kernel.c revision 289805
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, 2014 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); 370 371top: 372 delta = tim - gethrtime(); 373 if (delta <= 0) 374 return (-1); 375 376 ts.tv_sec = delta / NANOSEC; 377 ts.tv_nsec = delta % NANOSEC; 378 379 ASSERT(mutex_owner(mp) == curthread); 380 mp->m_owner = NULL; 381 error = pthread_cond_timedwait(cv, &mp->m_lock, &ts); 382 mp->m_owner = curthread; 383 384 if (error == ETIMEDOUT) 385 return (-1); 386 387 if (error == EINTR) 388 goto top; 389 390 ASSERT(error == 0); 391 392 return (1); 393} 394 395void 396cv_signal(kcondvar_t *cv) 397{ 398 VERIFY(cond_signal(cv) == 0); 399} 400 401void 402cv_broadcast(kcondvar_t *cv) 403{ 404 VERIFY(cond_broadcast(cv) == 0); 405} 406 407/* 408 * ========================================================================= 409 * vnode operations 410 * ========================================================================= 411 */ 412/* 413 * Note: for the xxxat() versions of these functions, we assume that the 414 * starting vp is always rootdir (which is true for spa_directory.c, the only 415 * ZFS consumer of these interfaces). We assert this is true, and then emulate 416 * them by adding '/' in front of the path. 417 */ 418 419/*ARGSUSED*/ 420int 421vn_open(char *path, int x1, int flags, int mode, vnode_t **vpp, int x2, int x3) 422{ 423 int fd; 424 int dump_fd; 425 vnode_t *vp; 426 int old_umask; 427 char realpath[MAXPATHLEN]; 428 struct stat64 st; 429 430 /* 431 * If we're accessing a real disk from userland, we need to use 432 * the character interface to avoid caching. This is particularly 433 * important if we're trying to look at a real in-kernel storage 434 * pool from userland, e.g. via zdb, because otherwise we won't 435 * see the changes occurring under the segmap cache. 436 * On the other hand, the stupid character device returns zero 437 * for its size. So -- gag -- we open the block device to get 438 * its size, and remember it for subsequent VOP_GETATTR(). 439 */ 440 if (strncmp(path, "/dev/", 5) == 0) { 441 char *dsk; 442 fd = open64(path, O_RDONLY); 443 if (fd == -1) 444 return (errno); 445 if (fstat64(fd, &st) == -1) { 446 close(fd); 447 return (errno); 448 } 449 close(fd); 450 (void) sprintf(realpath, "%s", path); 451 dsk = strstr(path, "/dsk/"); 452 if (dsk != NULL) 453 (void) sprintf(realpath + (dsk - path) + 1, "r%s", 454 dsk + 1); 455 } else { 456 (void) sprintf(realpath, "%s", path); 457 if (!(flags & FCREAT) && stat64(realpath, &st) == -1) 458 return (errno); 459 } 460 461 if (flags & FCREAT) 462 old_umask = umask(0); 463 464 /* 465 * The construct 'flags - FREAD' conveniently maps combinations of 466 * FREAD and FWRITE to the corresponding O_RDONLY, O_WRONLY, and O_RDWR. 467 */ 468 fd = open64(realpath, flags - FREAD, mode); 469 470 if (flags & FCREAT) 471 (void) umask(old_umask); 472 473 if (vn_dumpdir != NULL) { 474 char dumppath[MAXPATHLEN]; 475 (void) snprintf(dumppath, sizeof (dumppath), 476 "%s/%s", vn_dumpdir, basename(realpath)); 477 dump_fd = open64(dumppath, O_CREAT | O_WRONLY, 0666); 478 if (dump_fd == -1) 479 return (errno); 480 } else { 481 dump_fd = -1; 482 } 483 484 if (fd == -1) 485 return (errno); 486 487 if (fstat64(fd, &st) == -1) { 488 close(fd); 489 return (errno); 490 } 491 492 (void) fcntl(fd, F_SETFD, FD_CLOEXEC); 493 494 *vpp = vp = umem_zalloc(sizeof (vnode_t), UMEM_NOFAIL); 495 496 vp->v_fd = fd; 497 vp->v_size = st.st_size; 498 vp->v_path = spa_strdup(path); 499 vp->v_dump_fd = dump_fd; 500 501 return (0); 502} 503 504/*ARGSUSED*/ 505int 506vn_openat(char *path, int x1, int flags, int mode, vnode_t **vpp, int x2, 507 int x3, vnode_t *startvp, int fd) 508{ 509 char *realpath = umem_alloc(strlen(path) + 2, UMEM_NOFAIL); 510 int ret; 511 512 ASSERT(startvp == rootdir); 513 (void) sprintf(realpath, "/%s", path); 514 515 /* fd ignored for now, need if want to simulate nbmand support */ 516 ret = vn_open(realpath, x1, flags, mode, vpp, x2, x3); 517 518 umem_free(realpath, strlen(path) + 2); 519 520 return (ret); 521} 522 523/*ARGSUSED*/ 524int 525vn_rdwr(int uio, vnode_t *vp, void *addr, ssize_t len, offset_t offset, 526 int x1, int x2, rlim64_t x3, void *x4, ssize_t *residp) 527{ 528 ssize_t iolen, split; 529 530 if (uio == UIO_READ) { 531 iolen = pread64(vp->v_fd, addr, len, offset); 532 if (vp->v_dump_fd != -1) { 533 int status = 534 pwrite64(vp->v_dump_fd, addr, iolen, offset); 535 ASSERT(status != -1); 536 } 537 } else { 538 /* 539 * To simulate partial disk writes, we split writes into two 540 * system calls so that the process can be killed in between. 541 */ 542 int sectors = len >> SPA_MINBLOCKSHIFT; 543 split = (sectors > 0 ? rand() % sectors : 0) << 544 SPA_MINBLOCKSHIFT; 545 iolen = pwrite64(vp->v_fd, addr, split, offset); 546 iolen += pwrite64(vp->v_fd, (char *)addr + split, 547 len - split, offset + split); 548 } 549 550 if (iolen == -1) 551 return (errno); 552 if (residp) 553 *residp = len - iolen; 554 else if (iolen != len) 555 return (EIO); 556 return (0); 557} 558 559void 560vn_close(vnode_t *vp, int openflag, cred_t *cr, kthread_t *td) 561{ 562 close(vp->v_fd); 563 if (vp->v_dump_fd != -1) 564 close(vp->v_dump_fd); 565 spa_strfree(vp->v_path); 566 umem_free(vp, sizeof (vnode_t)); 567} 568 569/* 570 * At a minimum we need to update the size since vdev_reopen() 571 * will no longer call vn_openat(). 572 */ 573int 574fop_getattr(vnode_t *vp, vattr_t *vap) 575{ 576 struct stat64 st; 577 578 if (fstat64(vp->v_fd, &st) == -1) { 579 close(vp->v_fd); 580 return (errno); 581 } 582 583 vap->va_size = st.st_size; 584 return (0); 585} 586 587#ifdef ZFS_DEBUG 588 589/* 590 * ========================================================================= 591 * Figure out which debugging statements to print 592 * ========================================================================= 593 */ 594 595static char *dprintf_string; 596static int dprintf_print_all; 597 598int 599dprintf_find_string(const char *string) 600{ 601 char *tmp_str = dprintf_string; 602 int len = strlen(string); 603 604 /* 605 * Find out if this is a string we want to print. 606 * String format: file1.c,function_name1,file2.c,file3.c 607 */ 608 609 while (tmp_str != NULL) { 610 if (strncmp(tmp_str, string, len) == 0 && 611 (tmp_str[len] == ',' || tmp_str[len] == '\0')) 612 return (1); 613 tmp_str = strchr(tmp_str, ','); 614 if (tmp_str != NULL) 615 tmp_str++; /* Get rid of , */ 616 } 617 return (0); 618} 619 620void 621dprintf_setup(int *argc, char **argv) 622{ 623 int i, j; 624 625 /* 626 * Debugging can be specified two ways: by setting the 627 * environment variable ZFS_DEBUG, or by including a 628 * "debug=..." argument on the command line. The command 629 * line setting overrides the environment variable. 630 */ 631 632 for (i = 1; i < *argc; i++) { 633 int len = strlen("debug="); 634 /* First look for a command line argument */ 635 if (strncmp("debug=", argv[i], len) == 0) { 636 dprintf_string = argv[i] + len; 637 /* Remove from args */ 638 for (j = i; j < *argc; j++) 639 argv[j] = argv[j+1]; 640 argv[j] = NULL; 641 (*argc)--; 642 } 643 } 644 645 if (dprintf_string == NULL) { 646 /* Look for ZFS_DEBUG environment variable */ 647 dprintf_string = getenv("ZFS_DEBUG"); 648 } 649 650 /* 651 * Are we just turning on all debugging? 652 */ 653 if (dprintf_find_string("on")) 654 dprintf_print_all = 1; 655} 656 657int 658sysctl_handle_64(SYSCTL_HANDLER_ARGS) 659{ 660 return (0); 661} 662 663/* 664 * ========================================================================= 665 * debug printfs 666 * ========================================================================= 667 */ 668void 669__dprintf(const char *file, const char *func, int line, const char *fmt, ...) 670{ 671 const char *newfile; 672 va_list adx; 673 674 /* 675 * Get rid of annoying "../common/" prefix to filename. 676 */ 677 newfile = strrchr(file, '/'); 678 if (newfile != NULL) { 679 newfile = newfile + 1; /* Get rid of leading / */ 680 } else { 681 newfile = file; 682 } 683 684 if (dprintf_print_all || 685 dprintf_find_string(newfile) || 686 dprintf_find_string(func)) { 687 /* Print out just the function name if requested */ 688 flockfile(stdout); 689 if (dprintf_find_string("pid")) 690 (void) printf("%d ", getpid()); 691 if (dprintf_find_string("tid")) 692 (void) printf("%ul ", thr_self()); 693#if 0 694 if (dprintf_find_string("cpu")) 695 (void) printf("%u ", getcpuid()); 696#endif 697 if (dprintf_find_string("time")) 698 (void) printf("%llu ", gethrtime()); 699 if (dprintf_find_string("long")) 700 (void) printf("%s, line %d: ", newfile, line); 701 (void) printf("%s: ", func); 702 va_start(adx, fmt); 703 (void) vprintf(fmt, adx); 704 va_end(adx); 705 funlockfile(stdout); 706 } 707} 708 709#endif /* ZFS_DEBUG */ 710 711/* 712 * ========================================================================= 713 * cmn_err() and panic() 714 * ========================================================================= 715 */ 716static char ce_prefix[CE_IGNORE][10] = { "", "NOTICE: ", "WARNING: ", "" }; 717static char ce_suffix[CE_IGNORE][2] = { "", "\n", "\n", "" }; 718 719void 720vpanic(const char *fmt, va_list adx) 721{ 722 (void) fprintf(stderr, "error: "); 723 (void) vfprintf(stderr, fmt, adx); 724 (void) fprintf(stderr, "\n"); 725 726 abort(); /* think of it as a "user-level crash dump" */ 727} 728 729void 730panic(const char *fmt, ...) 731{ 732 va_list adx; 733 734 va_start(adx, fmt); 735 vpanic(fmt, adx); 736 va_end(adx); 737} 738 739void 740vcmn_err(int ce, const char *fmt, va_list adx) 741{ 742 if (ce == CE_PANIC) 743 vpanic(fmt, adx); 744 if (ce != CE_NOTE) { /* suppress noise in userland stress testing */ 745 (void) fprintf(stderr, "%s", ce_prefix[ce]); 746 (void) vfprintf(stderr, fmt, adx); 747 (void) fprintf(stderr, "%s", ce_suffix[ce]); 748 } 749} 750 751/*PRINTFLIKE2*/ 752void 753cmn_err(int ce, const char *fmt, ...) 754{ 755 va_list adx; 756 757 va_start(adx, fmt); 758 vcmn_err(ce, fmt, adx); 759 va_end(adx); 760} 761 762/* 763 * ========================================================================= 764 * kobj interfaces 765 * ========================================================================= 766 */ 767struct _buf * 768kobj_open_file(char *name) 769{ 770 struct _buf *file; 771 vnode_t *vp; 772 773 /* set vp as the _fd field of the file */ 774 if (vn_openat(name, UIO_SYSSPACE, FREAD, 0, &vp, 0, 0, rootdir, 775 -1) != 0) 776 return ((void *)-1UL); 777 778 file = umem_zalloc(sizeof (struct _buf), UMEM_NOFAIL); 779 file->_fd = (intptr_t)vp; 780 return (file); 781} 782 783int 784kobj_read_file(struct _buf *file, char *buf, unsigned size, unsigned off) 785{ 786 ssize_t resid; 787 788 vn_rdwr(UIO_READ, (vnode_t *)file->_fd, buf, size, (offset_t)off, 789 UIO_SYSSPACE, 0, 0, 0, &resid); 790 791 return (size - resid); 792} 793 794void 795kobj_close_file(struct _buf *file) 796{ 797 vn_close((vnode_t *)file->_fd, 0, NULL, NULL); 798 umem_free(file, sizeof (struct _buf)); 799} 800 801int 802kobj_get_filesize(struct _buf *file, uint64_t *size) 803{ 804 struct stat64 st; 805 vnode_t *vp = (vnode_t *)file->_fd; 806 807 if (fstat64(vp->v_fd, &st) == -1) { 808 vn_close(vp, 0, NULL, NULL); 809 return (errno); 810 } 811 *size = st.st_size; 812 return (0); 813} 814 815/* 816 * ========================================================================= 817 * misc routines 818 * ========================================================================= 819 */ 820 821void 822delay(clock_t ticks) 823{ 824 poll(0, 0, ticks * (1000 / hz)); 825} 826 827#if 0 828/* 829 * Find highest one bit set. 830 * Returns bit number + 1 of highest bit that is set, otherwise returns 0. 831 */ 832int 833highbit64(uint64_t i) 834{ 835 int h = 1; 836 837 if (i == 0) 838 return (0); 839 if (i & 0xffffffff00000000ULL) { 840 h += 32; i >>= 32; 841 } 842 if (i & 0xffff0000) { 843 h += 16; i >>= 16; 844 } 845 if (i & 0xff00) { 846 h += 8; i >>= 8; 847 } 848 if (i & 0xf0) { 849 h += 4; i >>= 4; 850 } 851 if (i & 0xc) { 852 h += 2; i >>= 2; 853 } 854 if (i & 0x2) { 855 h += 1; 856 } 857 return (h); 858} 859#endif 860 861static int random_fd = -1, urandom_fd = -1; 862 863static int 864random_get_bytes_common(uint8_t *ptr, size_t len, int fd) 865{ 866 size_t resid = len; 867 ssize_t bytes; 868 869 ASSERT(fd != -1); 870 871 while (resid != 0) { 872 bytes = read(fd, ptr, resid); 873 ASSERT3S(bytes, >=, 0); 874 ptr += bytes; 875 resid -= bytes; 876 } 877 878 return (0); 879} 880 881int 882random_get_bytes(uint8_t *ptr, size_t len) 883{ 884 return (random_get_bytes_common(ptr, len, random_fd)); 885} 886 887int 888random_get_pseudo_bytes(uint8_t *ptr, size_t len) 889{ 890 return (random_get_bytes_common(ptr, len, urandom_fd)); 891} 892 893int 894ddi_strtoul(const char *hw_serial, char **nptr, int base, unsigned long *result) 895{ 896 char *end; 897 898 *result = strtoul(hw_serial, &end, base); 899 if (*result == 0) 900 return (errno); 901 return (0); 902} 903 904int 905ddi_strtoull(const char *str, char **nptr, int base, u_longlong_t *result) 906{ 907 char *end; 908 909 *result = strtoull(str, &end, base); 910 if (*result == 0) 911 return (errno); 912 return (0); 913} 914 915#ifdef illumos 916/* ARGSUSED */ 917cyclic_id_t 918cyclic_add(cyc_handler_t *hdlr, cyc_time_t *when) 919{ 920 return (1); 921} 922 923/* ARGSUSED */ 924void 925cyclic_remove(cyclic_id_t id) 926{ 927} 928 929/* ARGSUSED */ 930int 931cyclic_reprogram(cyclic_id_t id, hrtime_t expiration) 932{ 933 return (1); 934} 935#endif 936 937/* 938 * ========================================================================= 939 * kernel emulation setup & teardown 940 * ========================================================================= 941 */ 942static int 943umem_out_of_memory(void) 944{ 945 char errmsg[] = "out of memory -- generating core dump\n"; 946 947 write(fileno(stderr), errmsg, sizeof (errmsg)); 948 abort(); 949 return (0); 950} 951 952void 953kernel_init(int mode) 954{ 955 extern uint_t rrw_tsd_key; 956 957 umem_nofail_callback(umem_out_of_memory); 958 959 physmem = sysconf(_SC_PHYS_PAGES); 960 961 dprintf("physmem = %llu pages (%.2f GB)\n", physmem, 962 (double)physmem * sysconf(_SC_PAGE_SIZE) / (1ULL << 30)); 963 964 (void) snprintf(hw_serial, sizeof (hw_serial), "%lu", 965 (mode & FWRITE) ? (unsigned long)gethostid() : 0); 966 967 VERIFY((random_fd = open("/dev/random", O_RDONLY)) != -1); 968 VERIFY((urandom_fd = open("/dev/urandom", O_RDONLY)) != -1); 969 970 system_taskq_init(); 971 972#ifdef illumos 973 mutex_init(&cpu_lock, NULL, MUTEX_DEFAULT, NULL); 974#endif 975 976 spa_init(mode); 977 978 tsd_create(&rrw_tsd_key, rrw_tsd_destroy); 979} 980 981void 982kernel_fini(void) 983{ 984 spa_fini(); 985 986 system_taskq_fini(); 987 988 close(random_fd); 989 close(urandom_fd); 990 991 random_fd = -1; 992 urandom_fd = -1; 993} 994 995int 996z_uncompress(void *dst, size_t *dstlen, const void *src, size_t srclen) 997{ 998 int ret; 999 uLongf len = *dstlen; 1000 1001 if ((ret = uncompress(dst, &len, src, srclen)) == Z_OK) 1002 *dstlen = (size_t)len; 1003 1004 return (ret); 1005} 1006 1007int 1008z_compress_level(void *dst, size_t *dstlen, const void *src, size_t srclen, 1009 int level) 1010{ 1011 int ret; 1012 uLongf len = *dstlen; 1013 1014 if ((ret = compress2(dst, &len, src, srclen, level)) == Z_OK) 1015 *dstlen = (size_t)len; 1016 1017 return (ret); 1018} 1019 1020uid_t 1021crgetuid(cred_t *cr) 1022{ 1023 return (0); 1024} 1025 1026uid_t 1027crgetruid(cred_t *cr) 1028{ 1029 return (0); 1030} 1031 1032gid_t 1033crgetgid(cred_t *cr) 1034{ 1035 return (0); 1036} 1037 1038int 1039crgetngroups(cred_t *cr) 1040{ 1041 return (0); 1042} 1043 1044gid_t * 1045crgetgroups(cred_t *cr) 1046{ 1047 return (NULL); 1048} 1049 1050int 1051zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr) 1052{ 1053 return (0); 1054} 1055 1056int 1057zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr) 1058{ 1059 return (0); 1060} 1061 1062int 1063zfs_secpolicy_destroy_perms(const char *name, cred_t *cr) 1064{ 1065 return (0); 1066} 1067 1068ksiddomain_t * 1069ksid_lookupdomain(const char *dom) 1070{ 1071 ksiddomain_t *kd; 1072 1073 kd = umem_zalloc(sizeof (ksiddomain_t), UMEM_NOFAIL); 1074 kd->kd_name = spa_strdup(dom); 1075 return (kd); 1076} 1077 1078void 1079ksiddomain_rele(ksiddomain_t *ksid) 1080{ 1081 spa_strfree(ksid->kd_name); 1082 umem_free(ksid, sizeof (ksiddomain_t)); 1083} 1084 1085/* 1086 * Do not change the length of the returned string; it must be freed 1087 * with strfree(). 1088 */ 1089char * 1090kmem_asprintf(const char *fmt, ...) 1091{ 1092 int size; 1093 va_list adx; 1094 char *buf; 1095 1096 va_start(adx, fmt); 1097 size = vsnprintf(NULL, 0, fmt, adx) + 1; 1098 va_end(adx); 1099 1100 buf = kmem_alloc(size, KM_SLEEP); 1101 1102 va_start(adx, fmt); 1103 size = vsnprintf(buf, size, fmt, adx); 1104 va_end(adx); 1105 1106 return (buf); 1107} 1108 1109/* ARGSUSED */ 1110int 1111zfs_onexit_fd_hold(int fd, minor_t *minorp) 1112{ 1113 *minorp = 0; 1114 return (0); 1115} 1116 1117/* ARGSUSED */ 1118void 1119zfs_onexit_fd_rele(int fd) 1120{ 1121} 1122 1123/* ARGSUSED */ 1124int 1125zfs_onexit_add_cb(minor_t minor, void (*func)(void *), void *data, 1126 uint64_t *action_handle) 1127{ 1128 return (0); 1129} 1130 1131/* ARGSUSED */ 1132int 1133zfs_onexit_del_cb(minor_t minor, uint64_t action_handle, boolean_t fire) 1134{ 1135 return (0); 1136} 1137 1138/* ARGSUSED */ 1139int 1140zfs_onexit_cb_data(minor_t minor, uint64_t action_handle, void **data) 1141{ 1142 return (0); 1143} 1144 1145#ifdef __FreeBSD__ 1146/* ARGSUSED */ 1147int 1148zvol_create_minors(const char *name) 1149{ 1150 return (0); 1151} 1152#endif 1153 1154#ifdef illumos 1155void 1156bioinit(buf_t *bp) 1157{ 1158 bzero(bp, sizeof (buf_t)); 1159} 1160 1161void 1162biodone(buf_t *bp) 1163{ 1164 if (bp->b_iodone != NULL) { 1165 (*(bp->b_iodone))(bp); 1166 return; 1167 } 1168 ASSERT((bp->b_flags & B_DONE) == 0); 1169 bp->b_flags |= B_DONE; 1170} 1171 1172void 1173bioerror(buf_t *bp, int error) 1174{ 1175 ASSERT(bp != NULL); 1176 ASSERT(error >= 0); 1177 1178 if (error != 0) { 1179 bp->b_flags |= B_ERROR; 1180 } else { 1181 bp->b_flags &= ~B_ERROR; 1182 } 1183 bp->b_error = error; 1184} 1185 1186 1187int 1188geterror(struct buf *bp) 1189{ 1190 int error = 0; 1191 1192 if (bp->b_flags & B_ERROR) { 1193 error = bp->b_error; 1194 if (!error) 1195 error = EIO; 1196 } 1197 return (error); 1198} 1199#endif 1200