kernel.c revision 299429
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 timestruc_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 ts.tv_sec = delta / NANOSEC; 380 ts.tv_nsec = delta % NANOSEC; 381 382 ASSERT(mutex_owner(mp) == curthread); 383 mp->m_owner = NULL; 384 error = pthread_cond_timedwait(cv, &mp->m_lock, &ts); 385 mp->m_owner = curthread; 386 387 if (error == ETIMEDOUT) 388 return (-1); 389 390 if (error == EINTR) 391 goto top; 392 393 ASSERT(error == 0); 394 395 return (1); 396} 397 398void 399cv_signal(kcondvar_t *cv) 400{ 401 VERIFY(cond_signal(cv) == 0); 402} 403 404void 405cv_broadcast(kcondvar_t *cv) 406{ 407 VERIFY(cond_broadcast(cv) == 0); 408} 409 410/* 411 * ========================================================================= 412 * vnode operations 413 * ========================================================================= 414 */ 415/* 416 * Note: for the xxxat() versions of these functions, we assume that the 417 * starting vp is always rootdir (which is true for spa_directory.c, the only 418 * ZFS consumer of these interfaces). We assert this is true, and then emulate 419 * them by adding '/' in front of the path. 420 */ 421 422/*ARGSUSED*/ 423int 424vn_open(char *path, int x1, int flags, int mode, vnode_t **vpp, int x2, int x3) 425{ 426 int fd; 427 int dump_fd; 428 vnode_t *vp; 429 int old_umask; 430 char realpath[MAXPATHLEN]; 431 struct stat64 st; 432 433 /* 434 * If we're accessing a real disk from userland, we need to use 435 * the character interface to avoid caching. This is particularly 436 * important if we're trying to look at a real in-kernel storage 437 * pool from userland, e.g. via zdb, because otherwise we won't 438 * see the changes occurring under the segmap cache. 439 * On the other hand, the stupid character device returns zero 440 * for its size. So -- gag -- we open the block device to get 441 * its size, and remember it for subsequent VOP_GETATTR(). 442 */ 443 if (strncmp(path, "/dev/", 5) == 0) { 444 char *dsk; 445 fd = open64(path, O_RDONLY); 446 if (fd == -1) 447 return (errno); 448 if (fstat64(fd, &st) == -1) { 449 close(fd); 450 return (errno); 451 } 452 close(fd); 453 (void) sprintf(realpath, "%s", path); 454 dsk = strstr(path, "/dsk/"); 455 if (dsk != NULL) 456 (void) sprintf(realpath + (dsk - path) + 1, "r%s", 457 dsk + 1); 458 } else { 459 (void) sprintf(realpath, "%s", path); 460 if (!(flags & FCREAT) && stat64(realpath, &st) == -1) 461 return (errno); 462 } 463 464 if (flags & FCREAT) 465 old_umask = umask(0); 466 467 /* 468 * The construct 'flags - FREAD' conveniently maps combinations of 469 * FREAD and FWRITE to the corresponding O_RDONLY, O_WRONLY, and O_RDWR. 470 */ 471 fd = open64(realpath, flags - FREAD, mode); 472 473 if (flags & FCREAT) 474 (void) umask(old_umask); 475 476 if (vn_dumpdir != NULL) { 477 char dumppath[MAXPATHLEN]; 478 (void) snprintf(dumppath, sizeof (dumppath), 479 "%s/%s", vn_dumpdir, basename(realpath)); 480 dump_fd = open64(dumppath, O_CREAT | O_WRONLY, 0666); 481 if (dump_fd == -1) 482 return (errno); 483 } else { 484 dump_fd = -1; 485 } 486 487 if (fd == -1) 488 return (errno); 489 490 if (fstat64(fd, &st) == -1) { 491 close(fd); 492 return (errno); 493 } 494 495 (void) fcntl(fd, F_SETFD, FD_CLOEXEC); 496 497 *vpp = vp = umem_zalloc(sizeof (vnode_t), UMEM_NOFAIL); 498 499 vp->v_fd = fd; 500 vp->v_size = st.st_size; 501 vp->v_path = spa_strdup(path); 502 vp->v_dump_fd = dump_fd; 503 504 return (0); 505} 506 507/*ARGSUSED*/ 508int 509vn_openat(char *path, int x1, int flags, int mode, vnode_t **vpp, int x2, 510 int x3, vnode_t *startvp, int fd) 511{ 512 char *realpath = umem_alloc(strlen(path) + 2, UMEM_NOFAIL); 513 int ret; 514 515 ASSERT(startvp == rootdir); 516 (void) sprintf(realpath, "/%s", path); 517 518 /* fd ignored for now, need if want to simulate nbmand support */ 519 ret = vn_open(realpath, x1, flags, mode, vpp, x2, x3); 520 521 umem_free(realpath, strlen(path) + 2); 522 523 return (ret); 524} 525 526/*ARGSUSED*/ 527int 528vn_rdwr(int uio, vnode_t *vp, void *addr, ssize_t len, offset_t offset, 529 int x1, int x2, rlim64_t x3, void *x4, ssize_t *residp) 530{ 531 ssize_t iolen, split; 532 533 if (uio == UIO_READ) { 534 iolen = pread64(vp->v_fd, addr, len, offset); 535 if (vp->v_dump_fd != -1) { 536 int status = 537 pwrite64(vp->v_dump_fd, addr, iolen, offset); 538 ASSERT(status != -1); 539 } 540 } else { 541 /* 542 * To simulate partial disk writes, we split writes into two 543 * system calls so that the process can be killed in between. 544 */ 545 int sectors = len >> SPA_MINBLOCKSHIFT; 546 split = (sectors > 0 ? rand() % sectors : 0) << 547 SPA_MINBLOCKSHIFT; 548 iolen = pwrite64(vp->v_fd, addr, split, offset); 549 iolen += pwrite64(vp->v_fd, (char *)addr + split, 550 len - split, offset + split); 551 } 552 553 if (iolen == -1) 554 return (errno); 555 if (residp) 556 *residp = len - iolen; 557 else if (iolen != len) 558 return (EIO); 559 return (0); 560} 561 562void 563vn_close(vnode_t *vp, int openflag, cred_t *cr, kthread_t *td) 564{ 565 close(vp->v_fd); 566 if (vp->v_dump_fd != -1) 567 close(vp->v_dump_fd); 568 spa_strfree(vp->v_path); 569 umem_free(vp, sizeof (vnode_t)); 570} 571 572/* 573 * At a minimum we need to update the size since vdev_reopen() 574 * will no longer call vn_openat(). 575 */ 576int 577fop_getattr(vnode_t *vp, vattr_t *vap) 578{ 579 struct stat64 st; 580 581 if (fstat64(vp->v_fd, &st) == -1) { 582 close(vp->v_fd); 583 return (errno); 584 } 585 586 vap->va_size = st.st_size; 587 return (0); 588} 589 590#ifdef ZFS_DEBUG 591 592/* 593 * ========================================================================= 594 * Figure out which debugging statements to print 595 * ========================================================================= 596 */ 597 598static char *dprintf_string; 599static int dprintf_print_all; 600 601int 602dprintf_find_string(const char *string) 603{ 604 char *tmp_str = dprintf_string; 605 int len = strlen(string); 606 607 /* 608 * Find out if this is a string we want to print. 609 * String format: file1.c,function_name1,file2.c,file3.c 610 */ 611 612 while (tmp_str != NULL) { 613 if (strncmp(tmp_str, string, len) == 0 && 614 (tmp_str[len] == ',' || tmp_str[len] == '\0')) 615 return (1); 616 tmp_str = strchr(tmp_str, ','); 617 if (tmp_str != NULL) 618 tmp_str++; /* Get rid of , */ 619 } 620 return (0); 621} 622 623void 624dprintf_setup(int *argc, char **argv) 625{ 626 int i, j; 627 628 /* 629 * Debugging can be specified two ways: by setting the 630 * environment variable ZFS_DEBUG, or by including a 631 * "debug=..." argument on the command line. The command 632 * line setting overrides the environment variable. 633 */ 634 635 for (i = 1; i < *argc; i++) { 636 int len = strlen("debug="); 637 /* First look for a command line argument */ 638 if (strncmp("debug=", argv[i], len) == 0) { 639 dprintf_string = argv[i] + len; 640 /* Remove from args */ 641 for (j = i; j < *argc; j++) 642 argv[j] = argv[j+1]; 643 argv[j] = NULL; 644 (*argc)--; 645 } 646 } 647 648 if (dprintf_string == NULL) { 649 /* Look for ZFS_DEBUG environment variable */ 650 dprintf_string = getenv("ZFS_DEBUG"); 651 } 652 653 /* 654 * Are we just turning on all debugging? 655 */ 656 if (dprintf_find_string("on")) 657 dprintf_print_all = 1; 658} 659 660int 661sysctl_handle_64(SYSCTL_HANDLER_ARGS) 662{ 663 return (0); 664} 665 666/* 667 * ========================================================================= 668 * debug printfs 669 * ========================================================================= 670 */ 671void 672__dprintf(const char *file, const char *func, int line, const char *fmt, ...) 673{ 674 const char *newfile; 675 va_list adx; 676 677 /* 678 * Get rid of annoying "../common/" prefix to filename. 679 */ 680 newfile = strrchr(file, '/'); 681 if (newfile != NULL) { 682 newfile = newfile + 1; /* Get rid of leading / */ 683 } else { 684 newfile = file; 685 } 686 687 if (dprintf_print_all || 688 dprintf_find_string(newfile) || 689 dprintf_find_string(func)) { 690 /* Print out just the function name if requested */ 691 flockfile(stdout); 692 if (dprintf_find_string("pid")) 693 (void) printf("%d ", getpid()); 694 if (dprintf_find_string("tid")) 695 (void) printf("%ul ", thr_self()); 696#if 0 697 if (dprintf_find_string("cpu")) 698 (void) printf("%u ", getcpuid()); 699#endif 700 if (dprintf_find_string("time")) 701 (void) printf("%llu ", gethrtime()); 702 if (dprintf_find_string("long")) 703 (void) printf("%s, line %d: ", newfile, line); 704 (void) printf("%s: ", func); 705 va_start(adx, fmt); 706 (void) vprintf(fmt, adx); 707 va_end(adx); 708 funlockfile(stdout); 709 } 710} 711 712#endif /* ZFS_DEBUG */ 713 714/* 715 * ========================================================================= 716 * cmn_err() and panic() 717 * ========================================================================= 718 */ 719static char ce_prefix[CE_IGNORE][10] = { "", "NOTICE: ", "WARNING: ", "" }; 720static char ce_suffix[CE_IGNORE][2] = { "", "\n", "\n", "" }; 721 722void 723vpanic(const char *fmt, va_list adx) 724{ 725 (void) fprintf(stderr, "error: "); 726 (void) vfprintf(stderr, fmt, adx); 727 (void) fprintf(stderr, "\n"); 728 729 abort(); /* think of it as a "user-level crash dump" */ 730} 731 732void 733panic(const char *fmt, ...) 734{ 735 va_list adx; 736 737 va_start(adx, fmt); 738 vpanic(fmt, adx); 739 va_end(adx); 740} 741 742void 743vcmn_err(int ce, const char *fmt, va_list adx) 744{ 745 if (ce == CE_PANIC) 746 vpanic(fmt, adx); 747 if (ce != CE_NOTE) { /* suppress noise in userland stress testing */ 748 (void) fprintf(stderr, "%s", ce_prefix[ce]); 749 (void) vfprintf(stderr, fmt, adx); 750 (void) fprintf(stderr, "%s", ce_suffix[ce]); 751 } 752} 753 754/*PRINTFLIKE2*/ 755void 756cmn_err(int ce, const char *fmt, ...) 757{ 758 va_list adx; 759 760 va_start(adx, fmt); 761 vcmn_err(ce, fmt, adx); 762 va_end(adx); 763} 764 765/* 766 * ========================================================================= 767 * kobj interfaces 768 * ========================================================================= 769 */ 770struct _buf * 771kobj_open_file(char *name) 772{ 773 struct _buf *file; 774 vnode_t *vp; 775 776 /* set vp as the _fd field of the file */ 777 if (vn_openat(name, UIO_SYSSPACE, FREAD, 0, &vp, 0, 0, rootdir, 778 -1) != 0) 779 return ((void *)-1UL); 780 781 file = umem_zalloc(sizeof (struct _buf), UMEM_NOFAIL); 782 file->_fd = (intptr_t)vp; 783 return (file); 784} 785 786int 787kobj_read_file(struct _buf *file, char *buf, unsigned size, unsigned off) 788{ 789 ssize_t resid; 790 791 vn_rdwr(UIO_READ, (vnode_t *)file->_fd, buf, size, (offset_t)off, 792 UIO_SYSSPACE, 0, 0, 0, &resid); 793 794 return (size - resid); 795} 796 797void 798kobj_close_file(struct _buf *file) 799{ 800 vn_close((vnode_t *)file->_fd, 0, NULL, NULL); 801 umem_free(file, sizeof (struct _buf)); 802} 803 804int 805kobj_get_filesize(struct _buf *file, uint64_t *size) 806{ 807 struct stat64 st; 808 vnode_t *vp = (vnode_t *)file->_fd; 809 810 if (fstat64(vp->v_fd, &st) == -1) { 811 vn_close(vp, 0, NULL, NULL); 812 return (errno); 813 } 814 *size = st.st_size; 815 return (0); 816} 817 818/* 819 * ========================================================================= 820 * misc routines 821 * ========================================================================= 822 */ 823 824void 825delay(clock_t ticks) 826{ 827 poll(0, 0, ticks * (1000 / hz)); 828} 829 830#if 0 831/* 832 * Find highest one bit set. 833 * Returns bit number + 1 of highest bit that is set, otherwise returns 0. 834 */ 835int 836highbit64(uint64_t i) 837{ 838 int h = 1; 839 840 if (i == 0) 841 return (0); 842 if (i & 0xffffffff00000000ULL) { 843 h += 32; i >>= 32; 844 } 845 if (i & 0xffff0000) { 846 h += 16; i >>= 16; 847 } 848 if (i & 0xff00) { 849 h += 8; i >>= 8; 850 } 851 if (i & 0xf0) { 852 h += 4; i >>= 4; 853 } 854 if (i & 0xc) { 855 h += 2; i >>= 2; 856 } 857 if (i & 0x2) { 858 h += 1; 859 } 860 return (h); 861} 862#endif 863 864static int random_fd = -1, urandom_fd = -1; 865 866static int 867random_get_bytes_common(uint8_t *ptr, size_t len, int fd) 868{ 869 size_t resid = len; 870 ssize_t bytes; 871 872 ASSERT(fd != -1); 873 874 while (resid != 0) { 875 bytes = read(fd, ptr, resid); 876 ASSERT3S(bytes, >=, 0); 877 ptr += bytes; 878 resid -= bytes; 879 } 880 881 return (0); 882} 883 884int 885random_get_bytes(uint8_t *ptr, size_t len) 886{ 887 return (random_get_bytes_common(ptr, len, random_fd)); 888} 889 890int 891random_get_pseudo_bytes(uint8_t *ptr, size_t len) 892{ 893 return (random_get_bytes_common(ptr, len, urandom_fd)); 894} 895 896int 897ddi_strtoul(const char *hw_serial, char **nptr, int base, unsigned long *result) 898{ 899 char *end; 900 901 *result = strtoul(hw_serial, &end, base); 902 if (*result == 0) 903 return (errno); 904 return (0); 905} 906 907int 908ddi_strtoull(const char *str, char **nptr, int base, u_longlong_t *result) 909{ 910 char *end; 911 912 *result = strtoull(str, &end, base); 913 if (*result == 0) 914 return (errno); 915 return (0); 916} 917 918#ifdef illumos 919/* ARGSUSED */ 920cyclic_id_t 921cyclic_add(cyc_handler_t *hdlr, cyc_time_t *when) 922{ 923 return (1); 924} 925 926/* ARGSUSED */ 927void 928cyclic_remove(cyclic_id_t id) 929{ 930} 931 932/* ARGSUSED */ 933int 934cyclic_reprogram(cyclic_id_t id, hrtime_t expiration) 935{ 936 return (1); 937} 938#endif 939 940/* 941 * ========================================================================= 942 * kernel emulation setup & teardown 943 * ========================================================================= 944 */ 945static int 946umem_out_of_memory(void) 947{ 948 char errmsg[] = "out of memory -- generating core dump\n"; 949 950 write(fileno(stderr), errmsg, sizeof (errmsg)); 951 abort(); 952 return (0); 953} 954 955void 956kernel_init(int mode) 957{ 958 extern uint_t rrw_tsd_key; 959 960 umem_nofail_callback(umem_out_of_memory); 961 962 physmem = sysconf(_SC_PHYS_PAGES); 963 964 dprintf("physmem = %llu pages (%.2f GB)\n", physmem, 965 (double)physmem * sysconf(_SC_PAGE_SIZE) / (1ULL << 30)); 966 967 (void) snprintf(hw_serial, sizeof (hw_serial), "%lu", 968 (mode & FWRITE) ? (unsigned long)gethostid() : 0); 969 970 VERIFY((random_fd = open("/dev/random", O_RDONLY)) != -1); 971 VERIFY((urandom_fd = open("/dev/urandom", O_RDONLY)) != -1); 972 973 system_taskq_init(); 974 975#ifdef illumos 976 mutex_init(&cpu_lock, NULL, MUTEX_DEFAULT, NULL); 977#endif 978 979 spa_init(mode); 980 981 tsd_create(&rrw_tsd_key, rrw_tsd_destroy); 982} 983 984void 985kernel_fini(void) 986{ 987 spa_fini(); 988 989 system_taskq_fini(); 990 991 close(random_fd); 992 close(urandom_fd); 993 994 random_fd = -1; 995 urandom_fd = -1; 996} 997 998int 999z_uncompress(void *dst, size_t *dstlen, const void *src, size_t srclen) 1000{ 1001 int ret; 1002 uLongf len = *dstlen; 1003 1004 if ((ret = uncompress(dst, &len, src, srclen)) == Z_OK) 1005 *dstlen = (size_t)len; 1006 1007 return (ret); 1008} 1009 1010int 1011z_compress_level(void *dst, size_t *dstlen, const void *src, size_t srclen, 1012 int level) 1013{ 1014 int ret; 1015 uLongf len = *dstlen; 1016 1017 if ((ret = compress2(dst, &len, src, srclen, level)) == Z_OK) 1018 *dstlen = (size_t)len; 1019 1020 return (ret); 1021} 1022 1023uid_t 1024crgetuid(cred_t *cr) 1025{ 1026 return (0); 1027} 1028 1029uid_t 1030crgetruid(cred_t *cr) 1031{ 1032 return (0); 1033} 1034 1035gid_t 1036crgetgid(cred_t *cr) 1037{ 1038 return (0); 1039} 1040 1041int 1042crgetngroups(cred_t *cr) 1043{ 1044 return (0); 1045} 1046 1047gid_t * 1048crgetgroups(cred_t *cr) 1049{ 1050 return (NULL); 1051} 1052 1053int 1054zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr) 1055{ 1056 return (0); 1057} 1058 1059int 1060zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr) 1061{ 1062 return (0); 1063} 1064 1065int 1066zfs_secpolicy_destroy_perms(const char *name, cred_t *cr) 1067{ 1068 return (0); 1069} 1070 1071ksiddomain_t * 1072ksid_lookupdomain(const char *dom) 1073{ 1074 ksiddomain_t *kd; 1075 1076 kd = umem_zalloc(sizeof (ksiddomain_t), UMEM_NOFAIL); 1077 kd->kd_name = spa_strdup(dom); 1078 return (kd); 1079} 1080 1081void 1082ksiddomain_rele(ksiddomain_t *ksid) 1083{ 1084 spa_strfree(ksid->kd_name); 1085 umem_free(ksid, sizeof (ksiddomain_t)); 1086} 1087 1088/* 1089 * Do not change the length of the returned string; it must be freed 1090 * with strfree(). 1091 */ 1092char * 1093kmem_asprintf(const char *fmt, ...) 1094{ 1095 int size; 1096 va_list adx; 1097 char *buf; 1098 1099 va_start(adx, fmt); 1100 size = vsnprintf(NULL, 0, fmt, adx) + 1; 1101 va_end(adx); 1102 1103 buf = kmem_alloc(size, KM_SLEEP); 1104 1105 va_start(adx, fmt); 1106 size = vsnprintf(buf, size, fmt, adx); 1107 va_end(adx); 1108 1109 return (buf); 1110} 1111 1112/* ARGSUSED */ 1113int 1114zfs_onexit_fd_hold(int fd, minor_t *minorp) 1115{ 1116 *minorp = 0; 1117 return (0); 1118} 1119 1120/* ARGSUSED */ 1121void 1122zfs_onexit_fd_rele(int fd) 1123{ 1124} 1125 1126/* ARGSUSED */ 1127int 1128zfs_onexit_add_cb(minor_t minor, void (*func)(void *), void *data, 1129 uint64_t *action_handle) 1130{ 1131 return (0); 1132} 1133 1134/* ARGSUSED */ 1135int 1136zfs_onexit_del_cb(minor_t minor, uint64_t action_handle, boolean_t fire) 1137{ 1138 return (0); 1139} 1140 1141/* ARGSUSED */ 1142int 1143zfs_onexit_cb_data(minor_t minor, uint64_t action_handle, void **data) 1144{ 1145 return (0); 1146} 1147 1148#ifdef __FreeBSD__ 1149/* ARGSUSED */ 1150int 1151zvol_create_minors(const char *name) 1152{ 1153 return (0); 1154} 1155#endif 1156 1157#ifdef illumos 1158void 1159bioinit(buf_t *bp) 1160{ 1161 bzero(bp, sizeof (buf_t)); 1162} 1163 1164void 1165biodone(buf_t *bp) 1166{ 1167 if (bp->b_iodone != NULL) { 1168 (*(bp->b_iodone))(bp); 1169 return; 1170 } 1171 ASSERT((bp->b_flags & B_DONE) == 0); 1172 bp->b_flags |= B_DONE; 1173} 1174 1175void 1176bioerror(buf_t *bp, int error) 1177{ 1178 ASSERT(bp != NULL); 1179 ASSERT(error >= 0); 1180 1181 if (error != 0) { 1182 bp->b_flags |= B_ERROR; 1183 } else { 1184 bp->b_flags &= ~B_ERROR; 1185 } 1186 bp->b_error = error; 1187} 1188 1189 1190int 1191geterror(struct buf *bp) 1192{ 1193 int error = 0; 1194 1195 if (bp->b_flags & B_ERROR) { 1196 error = bp->b_error; 1197 if (!error) 1198 error = EIO; 1199 } 1200 return (error); 1201} 1202#endif 1203