kernel.c revision 324587
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 char buf[512]; 731 (void) vsnprintf(buf, 512, fmt, adx); 732 assfail(buf, NULL, 0); 733} 734 735void 736panic(const char *fmt, ...) 737{ 738 va_list adx; 739 740 va_start(adx, fmt); 741 vpanic(fmt, adx); 742 va_end(adx); 743} 744 745void 746vcmn_err(int ce, const char *fmt, va_list adx) 747{ 748 if (ce == CE_PANIC) 749 vpanic(fmt, adx); 750 if (ce != CE_NOTE) { /* suppress noise in userland stress testing */ 751 (void) fprintf(stderr, "%s", ce_prefix[ce]); 752 (void) vfprintf(stderr, fmt, adx); 753 (void) fprintf(stderr, "%s", ce_suffix[ce]); 754 } 755} 756 757/*PRINTFLIKE2*/ 758void 759cmn_err(int ce, const char *fmt, ...) 760{ 761 va_list adx; 762 763 va_start(adx, fmt); 764 vcmn_err(ce, fmt, adx); 765 va_end(adx); 766} 767 768/* 769 * ========================================================================= 770 * kobj interfaces 771 * ========================================================================= 772 */ 773struct _buf * 774kobj_open_file(char *name) 775{ 776 struct _buf *file; 777 vnode_t *vp; 778 779 /* set vp as the _fd field of the file */ 780 if (vn_openat(name, UIO_SYSSPACE, FREAD, 0, &vp, 0, 0, rootdir, 781 -1) != 0) 782 return ((void *)-1UL); 783 784 file = umem_zalloc(sizeof (struct _buf), UMEM_NOFAIL); 785 file->_fd = (intptr_t)vp; 786 return (file); 787} 788 789int 790kobj_read_file(struct _buf *file, char *buf, unsigned size, unsigned off) 791{ 792 ssize_t resid; 793 794 vn_rdwr(UIO_READ, (vnode_t *)file->_fd, buf, size, (offset_t)off, 795 UIO_SYSSPACE, 0, 0, 0, &resid); 796 797 return (size - resid); 798} 799 800void 801kobj_close_file(struct _buf *file) 802{ 803 vn_close((vnode_t *)file->_fd, 0, NULL, NULL); 804 umem_free(file, sizeof (struct _buf)); 805} 806 807int 808kobj_get_filesize(struct _buf *file, uint64_t *size) 809{ 810 struct stat64 st; 811 vnode_t *vp = (vnode_t *)file->_fd; 812 813 if (fstat64(vp->v_fd, &st) == -1) { 814 vn_close(vp, 0, NULL, NULL); 815 return (errno); 816 } 817 *size = st.st_size; 818 return (0); 819} 820 821/* 822 * ========================================================================= 823 * misc routines 824 * ========================================================================= 825 */ 826 827void 828delay(clock_t ticks) 829{ 830 poll(0, 0, ticks * (1000 / hz)); 831} 832 833#if 0 834/* 835 * Find highest one bit set. 836 * Returns bit number + 1 of highest bit that is set, otherwise returns 0. 837 */ 838int 839highbit64(uint64_t i) 840{ 841 int h = 1; 842 843 if (i == 0) 844 return (0); 845 if (i & 0xffffffff00000000ULL) { 846 h += 32; i >>= 32; 847 } 848 if (i & 0xffff0000) { 849 h += 16; i >>= 16; 850 } 851 if (i & 0xff00) { 852 h += 8; i >>= 8; 853 } 854 if (i & 0xf0) { 855 h += 4; i >>= 4; 856 } 857 if (i & 0xc) { 858 h += 2; i >>= 2; 859 } 860 if (i & 0x2) { 861 h += 1; 862 } 863 return (h); 864} 865#endif 866 867static int random_fd = -1, urandom_fd = -1; 868 869static int 870random_get_bytes_common(uint8_t *ptr, size_t len, int fd) 871{ 872 size_t resid = len; 873 ssize_t bytes; 874 875 ASSERT(fd != -1); 876 877 while (resid != 0) { 878 bytes = read(fd, ptr, resid); 879 ASSERT3S(bytes, >=, 0); 880 ptr += bytes; 881 resid -= bytes; 882 } 883 884 return (0); 885} 886 887int 888random_get_bytes(uint8_t *ptr, size_t len) 889{ 890 return (random_get_bytes_common(ptr, len, random_fd)); 891} 892 893int 894random_get_pseudo_bytes(uint8_t *ptr, size_t len) 895{ 896 return (random_get_bytes_common(ptr, len, urandom_fd)); 897} 898 899int 900ddi_strtoul(const char *hw_serial, char **nptr, int base, unsigned long *result) 901{ 902 char *end; 903 904 *result = strtoul(hw_serial, &end, base); 905 if (*result == 0) 906 return (errno); 907 return (0); 908} 909 910int 911ddi_strtoull(const char *str, char **nptr, int base, u_longlong_t *result) 912{ 913 char *end; 914 915 *result = strtoull(str, &end, base); 916 if (*result == 0) 917 return (errno); 918 return (0); 919} 920 921#ifdef illumos 922/* ARGSUSED */ 923cyclic_id_t 924cyclic_add(cyc_handler_t *hdlr, cyc_time_t *when) 925{ 926 return (1); 927} 928 929/* ARGSUSED */ 930void 931cyclic_remove(cyclic_id_t id) 932{ 933} 934 935/* ARGSUSED */ 936int 937cyclic_reprogram(cyclic_id_t id, hrtime_t expiration) 938{ 939 return (1); 940} 941#endif 942 943/* 944 * ========================================================================= 945 * kernel emulation setup & teardown 946 * ========================================================================= 947 */ 948static int 949umem_out_of_memory(void) 950{ 951 char errmsg[] = "out of memory -- generating core dump\n"; 952 953 write(fileno(stderr), errmsg, sizeof (errmsg)); 954 abort(); 955 return (0); 956} 957 958void 959kernel_init(int mode) 960{ 961 extern uint_t rrw_tsd_key; 962 963 umem_nofail_callback(umem_out_of_memory); 964 965 physmem = sysconf(_SC_PHYS_PAGES); 966 967 dprintf("physmem = %llu pages (%.2f GB)\n", physmem, 968 (double)physmem * sysconf(_SC_PAGE_SIZE) / (1ULL << 30)); 969 970 (void) snprintf(hw_serial, sizeof (hw_serial), "%lu", 971 (mode & FWRITE) ? (unsigned long)gethostid() : 0); 972 973 VERIFY((random_fd = open("/dev/random", O_RDONLY)) != -1); 974 VERIFY((urandom_fd = open("/dev/urandom", O_RDONLY)) != -1); 975 976 system_taskq_init(); 977 978#ifdef illumos 979 mutex_init(&cpu_lock, NULL, MUTEX_DEFAULT, NULL); 980#endif 981 982 spa_init(mode); 983 984 tsd_create(&rrw_tsd_key, rrw_tsd_destroy); 985} 986 987void 988kernel_fini(void) 989{ 990 spa_fini(); 991 992 system_taskq_fini(); 993 994 close(random_fd); 995 close(urandom_fd); 996 997 random_fd = -1; 998 urandom_fd = -1; 999} 1000 1001int 1002z_uncompress(void *dst, size_t *dstlen, const void *src, size_t srclen) 1003{ 1004 int ret; 1005 uLongf len = *dstlen; 1006 1007 if ((ret = uncompress(dst, &len, src, srclen)) == Z_OK) 1008 *dstlen = (size_t)len; 1009 1010 return (ret); 1011} 1012 1013int 1014z_compress_level(void *dst, size_t *dstlen, const void *src, size_t srclen, 1015 int level) 1016{ 1017 int ret; 1018 uLongf len = *dstlen; 1019 1020 if ((ret = compress2(dst, &len, src, srclen, level)) == Z_OK) 1021 *dstlen = (size_t)len; 1022 1023 return (ret); 1024} 1025 1026uid_t 1027crgetuid(cred_t *cr) 1028{ 1029 return (0); 1030} 1031 1032uid_t 1033crgetruid(cred_t *cr) 1034{ 1035 return (0); 1036} 1037 1038gid_t 1039crgetgid(cred_t *cr) 1040{ 1041 return (0); 1042} 1043 1044int 1045crgetngroups(cred_t *cr) 1046{ 1047 return (0); 1048} 1049 1050gid_t * 1051crgetgroups(cred_t *cr) 1052{ 1053 return (NULL); 1054} 1055 1056int 1057zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr) 1058{ 1059 return (0); 1060} 1061 1062int 1063zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr) 1064{ 1065 return (0); 1066} 1067 1068int 1069zfs_secpolicy_destroy_perms(const char *name, cred_t *cr) 1070{ 1071 return (0); 1072} 1073 1074ksiddomain_t * 1075ksid_lookupdomain(const char *dom) 1076{ 1077 ksiddomain_t *kd; 1078 1079 kd = umem_zalloc(sizeof (ksiddomain_t), UMEM_NOFAIL); 1080 kd->kd_name = spa_strdup(dom); 1081 return (kd); 1082} 1083 1084void 1085ksiddomain_rele(ksiddomain_t *ksid) 1086{ 1087 spa_strfree(ksid->kd_name); 1088 umem_free(ksid, sizeof (ksiddomain_t)); 1089} 1090 1091/* 1092 * Do not change the length of the returned string; it must be freed 1093 * with strfree(). 1094 */ 1095char * 1096kmem_asprintf(const char *fmt, ...) 1097{ 1098 int size; 1099 va_list adx; 1100 char *buf; 1101 1102 va_start(adx, fmt); 1103 size = vsnprintf(NULL, 0, fmt, adx) + 1; 1104 va_end(adx); 1105 1106 buf = kmem_alloc(size, KM_SLEEP); 1107 1108 va_start(adx, fmt); 1109 size = vsnprintf(buf, size, fmt, adx); 1110 va_end(adx); 1111 1112 return (buf); 1113} 1114 1115/* ARGSUSED */ 1116int 1117zfs_onexit_fd_hold(int fd, minor_t *minorp) 1118{ 1119 *minorp = 0; 1120 return (0); 1121} 1122 1123/* ARGSUSED */ 1124void 1125zfs_onexit_fd_rele(int fd) 1126{ 1127} 1128 1129/* ARGSUSED */ 1130int 1131zfs_onexit_add_cb(minor_t minor, void (*func)(void *), void *data, 1132 uint64_t *action_handle) 1133{ 1134 return (0); 1135} 1136 1137/* ARGSUSED */ 1138int 1139zfs_onexit_del_cb(minor_t minor, uint64_t action_handle, boolean_t fire) 1140{ 1141 return (0); 1142} 1143 1144/* ARGSUSED */ 1145int 1146zfs_onexit_cb_data(minor_t minor, uint64_t action_handle, void **data) 1147{ 1148 return (0); 1149} 1150 1151#ifdef __FreeBSD__ 1152/* ARGSUSED */ 1153int 1154zvol_create_minors(const char *name) 1155{ 1156 return (0); 1157} 1158#endif 1159 1160#ifdef illumos 1161void 1162bioinit(buf_t *bp) 1163{ 1164 bzero(bp, sizeof (buf_t)); 1165} 1166 1167void 1168biodone(buf_t *bp) 1169{ 1170 if (bp->b_iodone != NULL) { 1171 (*(bp->b_iodone))(bp); 1172 return; 1173 } 1174 ASSERT((bp->b_flags & B_DONE) == 0); 1175 bp->b_flags |= B_DONE; 1176} 1177 1178void 1179bioerror(buf_t *bp, int error) 1180{ 1181 ASSERT(bp != NULL); 1182 ASSERT(error >= 0); 1183 1184 if (error != 0) { 1185 bp->b_flags |= B_ERROR; 1186 } else { 1187 bp->b_flags &= ~B_ERROR; 1188 } 1189 bp->b_error = error; 1190} 1191 1192 1193int 1194geterror(struct buf *bp) 1195{ 1196 int error = 0; 1197 1198 if (bp->b_flags & B_ERROR) { 1199 error = bp->b_error; 1200 if (!error) 1201 error = EIO; 1202 } 1203 return (error); 1204} 1205#endif 1206