->rw_count == 0); 219 rwlp->rw_count = -1; 220 rwlp->rw_owner = curthread; 221 } 222} 223 224void 225rw_exit(krwlock_t *rwlp) 226{ 227 ASSERT(rwlp->initialized == B_TRUE); 228 ASSERT(rwlp->rw_owner != (void *)-1UL); 229 230 if (rwlp->rw_owner == curthread) { 231 /* Write locked. */ 232 ASSERT(rwlp->rw_count == -1); 233 rwlp->rw_count = 0; 234 rwlp->rw_owner = NULL; 235 } else { 236 /* Read locked. */ 237 ASSERT(rwlp->rw_count > 0); 238 atomic_add_int(&rwlp->rw_count, -1); 239 } 240 VERIFY(rw_unlock(&rwlp->rw_lock) == 0); 241} 242 243int 244rw_tryenter(krwlock_t *rwlp, krw_t rw) 245{ 246 int rv; 247 248 ASSERT(rwlp->initialized == B_TRUE); 249 ASSERT(rwlp->rw_owner != (void *)-1UL); 250 ASSERT(rwlp->rw_owner != curthread); 251 252 if (rw == RW_READER) 253 rv = rw_tryrdlock(&rwlp->rw_lock); 254 else 255 rv = rw_trywrlock(&rwlp->rw_lock); 256 257 if (rv == 0) { 258 ASSERT(rwlp->rw_owner == NULL); 259 if (rw == RW_READER) { 260 ASSERT(rwlp->rw_count >= 0); 261 atomic_add_int(&rwlp->rw_count, 1); 262 } else { 263 ASSERT(rwlp->rw_count == 0); 264 rwlp->rw_count = -1; 265 rwlp->rw_owner = curthread; 266 } 267 return (1); 268 } 269 270 return (0); 271} 272 273/*ARGSUSED*/ 274int 275rw_tryupgrade(krwlock_t *rwlp) 276{ 277 ASSERT(rwlp->initialized == B_TRUE); 278 ASSERT(rwlp->rw_owner != (void *)-1UL); 279 280 return (0); 281} 282 283int 284rw_lock_held(krwlock_t *rwlp) 285{ 286 287 return (rwlp->rw_count != 0); 288} 289 290/* 291 * ========================================================================= 292 * condition variables 293 * ========================================================================= 294 */ 295/*ARGSUSED*/ 296void 297cv_init(kcondvar_t *cv, char *name, int type, void *arg) 298{ 299 VERIFY(cond_init(cv, name, NULL) == 0); 300} 301 302void 303cv_destroy(kcondvar_t *cv) 304{ 305 VERIFY(cond_destroy(cv) == 0); 306} 307 308void 309cv_wait(kcondvar_t *cv, kmutex_t *mp) 310{ 311 ASSERT(mutex_owner(mp) == curthread); 312 mp->m_owner = NULL; 313 int ret = cond_wait(cv, &mp->m_lock); 314 VERIFY(ret == 0 || ret == EINTR); 315 mp->m_owner = curthread; 316} 317 318clock_t 319cv_timedwait(kcondvar_t *cv, kmutex_t *mp, clock_t abstime) 320{ 321 int error; 322 struct timespec ts; 323 struct timeval tv; 324 clock_t delta; 325 326 abstime += ddi_get_lbolt(); 327top: 328 delta = abstime - ddi_get_lbolt(); 329 if (delta <= 0) 330 return (-1); 331 332 if (gettimeofday(&tv, NULL) != 0) 333 assert(!"gettimeofday() failed"); 334 335 ts.tv_sec = tv.tv_sec + delta / hz; 336 ts.tv_nsec = tv.tv_usec * 1000 + (delta % hz) * (NANOSEC / hz); 337 ASSERT(ts.tv_nsec >= 0); 338 339 if (ts.tv_nsec >= NANOSEC) { 340 ts.tv_sec++; 341 ts.tv_nsec -= NANOSEC; 342 } 343 344 ASSERT(mutex_owner(mp) == curthread); 345 mp->m_owner = NULL; 346 error = pthread_cond_timedwait(cv, &mp->m_lock, &ts); 347 mp->m_owner = curthread; 348 349 if (error == EINTR) 350 goto top; 351 352 if (error == ETIMEDOUT) 353 return (-1); 354 355 ASSERT(error == 0); 356 357 return (1); 358} 359 360/*ARGSUSED*/ 361clock_t 362cv_timedwait_hires(kcondvar_t *cv, kmutex_t *mp, hrtime_t tim, hrtime_t res, 363 int flag) 364{ 365 int error; 366 timespec_t ts; 367 hrtime_t delta; 368 369 ASSERT(flag == 0 || flag == CALLOUT_FLAG_ABSOLUTE); 370 371top: 372 delta = tim; 373 if (flag & CALLOUT_FLAG_ABSOLUTE) 374 delta -= gethrtime(); 375 376 if (delta <= 0) 377 return (-1); 378 379 clock_gettime(CLOCK_REALTIME, &ts); 380 ts.tv_sec += delta / NANOSEC; 381 ts.tv_nsec += delta % NANOSEC; 382 if (ts.tv_nsec >= NANOSEC) { 383 ts.tv_sec++; 384 ts.tv_nsec -= NANOSEC; 385 } 386 387 ASSERT(mutex_owner(mp) == curthread); 388 mp->m_owner = NULL; 389 error = pthread_cond_timedwait(cv, &mp->m_lock, &ts); 390 mp->m_owner = curthread; 391 392 if (error == ETIMEDOUT) 393 return (-1); 394 395 if (error == EINTR) 396 goto top; 397 398 ASSERT(error == 0); 399 400 return (1); 401} 402 403void 404cv_signal(kcondvar_t *cv) 405{ 406 VERIFY(cond_signal(cv) == 0); 407} 408 409void 410cv_broadcast(kcondvar_t *cv) 411{ 412 VERIFY(cond_broadcast(cv) == 0); 413} 414 415/* 416 * ========================================================================= 417 * vnode operations 418 * ========================================================================= 419 */ 420/* 421 * Note: for the xxxat() versions of these functions, we assume that the 422 * starting vp is always rootdir (which is true for spa_directory.c, the only 423 * ZFS consumer of these interfaces). We assert this is true, and then emulate 424 * them by adding '/' in front of the path. 425 */ 426 427/*ARGSUSED*/ 428int 429vn_open(char *path, int x1, int flags, int mode, vnode_t **vpp, int x2, int x3) 430{ 431 int fd; 432 int dump_fd; 433 vnode_t *vp; 434 int old_umask; 435 char realpath[MAXPATHLEN]; 436 struct stat64 st; 437 438 /* 439 * If we're accessing a real disk from userland, we need to use 440 * the character interface to avoid caching. This is particularly 441 * important if we're trying to look at a real in-kernel storage 442 * pool from userland, e.g. via zdb, because otherwise we won't 443 * see the changes occurring under the segmap cache. 444 * On the other hand, the stupid character device returns zero 445 * for its size. So -- gag -- we open the block device to get 446 * its size, and remember it for subsequent VOP_GETATTR(). 447 */ 448 if (strncmp(path, "/dev/", 5) == 0) { 449 char *dsk; 450 fd = open64(path, O_RDONLY); 451 if (fd == -1) 452 return (errno); 453 if (fstat64(fd, &st) == -1) { 454 close(fd); 455 return (errno); 456 } 457 close(fd); 458 (void) sprintf(realpath, "%s", path); 459 dsk = strstr(path, "/dsk/"); 460 if (dsk != NULL) 461 (void) sprintf(realpath + (dsk - path) + 1, "r%s", 462 dsk + 1); 463 } else { 464 (void) sprintf(realpath, "%s", path); 465 if (!(flags & FCREAT) && stat64(realpath, &st) == -1) 466 return (errno); 467 } 468 469 if (flags & FCREAT) 470 old_umask = umask(0); 471 472 /* 473 * The construct 'flags - FREAD' conveniently maps combinations of 474 * FREAD and FWRITE to the corresponding O_RDONLY, O_WRONLY, and O_RDWR. 475 */ 476 fd = open64(realpath, flags - FREAD, mode); 477 478 if (flags & FCREAT) 479 (void) umask(old_umask); 480 481 if (vn_dumpdir != NULL) { 482 char dumppath[MAXPATHLEN]; 483 (void) snprintf(dumppath, sizeof (dumppath), 484 "%s/%s", vn_dumpdir, basename(realpath)); 485 dump_fd = open64(dumppath, O_CREAT | O_WRONLY, 0666); 486 if (dump_fd == -1) 487 return (errno); 488 } else { 489 dump_fd = -1; 490 } 491 492 if (fd == -1) 493 return (errno); 494 495 if (fstat64(fd, &st) == -1) { 496 close(fd); 497 return (errno); 498 } 499 500 (void) fcntl(fd, F_SETFD, FD_CLOEXEC); 501 502 *vpp = vp = umem_zalloc(sizeof (vnode_t), UMEM_NOFAIL); 503 504 vp->v_fd = fd; 505 vp->v_size = st.st_size; 506 vp->v_path = spa_strdup(path); 507 vp->v_dump_fd = dump_fd; 508 509 return (0); 510} 511 512/*ARGSUSED*/ 513int 514vn_openat(char *path, int x1, int flags, int mode, vnode_t **vpp, int x2, 515 int x3, vnode_t *startvp, int fd) 516{ 517 char *realpath = umem_alloc(strlen(path) + 2, UMEM_NOFAIL); 518 int ret; 519 520 ASSERT(startvp == rootdir); 521 (void) sprintf(realpath, "/%s", path); 522 523 /* fd ignored for now, need if want to simulate nbmand support */ 524 ret = vn_open(realpath, x1, flags, mode, vpp, x2, x3); 525 526 umem_free(realpath, strlen(path) + 2); 527 528 return (ret); 529} 530 531/*ARGSUSED*/ 532int 533vn_rdwr(int uio, vnode_t *vp, void *addr, ssize_t len, offset_t offset, 534 int x1, int x2, rlim64_t x3, void *x4, ssize_t *residp) 535{ 536 ssize_t iolen, split; 537 538 if (uio == UIO_READ) { 539 iolen = pread64(vp->v_fd, addr, len, offset); 540 if (vp->v_dump_fd != -1) { 541 int status = 542 pwrite64(vp->v_dump_fd, addr, iolen, offset); 543 ASSERT(status != -1); 544 } 545 } else { 546 /* 547 * To simulate partial disk writes, we split writes into two 548 * system calls so that the process can be killed in between. 549 */ 550 int sectors = len >> SPA_MINBLOCKSHIFT; 551 split = (sectors > 0 ? rand() % sectors : 0) << 552 SPA_MINBLOCKSHIFT; 553 iolen = pwrite64(vp->v_fd, addr, split, offset); 554 iolen += pwrite64(vp->v_fd, (char *)addr + split, 555 len - split, offset + split); 556 } 557 558 if (iolen == -1) 559 return (errno); 560 if (residp) 561 *residp = len - iolen; 562 else if (iolen != len) 563 return (EIO); 564 return (0); 565} 566 567void 568vn_close(vnode_t *vp, int openflag, cred_t *cr, kthread_t *td) 569{ 570 close(vp->v_fd); 571 if (vp->v_dump_fd != -1) 572 close(vp->v_dump_fd); 573 spa_strfree(vp->v_path); 574 umem_free(vp, sizeof (vnode_t)); 575} 576 577/* 578 * At a minimum we need to update the size since vdev_reopen() 579 * will no longer call vn_openat(). 580 */ 581int 582fop_getattr(vnode_t *vp, vattr_t *vap) 583{ 584 struct stat64 st; 585 586 if (fstat64(vp->v_fd, &st) == -1) { 587 close(vp->v_fd); 588 return (errno); 589 } 590 591 vap->va_size = st.st_size; 592 return (0); 593} 594 595#ifdef ZFS_DEBUG 596 597/* 598 * ========================================================================= 599 * Figure out which debugging statements to print 600 * ========================================================================= 601 */ 602 603static char *dprintf_string; 604static int dprintf_print_all; 605 606int 607dprintf_find_string(const char *string) 608{ 609 char *tmp_str = dprintf_string; 610 int len = strlen(string); 611 612 /* 613 * Find out if this is a string we want to print. 614 * String format: file1.c,function_name1,file2.c,file3.c 615 */ 616 617 while (tmp_str != NULL) { 618 if (strncmp(tmp_str, string, len) == 0 && 619 (tmp_str[len] == ',' || tmp_str[len] == '\0')) 620 return (1); 621 tmp_str = strchr(tmp_str, ','); 622 if (tmp_str != NULL) 623 tmp_str++; /* Get rid of , */ 624 } 625 return (0); 626} 627 628void 629dprintf_setup(int *argc, char **argv) 630{ 631 int i, j; 632 633 /* 634 * Debugging can be specified two ways: by setting the 635 * environment variable ZFS_DEBUG, or by including a 636 * "debug=..." argument on the command line. The command 637 * line setting overrides the environment variable. 638 */ 639 640 for (i = 1; i < *argc; i++) { 641 int len = strlen("debug="); 642 /* First look for a command line argument */ 643 if (strncmp("debug=", argv[i], len) == 0) { 644 dprintf_string = argv[i] + len; 645 /* Remove from args */ 646 for (j = i; j < *argc; j++) 647 argv[j] = argv[j+1]; 648 argv[j] = NULL; 649 (*argc)--; 650 } 651 } 652 653 if (dprintf_string == NULL) { 654 /* Look for ZFS_DEBUG environment variable */ 655 dprintf_string = getenv("ZFS_DEBUG"); 656 } 657 658 /* 659 * Are we just turning on all debugging? 660 */ 661 if (dprintf_find_string("on")) 662 dprintf_print_all = 1; 663} 664 665int 666sysctl_handle_64(SYSCTL_HANDLER_ARGS) 667{ 668 return (0); 669} 670 671/* 672 * ========================================================================= 673 * debug printfs 674 * ========================================================================= 675 */ 676void 677__dprintf(const char *file, const char *func, int line, const char *fmt, ...) 678{ 679 const char *newfile; 680 va_list adx; 681 682 /* 683 * Get rid of annoying "../common/" prefix to filename. 684 */ 685 newfile = strrchr(file, '/'); 686 if (newfile != NULL) { 687 newfile = newfile + 1; /* Get rid of leading / */ 688 } else { 689 newfile = file; 690 } 691 692 if (dprintf_print_all || 693 dprintf_find_string(newfile) || 694 dprintf_find_string(func)) { 695 /* Print out just the function name if requested */ 696 flockfile(stdout); 697 if (dprintf_find_string("pid")) 698 (void) printf("%d ", getpid()); 699 if (dprintf_find_string("tid")) 700 (void) printf("%ul ", thr_self()); 701#if 0 702 if (dprintf_find_string("cpu")) 703 (void) printf("%u ", getcpuid()); 704#endif 705 if (dprintf_find_string("time")) 706 (void) printf("%llu ", gethrtime()); 707 if (dprintf_find_string("long")) 708 (void) printf("%s, line %d: ", newfile, line); 709 (void) printf("%s: ", func); 710 va_start(adx, fmt); 711 (void) vprintf(fmt, adx); 712 va_end(adx); 713 funlockfile(stdout); 714 } 715} 716 717#endif /* ZFS_DEBUG */ 718 719/* 720 * ========================================================================= 721 * cmn_err() and panic() 722 * ========================================================================= 723 */ 724static char ce_prefix[CE_IGNORE][10] = { "", "NOTICE: ", "WARNING: ", "" }; 725static char ce_suffix[CE_IGNORE][2] = { "", "\n", "\n", "" }; 726 727void 728vpanic(const char *fmt, va_list adx) 729{ 730 (void) fprintf(stderr, "error: "); 731 (void) vfprintf(stderr, fmt, adx); 732 (void) fprintf(stderr, "\n"); 733 734 abort(); /* think of it as a "user-level crash dump" */ 735} 736 737void 738panic(const char *fmt, ...) 739{ 740 va_list adx; 741 742 va_start(adx, fmt); 743 vpanic(fmt, adx); 744 va_end(adx); 745} 746 747void 748vcmn_err(int ce, const char *fmt, va_list adx) 749{ 750 if (ce == CE_PANIC) 751 vpanic(fmt, adx); 752 if (ce != CE_NOTE) { /* suppress noise in userland stress testing */ 753 (void) fprintf(stderr, "%s", ce_prefix[ce]); 754 (void) vfprintf(stderr, fmt, adx); 755 (void) fprintf(stderr, "%s", ce_suffix[ce]); 756 } 757} 758 759/*PRINTFLIKE2*/ 760void 761cmn_err(int ce, const char *fmt, ...) 762{ 763 va_list adx; 764 765 va_start(adx, fmt); 766 vcmn_err(ce, fmt, adx); 767 va_end(adx); 768} 769 770/* 771 * ========================================================================= 772 * kobj interfaces 773 * ========================================================================= 774 */ 775struct _buf * 776kobj_open_file(char *name) 777{ 778 struct _buf *file; 779 vnode_t *vp; 780 781 /* set vp as the _fd field of the file */ 782 if (vn_openat(name, UIO_SYSSPACE, FREAD, 0, &vp, 0, 0, rootdir, 783 -1) != 0) 784 return ((void *)-1UL); 785 786 file = umem_zalloc(sizeof (struct _buf), UMEM_NOFAIL); 787 file->_fd = (intptr_t)vp; 788 return (file); 789} 790 791int 792kobj_read_file(struct _buf *file, char *buf, unsigned size, unsigned off) 793{ 794 ssize_t resid; 795 796 vn_rdwr(UIO_READ, (vnode_t *)file->_fd, buf, size, (offset_t)off, 797 UIO_SYSSPACE, 0, 0, 0, &resid); 798 799 return (size - resid); 800} 801 802void 803kobj_close_file(struct _buf *file) 804{ 805 vn_close((vnode_t *)file->_fd, 0, NULL, NULL); 806 umem_free(file, sizeof (struct _buf)); 807} 808 809int 810kobj_get_filesize(struct _buf *file, uint64_t *size) 811{ 812 struct stat64 st; 813 vnode_t *vp = (vnode_t *)file->_fd; 814 815 if (fstat64(vp->v_fd, &st) == -1) { 816 vn_close(vp, 0, NULL, NULL); 817 return (errno); 818 } 819 *size = st.st_size; 820 return (0); 821} 822 823/* 824 * ========================================================================= 825 * misc routines 826 * ========================================================================= 827 */ 828 829void 830delay(clock_t ticks) 831{ 832 poll(0, 0, ticks * (1000 / hz)); 833} 834 835#if 0 836/* 837 * Find highest one bit set. 838 * Returns bit number + 1 of highest bit that is set, otherwise returns 0. 839 */ 840int 841highbit64(uint64_t i) 842{ 843 int h = 1; 844 845 if (i == 0) 846 return (0); 847 if (i & 0xffffffff00000000ULL) { 848 h += 32; i >>= 32; 849 } 850 if (i & 0xffff0000) { 851 h += 16; i >>= 16; 852 } 853 if (i & 0xff00) { 854 h += 8; i >>= 8; 855 } 856 if (i & 0xf0) { 857 h += 4; i >>= 4; 858 } 859 if (i & 0xc) { 860 h += 2; i >>= 2; 861 } 862 if (i & 0x2) { 863 h += 1; 864 } 865 return (h); 866} 867#endif 868 869static int random_fd = -1, urandom_fd = -1; 870 871static int 872random_get_bytes_common(uint8_t *ptr, size_t len, int fd) 873{ 874 size_t resid = len; 875 ssize_t bytes; 876 877 ASSERT(fd != -1); 878 879 while (resid != 0) { 880 bytes = read(fd, ptr, resid); 881 ASSERT3S(bytes, >=, 0); 882 ptr += bytes; 883 resid -= bytes; 884 } 885 886 return (0); 887} 888 889int 890random_get_bytes(uint8_t *ptr, size_t len) 891{ 892 return (random_get_bytes_common(ptr, len, random_fd)); 893} 894 895int 896random_get_pseudo_bytes(uint8_t *ptr, size_t len) 897{ 898 return (random_get_bytes_common(ptr, len, urandom_fd)); 899} 900 901int 902ddi_strtoul(const char *hw_serial, char **nptr, int base, unsigned long *result) 903{ 904 char *end; 905 906 *result = strtoul(hw_serial, &end, base); 907 if (*result == 0) 908 return (errno); 909 return (0); 910} 911 912int 913ddi_strtoull(const char *str, char **nptr, int base, u_longlong_t *result) 914{ 915 char *end; 916 917 *result = strtoull(str, &end, base); 918 if (*result == 0) 919 return (errno); 920 return (0); 921} 922 923#ifdef illumos 924/* ARGSUSED */ 925cyclic_id_t 926cyclic_add(cyc_handler_t *hdlr, cyc_time_t *when) 927{ 928 return (1); 929} 930 931/* ARGSUSED */ 932void 933cyclic_remove(cyclic_id_t id) 934{ 935} 936 937/* ARGSUSED */ 938int 939cyclic_reprogram(cyclic_id_t id, hrtime_t expiration) 940{ 941 return (1); 942} 943#endif 944 945/* 946 * ========================================================================= 947 * kernel emulation setup & teardown 948 * ========================================================================= 949 */ 950static int 951umem_out_of_memory(void) 952{ 953 char errmsg[] = "out of memory -- generating core dump\n"; 954 955 write(fileno(stderr), errmsg, sizeof (errmsg)); 956 abort(); 957 return (0); 958} 959 960void 961kernel_init(int mode) 962{ 963 extern uint_t rrw_tsd_key; 964 965 umem_nofail_callback(umem_out_of_memory); 966 967 physmem = sysconf(_SC_PHYS_PAGES); 968 969 dprintf("physmem = %llu pages (%.2f GB)\n", physmem, 970 (double)physmem * sysconf(_SC_PAGE_SIZE) / (1ULL << 30)); 971 972 (void) snprintf(hw_serial, sizeof (hw_serial), "%lu", 973 (mode & FWRITE) ? (unsigned long)gethostid() : 0); 974 975 VERIFY((random_fd = open("/dev/random", O_RDONLY)) != -1); 976 VERIFY((urandom_fd = open("/dev/urandom", O_RDONLY)) != -1); 977 978 system_taskq_init(); 979 980#ifdef illumos 981 mutex_init(&cpu_lock, NULL, MUTEX_DEFAULT, NULL); 982#endif 983 984 spa_init(mode); 985 986 tsd_create(&rrw_tsd_key, rrw_tsd_destroy); 987} 988 989void 990kernel_fini(void) 991{ 992 spa_fini(); 993 994 system_taskq_fini(); 995 996 close(random_fd); 997 close(urandom_fd); 998 999 random_fd = -1; 1000 urandom_fd = -1; 1001} 1002 1003int 1004z_uncompress(void *dst, size_t *dstlen, const void *src, size_t srclen) 1005{ 1006 int ret; 1007 uLongf len = *dstlen; 1008 1009 if ((ret = uncompress(dst, &len, src, srclen)) == Z_OK) 1010 *dstlen = (size_t)len; 1011 1012 return (ret); 1013} 1014 1015int 1016z_compress_level(void *dst, size_t *dstlen, const void *src, size_t srclen, 1017 int level) 1018{ 1019 int ret; 1020 uLongf len = *dstlen; 1021 1022 if ((ret = compress2(dst, &len, src, srclen, level)) == Z_OK) 1023 *dstlen = (size_t)len; 1024 1025 return (ret); 1026} 1027 1028uid_t 1029crgetuid(cred_t *cr) 1030{ 1031 return (0); 1032} 1033 1034uid_t 1035crgetruid(cred_t *cr) 1036{ 1037 return (0); 1038} 1039 1040gid_t 1041crgetgid(cred_t *cr) 1042{ 1043 return (0); 1044} 1045 1046int 1047crgetngroups(cred_t *cr) 1048{ 1049 return (0); 1050} 1051 1052gid_t * 1053crgetgroups(cred_t *cr) 1054{ 1055 return (NULL); 1056} 1057 1058int 1059zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr) 1060{ 1061 return (0); 1062} 1063 1064int 1065zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr) 1066{ 1067 return (0); 1068} 1069 1070int 1071zfs_secpolicy_destroy_perms(const char *name, cred_t *cr) 1072{ 1073 return (0); 1074} 1075 1076ksiddomain_t * 1077ksid_lookupdomain(const char *dom) 1078{ 1079 ksiddomain_t *kd; 1080 1081 kd = umem_zalloc(sizeof (ksiddomain_t), UMEM_NOFAIL); 1082 kd->kd_name = spa_strdup(dom); 1083 return (kd); 1084} 1085 1086void 1087ksiddomain_rele(ksiddomain_t *ksid) 1088{ 1089 spa_strfree(ksid->kd_name); 1090 umem_free(ksid, sizeof (ksiddomain_t)); 1091} 1092 1093/* 1094 * Do not change the length of the returned string; it must be freed 1095 * with strfree(). 1096 */ 1097char * 1098kmem_asprintf(const char *fmt, ...) 1099{ 1100 int size; 1101 va_list adx; 1102 char *buf; 1103 1104 va_start(adx, fmt); 1105 size = vsnprintf(NULL, 0, fmt, adx) + 1; 1106 va_end(adx); 1107 1108 buf = kmem_alloc(size, KM_SLEEP); 1109 1110 va_start(adx, fmt); 1111 size = vsnprintf(buf, size, fmt, adx); 1112 va_end(adx); 1113 1114 return (buf); 1115} 1116 1117/* ARGSUSED */ 1118int 1119zfs_onexit_fd_hold(int fd, minor_t *minorp) 1120{ 1121 *minorp = 0; 1122 return (0); 1123} 1124 1125/* ARGSUSED */ 1126void 1127zfs_onexit_fd_rele(int fd) 1128{ 1129} 1130 1131/* ARGSUSED */ 1132int 1133zfs_onexit_add_cb(minor_t minor, void (*func)(void *), void *data, 1134 uint64_t *action_handle) 1135{ 1136 return (0); 1137} 1138 1139/* ARGSUSED */ 1140int 1141zfs_onexit_del_cb(minor_t minor, uint64_t action_handle, boolean_t fire) 1142{ 1143 return (0); 1144} 1145 1146/* ARGSUSED */ 1147int 1148zfs_onexit_cb_data(minor_t minor, uint64_t action_handle, void **data) 1149{ 1150 return (0); 1151} 1152 1153#ifdef __FreeBSD__ 1154/* ARGSUSED */ 1155int 1156zvol_create_minors(const char *name) 1157{ 1158 return (0); 1159} 1160#endif 1161 1162#ifdef illumos 1163void 1164bioinit(buf_t *bp) 1165{ 1166 bzero(bp, sizeof (buf_t)); 1167} 1168 1169void 1170biodone(buf_t *bp) 1171{ 1172 if (bp->b_iodone != NULL) { 1173 (*(bp->b_iodone))(bp); 1174 return; 1175 } 1176 ASSERT((bp->b_flags & B_DONE) == 0); 1177 bp->b_flags |= B_DONE; 1178} 1179 1180void 1181bioerror(buf_t *bp, int error) 1182{ 1183 ASSERT(bp != NULL); 1184 ASSERT(error >= 0); 1185 1186 if (error != 0) { 1187 bp->b_flags |= B_ERROR; 1188 } else { 1189 bp->b_flags &= ~B_ERROR; 1190 } 1191 bp->b_error = error; 1192} 1193 1194 1195int 1196geterror(struct buf *bp) 1197{ 1198 int error = 0; 1199 1200 if (bp->b_flags & B_ERROR) { 1201 error = bp->b_error; 1202 if (!error) 1203 error = EIO; 1204 } 1205 return (error); 1206} 1207#endif 1208