lib/libzpool/kernel.c

   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) 2016 Actifio, Inc. All rights reserved.
  24  */
  25
  26 #include <assert.h>
  27 #include <fcntl.h>
  28 #include <poll.h>
  29 #include <stdio.h>
  30 #include <stdlib.h>
  31 #include <string.h>
  32 #include <zlib.h>
  33 #include <libgen.h>
  34 #include <sys/signal.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/utsname.h>
  41 #include <sys/time.h>
  42 #include <sys/systeminfo.h>
  43 #include <zfs_fletcher.h>
  44 #include <sys/crypto/icp.h>
  45
  46 /*
  47  * Emulation of kernel services in userland.
  48  */
  49
  50 int aok;
  51 uint64_t physmem;
  52 vnode_t *rootdir = (vnode_t *)0xabcd1234;
  53 char hw_serial[HW_HOSTID_LEN];
  54 struct utsname hw_utsname;
  55 vmem_t *zio_arena = NULL;
  56
  57 /* If set, all blocks read will be copied to the specified directory. */
  58 char *vn_dumpdir = NULL;
  59
  60 /* this only exists to have its address taken */
  61 struct proc p0;
  62
  63 /*
  64  * =========================================================================
  65  * threads
  66  * =========================================================================
  67  *
  68  * TS_STACK_MIN is dictated by the minimum allowed pthread stack size.  While
  69  * TS_STACK_MAX is somewhat arbitrary, it was selected to be large enough for
  70  * the expected stack depth while small enough to avoid exhausting address
  71  * space with high thread counts.
  72  */
  73 #define TS_STACK_MIN    MAX(PTHREAD_STACK_MIN, 32768)
  74 #define TS_STACK_MAX    (256 * 1024)
  75
  76 /*ARGSUSED*/
  77 kthread_t *
  78 zk_thread_create(void (*func)(void *), void *arg, size_t stksize, int state)
  79 {
  80         pthread_attr_t attr;
  81         pthread_t tid;
  82         char *stkstr;
  83         int detachstate = PTHREAD_CREATE_DETACHED;
  84
  85         VERIFY0(pthread_attr_init(&attr));
  86
  87         if (state & TS_JOINABLE)
  88                 detachstate = PTHREAD_CREATE_JOINABLE;
  89
  90         VERIFY0(pthread_attr_setdetachstate(&attr, detachstate));
  91
  92         /*
  93          * We allow the default stack size in user space to be specified by
  94          * setting the ZFS_STACK_SIZE environment variable.  This allows us
  95          * the convenience of observing and debugging stack overruns in
  96          * user space.  Explicitly specified stack sizes will be honored.
  97          * The usage of ZFS_STACK_SIZE is discussed further in the
  98          * ENVIRONMENT VARIABLES sections of the ztest(1) man page.
  99          */
 100         if (stksize == 0) {
 101                 stkstr = getenv("ZFS_STACK_SIZE");
 102
 103                 if (stkstr == NULL)
 104                         stksize = TS_STACK_MAX;
 105                 else
 106                         stksize = MAX(atoi(stkstr), TS_STACK_MIN);
 107         }
 108
 109         VERIFY3S(stksize, >, 0);
 110         stksize = P2ROUNDUP(MAX(stksize, TS_STACK_MIN), PAGESIZE);
 111
 112         /*
 113          * If this ever fails, it may be because the stack size is not a
 114          * multiple of system page size.
 115          */
 116         VERIFY0(pthread_attr_setstacksize(&attr, stksize));
 117         VERIFY0(pthread_attr_setguardsize(&attr, PAGESIZE));
 118
 119         VERIFY0(pthread_create(&tid, &attr, (void *(*)(void *))func, arg));
 120         VERIFY0(pthread_attr_destroy(&attr));
 121
 122         return ((void *)(uintptr_t)tid);
 123 }
 124
 125 /*
 126  * =========================================================================
 127  * kstats
 128  * =========================================================================
 129  */
 130 /*ARGSUSED*/
 131 kstat_t *
 132 kstat_create(const char *module, int instance, const char *name,
 133     const char *class, uchar_t type, ulong_t ndata, uchar_t ks_flag)
 134 {
 135         return (NULL);
 136 }
 137
 138 /*ARGSUSED*/
 139 void
 140 kstat_install(kstat_t *ksp)
 141 {}
 142
 143 /*ARGSUSED*/
 144 void
 145 kstat_delete(kstat_t *ksp)
 146 {}
 147
 148 /*ARGSUSED*/
 149 void
 150 kstat_waitq_enter(kstat_io_t *kiop)
 151 {}
 152
 153 /*ARGSUSED*/
 154 void
 155 kstat_waitq_exit(kstat_io_t *kiop)
 156 {}
 157
 158 /*ARGSUSED*/
 159 void
 160 kstat_runq_enter(kstat_io_t *kiop)
 161 {}
 162
 163 /*ARGSUSED*/
 164 void
 165 kstat_runq_exit(kstat_io_t *kiop)
 166 {}
 167
 168 /*ARGSUSED*/
 169 void
 170 kstat_waitq_to_runq(kstat_io_t *kiop)
 171 {}
 172
 173 /*ARGSUSED*/
 174 void
 175 kstat_runq_back_to_waitq(kstat_io_t *kiop)
 176 {}
 177
 178 void
 179 kstat_set_raw_ops(kstat_t *ksp,
 180     int (*headers)(char *buf, size_t size),
 181     int (*data)(char *buf, size_t size, void *data),
 182     void *(*addr)(kstat_t *ksp, loff_t index))
 183 {}
 184
 185 /*
 186  * =========================================================================
 187  * mutexes
 188  * =========================================================================
 189  */
 190
 191 void
 192 mutex_init(kmutex_t *mp, char *name, int type, void *cookie)
 193 {
 194         VERIFY0(pthread_mutex_init(&mp->m_lock, NULL));
 195         memset(&mp->m_owner, 0, sizeof (pthread_t));
 196 }
 197
 198 void
 199 mutex_destroy(kmutex_t *mp)
 200 {
 201         VERIFY0(pthread_mutex_destroy(&mp->m_lock));
 202 }
 203
 204 void
 205 mutex_enter(kmutex_t *mp)
 206 {
 207         VERIFY0(pthread_mutex_lock(&mp->m_lock));
 208         mp->m_owner = pthread_self();
 209 }
 210
 211 int
 212 mutex_tryenter(kmutex_t *mp)
 213 {
 214         int error;
 215
 216         error = pthread_mutex_trylock(&mp->m_lock);
 217         if (error == 0) {
 218                 mp->m_owner = pthread_self();
 219                 return (1);
 220         } else {
 221                 VERIFY3S(error, ==, EBUSY);
 222                 return (0);
 223         }
 224 }
 225
 226 void
 227 mutex_exit(kmutex_t *mp)
 228 {
 229         memset(&mp->m_owner, 0, sizeof (pthread_t));
 230         VERIFY0(pthread_mutex_unlock(&mp->m_lock));
 231 }
 232
 233 /*
 234  * =========================================================================
 235  * rwlocks
 236  * =========================================================================
 237  */
 238
 239 void
 240 rw_init(krwlock_t *rwlp, char *name, int type, void *arg)
 241 {
 242         VERIFY0(pthread_rwlock_init(&rwlp->rw_lock, NULL));
 243         rwlp->rw_readers = 0;
 244         rwlp->rw_owner = 0;
 245 }
 246
 247 void
 248 rw_destroy(krwlock_t *rwlp)
 249 {
 250         VERIFY0(pthread_rwlock_destroy(&rwlp->rw_lock));
 251 }
 252
 253 void
 254 rw_enter(krwlock_t *rwlp, krw_t rw)
 255 {
 256         if (rw == RW_READER) {
 257                 VERIFY0(pthread_rwlock_rdlock(&rwlp->rw_lock));
 258                 atomic_inc_uint(&rwlp->rw_readers);
 259         } else {
 260                 VERIFY0(pthread_rwlock_wrlock(&rwlp->rw_lock));
 261                 rwlp->rw_owner = pthread_self();
 262         }
 263 }
 264
 265 void
 266 rw_exit(krwlock_t *rwlp)
 267 {
 268         if (RW_READ_HELD(rwlp))
 269                 atomic_dec_uint(&rwlp->rw_readers);
 270         else
 271                 rwlp->rw_owner = 0;
 272
 273         VERIFY0(pthread_rwlock_unlock(&rwlp->rw_lock));
 274 }
 275
 276 int
 277 rw_tryenter(krwlock_t *rwlp, krw_t rw)
 278 {
 279         int error;
 280
 281         if (rw == RW_READER)
 282                 error = pthread_rwlock_tryrdlock(&rwlp->rw_lock);
 283         else
 284                 error = pthread_rwlock_trywrlock(&rwlp->rw_lock);
 285
 286         if (error == 0) {
 287                 if (rw == RW_READER)
 288                         atomic_inc_uint(&rwlp->rw_readers);
 289                 else
 290                         rwlp->rw_owner = pthread_self();
 291
 292                 return (1);
 293         }
 294
 295         VERIFY3S(error, ==, EBUSY);
 296
 297         return (0);
 298 }
 299
 300 /* ARGSUSED */
 301 uint32_t
 302 zone_get_hostid(void *zonep)
 303 {
 304         /*
 305          * We're emulating the system's hostid in userland.
 306          */
 307         return (strtoul(hw_serial, NULL, 10));
 308 }
 309
 310 int
 311 rw_tryupgrade(krwlock_t *rwlp)
 312 {
 313         return (0);
 314 }
 315
 316 /*
 317  * =========================================================================
 318  * condition variables
 319  * =========================================================================
 320  */
 321
 322 void
 323 cv_init(kcondvar_t *cv, char *name, int type, void *arg)
 324 {
 325         VERIFY0(pthread_cond_init(cv, NULL));
 326 }
 327
 328 void
 329 cv_destroy(kcondvar_t *cv)
 330 {
 331         VERIFY0(pthread_cond_destroy(cv));
 332 }
 333
 334 void
 335 cv_wait(kcondvar_t *cv, kmutex_t *mp)
 336 {
 337         memset(&mp->m_owner, 0, sizeof (pthread_t));
 338         VERIFY0(pthread_cond_wait(cv, &mp->m_lock));
 339         mp->m_owner = pthread_self();
 340 }
 341
 342 clock_t
 343 cv_timedwait(kcondvar_t *cv, kmutex_t *mp, clock_t abstime)
 344 {
 345         int error;
 346         struct timeval tv;
 347         timestruc_t ts;
 348         clock_t delta;
 349
 350         delta = abstime - ddi_get_lbolt();
 351         if (delta <= 0)
 352                 return (-1);
 353
 354         VERIFY(gettimeofday(&tv, NULL) == 0);
 355
 356         ts.tv_sec = tv.tv_sec + delta / hz;
 357         ts.tv_nsec = tv.tv_usec * NSEC_PER_USEC + (delta % hz) * (NANOSEC / hz);
 358         if (ts.tv_nsec >= NANOSEC) {
 359                 ts.tv_sec++;
 360                 ts.tv_nsec -= NANOSEC;
 361         }
 362
 363         memset(&mp->m_owner, 0, sizeof (pthread_t));
 364         error = pthread_cond_timedwait(cv, &mp->m_lock, &ts);
 365         mp->m_owner = pthread_self();
 366
 367         if (error == ETIMEDOUT)
 368                 return (-1);
 369
 370         VERIFY0(error);
 371
 372         return (1);
 373 }
 374
 375 /*ARGSUSED*/
 376 clock_t
 377 cv_timedwait_hires(kcondvar_t *cv, kmutex_t *mp, hrtime_t tim, hrtime_t res,
 378     int flag)
 379 {
 380         int error;
 381         struct timeval tv;
 382         timestruc_t ts;
 383         hrtime_t delta;
 384
 385         ASSERT(flag == 0 || flag == CALLOUT_FLAG_ABSOLUTE);
 386
 387         delta = tim;
 388         if (flag & CALLOUT_FLAG_ABSOLUTE)
 389                 delta -= gethrtime();
 390
 391         if (delta <= 0)
 392                 return (-1);
 393
 394         VERIFY0(gettimeofday(&tv, NULL));
 395
 396         ts.tv_sec = tv.tv_sec + delta / NANOSEC;
 397         ts.tv_nsec = tv.tv_usec * NSEC_PER_USEC + (delta % NANOSEC);
 398         if (ts.tv_nsec >= NANOSEC) {
 399                 ts.tv_sec++;
 400                 ts.tv_nsec -= NANOSEC;
 401         }
 402
 403         memset(&mp->m_owner, 0, sizeof (pthread_t));
 404         error = pthread_cond_timedwait(cv, &mp->m_lock, &ts);
 405         mp->m_owner = pthread_self();
 406
 407         if (error == ETIMEDOUT)
 408                 return (-1);
 409
 410         VERIFY0(error);
 411
 412         return (1);
 413 }
 414
 415 void
 416 cv_signal(kcondvar_t *cv)
 417 {
 418         VERIFY0(pthread_cond_signal(cv));
 419 }
 420
 421 void
 422 cv_broadcast(kcondvar_t *cv)
 423 {
 424         VERIFY0(pthread_cond_broadcast(cv));
 425 }
 426
 427 /*
 428  * =========================================================================
 429  * vnode operations
 430  * =========================================================================
 431  */
 432 /*
 433  * Note: for the xxxat() versions of these functions, we assume that the
 434  * starting vp is always rootdir (which is true for spa_directory.c, the only
 435  * ZFS consumer of these interfaces).  We assert this is true, and then emulate
 436  * them by adding '/' in front of the path.
 437  */
 438
 439 /*ARGSUSED*/
 440 int
 441 vn_open(char *path, int x1, int flags, int mode, vnode_t **vpp, int x2, int x3)
 442 {
 443         int fd = -1;
 444         int dump_fd = -1;
 445         vnode_t *vp;
 446         int old_umask = 0;
 447         char *realpath;
 448         struct stat64 st;
 449         int err;
 450
 451         realpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL);
 452
 453         /*
 454          * If we're accessing a real disk from userland, we need to use
 455          * the character interface to avoid caching.  This is particularly
 456          * important if we're trying to look at a real in-kernel storage
 457          * pool from userland, e.g. via zdb, because otherwise we won't
 458          * see the changes occurring under the segmap cache.
 459          * On the other hand, the stupid character device returns zero
 460          * for its size.  So -- gag -- we open the block device to get
 461          * its size, and remember it for subsequent VOP_GETATTR().
 462          */
 463 #if defined(__sun__) || defined(__sun)
 464         if (strncmp(path, "/dev/", 5) == 0) {
 465 #else
 466         if (0) {
 467 #endif
 468                 char *dsk;
 469                 fd = open64(path, O_RDONLY);
 470                 if (fd == -1) {
 471                         err = errno;
 472                         free(realpath);
 473                         return (err);
 474                 }
 475                 if (fstat64(fd, &st) == -1) {
 476                         err = errno;
 477                         close(fd);
 478                         free(realpath);
 479                         return (err);
 480                 }
 481                 close(fd);
 482                 (void) sprintf(realpath, "%s", path);
 483                 dsk = strstr(path, "/dsk/");
 484                 if (dsk != NULL)
 485                         (void) sprintf(realpath + (dsk - path) + 1, "r%s",
 486                             dsk + 1);
 487         } else {
 488                 (void) sprintf(realpath, "%s", path);
 489                 if (!(flags & FCREAT) && stat64(realpath, &st) == -1) {
 490                         err = errno;
 491                         free(realpath);
 492                         return (err);
 493                 }
 494         }
 495
 496         if (!(flags & FCREAT) && S_ISBLK(st.st_mode)) {
 497 #ifdef __linux__
 498                 flags |= O_DIRECT;
 499 #endif
 500                 /* We shouldn't be writing to block devices in userspace */
 501                 VERIFY(!(flags & FWRITE));
 502         }
 503
 504         if (flags & FCREAT)
 505                 old_umask = umask(0);
 506
 507         /*
 508          * The construct 'flags - FREAD' conveniently maps combinations of
 509          * FREAD and FWRITE to the corresponding O_RDONLY, O_WRONLY, and O_RDWR.
 510          */
 511         fd = open64(realpath, flags - FREAD, mode);
 512         if (fd == -1) {
 513                 err = errno;
 514                 free(realpath);
 515                 return (err);
 516         }
 517
 518         if (flags & FCREAT)
 519                 (void) umask(old_umask);
 520
 521         if (vn_dumpdir != NULL) {
 522                 char *dumppath = umem_zalloc(MAXPATHLEN, UMEM_NOFAIL);
 523                 (void) snprintf(dumppath, MAXPATHLEN,
 524                     "%s/%s", vn_dumpdir, basename(realpath));
 525                 dump_fd = open64(dumppath, O_CREAT | O_WRONLY, 0666);
 526                 umem_free(dumppath, MAXPATHLEN);
 527                 if (dump_fd == -1) {
 528                         err = errno;
 529                         free(realpath);
 530                         close(fd);
 531                         return (err);
 532                 }
 533         } else {
 534                 dump_fd = -1;
 535         }
 536
 537         free(realpath);
 538
 539         if (fstat64_blk(fd, &st) == -1) {
 540                 err = errno;
 541                 close(fd);
 542                 if (dump_fd != -1)
 543                         close(dump_fd);
 544                 return (err);
 545         }
 546
 547         (void) fcntl(fd, F_SETFD, FD_CLOEXEC);
 548
 549         *vpp = vp = umem_zalloc(sizeof (vnode_t), UMEM_NOFAIL);
 550
 551         vp->v_fd = fd;
 552         vp->v_size = st.st_size;
 553         vp->v_path = spa_strdup(path);
 554         vp->v_dump_fd = dump_fd;
 555
 556         return (0);
 557 }
 558
 559 /*ARGSUSED*/
 560 int
 561 vn_openat(char *path, int x1, int flags, int mode, vnode_t **vpp, int x2,
 562     int x3, vnode_t *startvp, int fd)
 563 {
 564         char *realpath = umem_alloc(strlen(path) + 2, UMEM_NOFAIL);
 565         int ret;
 566
 567         ASSERT(startvp == rootdir);
 568         (void) sprintf(realpath, "/%s", path);
 569
 570         /* fd ignored for now, need if want to simulate nbmand support */
 571         ret = vn_open(realpath, x1, flags, mode, vpp, x2, x3);
 572
 573         umem_free(realpath, strlen(path) + 2);
 574
 575         return (ret);
 576 }
 577
 578 /*ARGSUSED*/
 579 int
 580 vn_rdwr(int uio, vnode_t *vp, void *addr, ssize_t len, offset_t offset,
 581     int x1, int x2, rlim64_t x3, void *x4, ssize_t *residp)
 582 {
 583         ssize_t rc, done = 0, split;
 584
 585         if (uio == UIO_READ) {
 586                 rc = pread64(vp->v_fd, addr, len, offset);
 587                 if (vp->v_dump_fd != -1 && rc != -1) {
 588                         int status;
 589                         status = pwrite64(vp->v_dump_fd, addr, rc, offset);
 590                         ASSERT(status != -1);
 591                 }
 592         } else {
 593                 /*
 594                  * To simulate partial disk writes, we split writes into two
 595                  * system calls so that the process can be killed in between.
 596                  */
 597                 int sectors = len >> SPA_MINBLOCKSHIFT;
 598                 split = (sectors > 0 ? rand() % sectors : 0) <<
 599                     SPA_MINBLOCKSHIFT;
 600                 rc = pwrite64(vp->v_fd, addr, split, offset);
 601                 if (rc != -1) {
 602                         done = rc;
 603                         rc = pwrite64(vp->v_fd, (char *)addr + split,
 604                             len - split, offset + split);
 605                 }
 606         }
 607
 608 #ifdef __linux__
 609         if (rc == -1 && errno == EINVAL) {
 610                 /*
 611                  * Under Linux, this most likely means an alignment issue
 612                  * (memory or disk) due to O_DIRECT, so we abort() in order to
 613                  * catch the offender.
 614                  */
 615                 abort();
 616         }
 617 #endif
 618         if (rc == -1)
 619                 return (errno);
 620
 621         done += rc;
 622
 623         if (residp)
 624                 *residp = len - done;
 625         else if (done != len)
 626                 return (EIO);
 627         return (0);
 628 }
 629
 630 void
 631 vn_close(vnode_t *vp)
 632 {
 633         close(vp->v_fd);
 634         if (vp->v_dump_fd != -1)
 635                 close(vp->v_dump_fd);
 636         spa_strfree(vp->v_path);
 637         umem_free(vp, sizeof (vnode_t));
 638 }
 639
 640 /*
 641  * At a minimum we need to update the size since vdev_reopen()
 642  * will no longer call vn_openat().
 643  */
 644 int
 645 fop_getattr(vnode_t *vp, vattr_t *vap)
 646 {
 647         struct stat64 st;
 648         int err;
 649
 650         if (fstat64_blk(vp->v_fd, &st) == -1) {
 651                 err = errno;
 652                 close(vp->v_fd);
 653                 return (err);
 654         }
 655
 656         vap->va_size = st.st_size;
 657         return (0);
 658 }
 659
 660 /*
 661  * =========================================================================
 662  * Figure out which debugging statements to print
 663  * =========================================================================
 664  */
 665
 666 static char *dprintf_string;
 667 static int dprintf_print_all;
 668
 669 int
 670 dprintf_find_string(const char *string)
 671 {
 672         char *tmp_str = dprintf_string;
 673         int len = strlen(string);
 674
 675         /*
 676          * Find out if this is a string we want to print.
 677          * String format: file1.c,function_name1,file2.c,file3.c
 678          */
 679
 680         while (tmp_str != NULL) {
 681                 if (strncmp(tmp_str, string, len) == 0 &&
 682                     (tmp_str[len] == ',' || tmp_str[len] == '\0'))
 683                         return (1);
 684                 tmp_str = strchr(tmp_str, ',');
 685                 if (tmp_str != NULL)
 686                         tmp_str++; /* Get rid of , */
 687         }
 688         return (0);
 689 }
 690
 691 void
 692 dprintf_setup(int *argc, char **argv)
 693 {
 694         int i, j;
 695
 696         /*
 697          * Debugging can be specified two ways: by setting the
 698          * environment variable ZFS_DEBUG, or by including a
 699          * "debug=..."  argument on the command line.  The command
 700          * line setting overrides the environment variable.
 701          */
 702
 703         for (i = 1; i < *argc; i++) {
 704                 int len = strlen("debug=");
 705                 /* First look for a command line argument */
 706                 if (strncmp("debug=", argv[i], len) == 0) {
 707                         dprintf_string = argv[i] + len;
 708                         /* Remove from args */
 709                         for (j = i; j < *argc; j++)
 710                                 argv[j] = argv[j+1];
 711                         argv[j] = NULL;
 712                         (*argc)--;
 713                 }
 714         }
 715
 716         if (dprintf_string == NULL) {
 717                 /* Look for ZFS_DEBUG environment variable */
 718                 dprintf_string = getenv("ZFS_DEBUG");
 719         }
 720
 721         /*
 722          * Are we just turning on all debugging?
 723          */
 724         if (dprintf_find_string("on"))
 725                 dprintf_print_all = 1;
 726
 727         if (dprintf_string != NULL)
 728                 zfs_flags |= ZFS_DEBUG_DPRINTF;
 729 }
 730
 731 /*
 732  * =========================================================================
 733  * debug printfs
 734  * =========================================================================
 735  */
 736 void
 737 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
 738 {
 739         const char *newfile;
 740         va_list adx;
 741
 742         /*
 743          * Get rid of annoying "../common/" prefix to filename.
 744          */
 745         newfile = strrchr(file, '/');
 746         if (newfile != NULL) {
 747                 newfile = newfile + 1; /* Get rid of leading / */
 748         } else {
 749                 newfile = file;
 750         }
 751
 752         if (dprintf_print_all ||
 753             dprintf_find_string(newfile) ||
 754             dprintf_find_string(func)) {
 755                 /* Print out just the function name if requested */
 756                 flockfile(stdout);
 757                 if (dprintf_find_string("pid"))
 758                         (void) printf("%d ", getpid());
 759                 if (dprintf_find_string("tid"))
 760                         (void) printf("%u ", (uint_t)pthread_self());
 761                 if (dprintf_find_string("cpu"))
 762                         (void) printf("%u ", getcpuid());
 763                 if (dprintf_find_string("time"))
 764                         (void) printf("%llu ", gethrtime());
 765                 if (dprintf_find_string("long"))
 766                         (void) printf("%s, line %d: ", newfile, line);
 767                 (void) printf("%s: ", func);
 768                 va_start(adx, fmt);
 769                 (void) vprintf(fmt, adx);
 770                 va_end(adx);
 771                 funlockfile(stdout);
 772         }
 773 }
 774
 775 /*
 776  * =========================================================================
 777  * cmn_err() and panic()
 778  * =========================================================================
 779  */
 780 static char ce_prefix[CE_IGNORE][10] = { "", "NOTICE: ", "WARNING: ", "" };
 781 static char ce_suffix[CE_IGNORE][2] = { "", "\n", "\n", "" };
 782
 783 void
 784 vpanic(const char *fmt, va_list adx)
 785 {
 786         (void) fprintf(stderr, "error: ");
 787         (void) vfprintf(stderr, fmt, adx);
 788         (void) fprintf(stderr, "\n");
 789
 790         abort();        /* think of it as a "user-level crash dump" */
 791 }
 792
 793 void
 794 panic(const char *fmt, ...)
 795 {
 796         va_list adx;
 797
 798         va_start(adx, fmt);
 799         vpanic(fmt, adx);
 800         va_end(adx);
 801 }
 802
 803 void
 804 vcmn_err(int ce, const char *fmt, va_list adx)
 805 {
 806         if (ce == CE_PANIC)
 807                 vpanic(fmt, adx);
 808         if (ce != CE_NOTE) {    /* suppress noise in userland stress testing */
 809                 (void) fprintf(stderr, "%s", ce_prefix[ce]);
 810                 (void) vfprintf(stderr, fmt, adx);
 811                 (void) fprintf(stderr, "%s", ce_suffix[ce]);
 812         }
 813 }
 814
 815 /*PRINTFLIKE2*/
 816 void
 817 cmn_err(int ce, const char *fmt, ...)
 818 {
 819         va_list adx;
 820
 821         va_start(adx, fmt);
 822         vcmn_err(ce, fmt, adx);
 823         va_end(adx);
 824 }
 825
 826 /*
 827  * =========================================================================
 828  * kobj interfaces
 829  * =========================================================================
 830  */
 831 struct _buf *
 832 kobj_open_file(char *name)
 833 {
 834         struct _buf *file;
 835         vnode_t *vp;
 836
 837         /* set vp as the _fd field of the file */
 838         if (vn_openat(name, UIO_SYSSPACE, FREAD, 0, &vp, 0, 0, rootdir,
 839             -1) != 0)
 840                 return ((void *)-1UL);
 841
 842         file = umem_zalloc(sizeof (struct _buf), UMEM_NOFAIL);
 843         file->_fd = (intptr_t)vp;
 844         return (file);
 845 }
 846
 847 int
 848 kobj_read_file(struct _buf *file, char *buf, unsigned size, unsigned off)
 849 {
 850         ssize_t resid = 0;
 851
 852         if (vn_rdwr(UIO_READ, (vnode_t *)file->_fd, buf, size, (offset_t)off,
 853             UIO_SYSSPACE, 0, 0, 0, &resid) != 0)
 854                 return (-1);
 855
 856         return (size - resid);
 857 }
 858
 859 void
 860 kobj_close_file(struct _buf *file)
 861 {
 862         vn_close((vnode_t *)file->_fd);
 863         umem_free(file, sizeof (struct _buf));
 864 }
 865
 866 int
 867 kobj_get_filesize(struct _buf *file, uint64_t *size)
 868 {
 869         struct stat64 st;
 870         vnode_t *vp = (vnode_t *)file->_fd;
 871
 872         if (fstat64(vp->v_fd, &st) == -1) {
 873                 vn_close(vp);
 874                 return (errno);
 875         }
 876         *size = st.st_size;
 877         return (0);
 878 }
 879
 880 /*
 881  * =========================================================================
 882  * misc routines
 883  * =========================================================================
 884  */
 885
 886 void
 887 delay(clock_t ticks)
 888 {
 889         (void) poll(0, 0, ticks * (1000 / hz));
 890 }
 891
 892 /*
 893  * Find highest one bit set.
 894  * Returns bit number + 1 of highest bit that is set, otherwise returns 0.
 895  * The __builtin_clzll() function is supported by both GCC and Clang.
 896  */
 897 int
 898 highbit64(uint64_t i)
 899 {
 900         if (i == 0)
 901         return (0);
 902
 903         return (NBBY * sizeof (uint64_t) - __builtin_clzll(i));
 904 }
 905
 906 /*
 907  * Find lowest one bit set.
 908  * Returns bit number + 1 of lowest bit that is set, otherwise returns 0.
 909  * The __builtin_ffsll() function is supported by both GCC and Clang.
 910  */
 911 int
 912 lowbit64(uint64_t i)
 913 {
 914         if (i == 0)
 915                 return (0);
 916
 917         return (__builtin_ffsll(i));
 918 }
 919
 920 char *random_path = "/dev/random";
 921 char *urandom_path = "/dev/urandom";
 922 static int random_fd = -1, urandom_fd = -1;
 923
 924 void
 925 random_init(void)
 926 {
 927         VERIFY((random_fd = open(random_path, O_RDONLY)) != -1);
 928         VERIFY((urandom_fd = open(urandom_path, O_RDONLY)) != -1);
 929 }
 930
 931 void
 932 random_fini(void)
 933 {
 934         close(random_fd);
 935         close(urandom_fd);
 936
 937         random_fd = -1;
 938         urandom_fd = -1;
 939 }
 940
 941 static int
 942 random_get_bytes_common(uint8_t *ptr, size_t len, int fd)
 943 {
 944         size_t resid = len;
 945         ssize_t bytes;
 946
 947         ASSERT(fd != -1);
 948
 949         while (resid != 0) {
 950                 bytes = read(fd, ptr, resid);
 951                 ASSERT3S(bytes, >=, 0);
 952                 ptr += bytes;
 953                 resid -= bytes;
 954         }
 955
 956         return (0);
 957 }
 958
 959 int
 960 random_get_bytes(uint8_t *ptr, size_t len)
 961 {
 962         return (random_get_bytes_common(ptr, len, random_fd));
 963 }
 964
 965 int
 966 random_get_pseudo_bytes(uint8_t *ptr, size_t len)
 967 {
 968         return (random_get_bytes_common(ptr, len, urandom_fd));
 969 }
 970
 971 int
 972 ddi_strtoul(const char *hw_serial, char **nptr, int base, unsigned long *result)
 973 {
 974         char *end;
 975
 976         *result = strtoul(hw_serial, &end, base);
 977         if (*result == 0)
 978                 return (errno);
 979         return (0);
 980 }
 981
 982 int
 983 ddi_strtoull(const char *str, char **nptr, int base, u_longlong_t *result)
 984 {
 985         char *end;
 986
 987         *result = strtoull(str, &end, base);
 988         if (*result == 0)
 989                 return (errno);
 990         return (0);
 991 }
 992
 993 utsname_t *
 994 utsname(void)
 995 {
 996         return (&hw_utsname);
 997 }
 998
 999 /*
1000  * =========================================================================
1001  * kernel emulation setup & teardown
1002  * =========================================================================
1003  */
1004 static int
1005 umem_out_of_memory(void)
1006 {
1007         char errmsg[] = "out of memory -- generating core dump\n";
1008
1009         (void) fprintf(stderr, "%s", errmsg);
1010         abort();
1011         return (0);
1012 }
1013
1014 void
1015 kernel_init(int mode)
1016 {
1017         extern uint_t rrw_tsd_key;
1018
1019         umem_nofail_callback(umem_out_of_memory);
1020
1021         physmem = sysconf(_SC_PHYS_PAGES);
1022
1023         dprintf("physmem = %llu pages (%.2f GB)\n", physmem,
1024             (double)physmem * sysconf(_SC_PAGE_SIZE) / (1ULL << 30));
1025
1026         (void) snprintf(hw_serial, sizeof (hw_serial), "%ld",
1027             (mode & FWRITE) ? get_system_hostid() : 0);
1028
1029         random_init();
1030
1031         VERIFY0(uname(&hw_utsname));
1032
1033         system_taskq_init();
1034         icp_init();
1035
1036         spa_init(mode);
1037
1038         fletcher_4_init();
1039
1040         tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
1041 }
1042
1043 void
1044 kernel_fini(void)
1045 {
1046         fletcher_4_fini();
1047         spa_fini();
1048
1049         icp_fini();
1050         system_taskq_fini();
1051
1052         random_fini();
1053 }
1054
1055 uid_t
1056 crgetuid(cred_t *cr)
1057 {
1058         return (0);
1059 }
1060
1061 uid_t
1062 crgetruid(cred_t *cr)
1063 {
1064         return (0);
1065 }
1066
1067 gid_t
1068 crgetgid(cred_t *cr)
1069 {
1070         return (0);
1071 }
1072
1073 int
1074 crgetngroups(cred_t *cr)
1075 {
1076         return (0);
1077 }
1078
1079 gid_t *
1080 crgetgroups(cred_t *cr)
1081 {
1082         return (NULL);
1083 }
1084
1085 int
1086 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
1087 {
1088         return (0);
1089 }
1090
1091 int
1092 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
1093 {
1094         return (0);
1095 }
1096
1097 int
1098 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
1099 {
1100         return (0);
1101 }
1102
1103 int
1104 secpolicy_zfs(const cred_t *cr)
1105 {
1106         return (0);
1107 }
1108
1109 ksiddomain_t *
1110 ksid_lookupdomain(const char *dom)
1111 {
1112         ksiddomain_t *kd;
1113
1114         kd = umem_zalloc(sizeof (ksiddomain_t), UMEM_NOFAIL);
1115         kd->kd_name = spa_strdup(dom);
1116         return (kd);
1117 }
1118
1119 void
1120 ksiddomain_rele(ksiddomain_t *ksid)
1121 {
1122         spa_strfree(ksid->kd_name);
1123         umem_free(ksid, sizeof (ksiddomain_t));
1124 }
1125
1126 char *
1127 kmem_vasprintf(const char *fmt, va_list adx)
1128 {
1129         char *buf = NULL;
1130         va_list adx_copy;
1131
1132         va_copy(adx_copy, adx);
1133         VERIFY(vasprintf(&buf, fmt, adx_copy) != -1);
1134         va_end(adx_copy);
1135
1136         return (buf);
1137 }
1138
1139 char *
1140 kmem_asprintf(const char *fmt, ...)
1141 {
1142         char *buf = NULL;
1143         va_list adx;
1144
1145         va_start(adx, fmt);
1146         VERIFY(vasprintf(&buf, fmt, adx) != -1);
1147         va_end(adx);
1148
1149         return (buf);
1150 }
1151
1152 /* ARGSUSED */
1153 int
1154 zfs_onexit_fd_hold(int fd, minor_t *minorp)
1155 {
1156         *minorp = 0;
1157         return (0);
1158 }
1159
1160 /* ARGSUSED */
1161 void
1162 zfs_onexit_fd_rele(int fd)
1163 {
1164 }
1165
1166 /* ARGSUSED */
1167 int
1168 zfs_onexit_add_cb(minor_t minor, void (*func)(void *), void *data,
1169     uint64_t *action_handle)
1170 {
1171         return (0);
1172 }
1173
1174 /* ARGSUSED */
1175 int
1176 zfs_onexit_del_cb(minor_t minor, uint64_t action_handle, boolean_t fire)
1177 {
1178         return (0);
1179 }
1180
1181 /* ARGSUSED */
1182 int
1183 zfs_onexit_cb_data(minor_t minor, uint64_t action_handle, void **data)
1184 {
1185         return (0);
1186 }
1187
1188 fstrans_cookie_t
1189 spl_fstrans_mark(void)
1190 {
1191         return ((fstrans_cookie_t)0);
1192 }
1193
1194 void
1195 spl_fstrans_unmark(fstrans_cookie_t cookie)
1196 {
1197 }
1198
1199 int
1200 __spl_pf_fstrans_check(void)
1201 {
1202         return (0);
1203 }
1204
1205 void *zvol_tag = "zvol_tag";
1206
1207 void
1208 zvol_create_minors(spa_t *spa, const char *name, boolean_t async)
1209 {
1210 }
1211
1212 void
1213 zvol_remove_minor(spa_t *spa, const char *name, boolean_t async)
1214 {
1215 }
1216
1217 void
1218 zvol_remove_minors(spa_t *spa, const char *name, boolean_t async)
1219 {
1220 }
1221
1222 void
1223 zvol_rename_minors(spa_t *spa, const char *oldname, const char *newname,
1224     boolean_t async)
1225 {
1226 }