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         struct timespec 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         struct timespec 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  * procfs list
 430  * =========================================================================
 431  */
 432
 433 void
 434 seq_printf(struct seq_file *m, const char *fmt, ...)
 435 {}
 436
 437 void
 438 procfs_list_install(const char *module,
 439     const char *name,
 440     mode_t mode,
 441     procfs_list_t *procfs_list,
 442     int (*show)(struct seq_file *f, void *p),
 443     int (*show_header)(struct seq_file *f),
 444     int (*clear)(procfs_list_t *procfs_list),
 445     size_t procfs_list_node_off)
 446 {
 447         mutex_init(&procfs_list->pl_lock, NULL, MUTEX_DEFAULT, NULL);
 448         list_create(&procfs_list->pl_list,
 449             procfs_list_node_off + sizeof (procfs_list_node_t),
 450             procfs_list_node_off + offsetof(procfs_list_node_t, pln_link));
 451         procfs_list->pl_next_id = 1;
 452         procfs_list->pl_node_offset = procfs_list_node_off;
 453 }
 454
 455 void
 456 procfs_list_uninstall(procfs_list_t *procfs_list)
 457 {}
 458
 459 void
 460 procfs_list_destroy(procfs_list_t *procfs_list)
 461 {
 462         ASSERT(list_is_empty(&procfs_list->pl_list));
 463         list_destroy(&procfs_list->pl_list);
 464         mutex_destroy(&procfs_list->pl_lock);
 465 }
 466
 467 #define NODE_ID(procfs_list, obj) \
 468                 (((procfs_list_node_t *)(((char *)obj) + \
 469                 (procfs_list)->pl_node_offset))->pln_id)
 470
 471 void
 472 procfs_list_add(procfs_list_t *procfs_list, void *p)
 473 {
 474         ASSERT(MUTEX_HELD(&procfs_list->pl_lock));
 475         NODE_ID(procfs_list, p) = procfs_list->pl_next_id++;
 476         list_insert_tail(&procfs_list->pl_list, p);
 477 }
 478
 479 /*
 480  * =========================================================================
 481  * vnode operations
 482  * =========================================================================
 483  */
 484 /*
 485  * Note: for the xxxat() versions of these functions, we assume that the
 486  * starting vp is always rootdir (which is true for spa_directory.c, the only
 487  * ZFS consumer of these interfaces).  We assert this is true, and then emulate
 488  * them by adding '/' in front of the path.
 489  */
 490
 491 /*ARGSUSED*/
 492 int
 493 vn_open(char *path, int x1, int flags, int mode, vnode_t **vpp, int x2, int x3)
 494 {
 495         int fd = -1;
 496         int dump_fd = -1;
 497         vnode_t *vp;
 498         int old_umask = 0;
 499         char *realpath;
 500         struct stat64 st;
 501         int err;
 502
 503         realpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL);
 504
 505         /*
 506          * If we're accessing a real disk from userland, we need to use
 507          * the character interface to avoid caching.  This is particularly
 508          * important if we're trying to look at a real in-kernel storage
 509          * pool from userland, e.g. via zdb, because otherwise we won't
 510          * see the changes occurring under the segmap cache.
 511          * On the other hand, the stupid character device returns zero
 512          * for its size.  So -- gag -- we open the block device to get
 513          * its size, and remember it for subsequent VOP_GETATTR().
 514          */
 515 #if defined(__sun__) || defined(__sun)
 516         if (strncmp(path, "/dev/", 5) == 0) {
 517 #else
 518         if (0) {
 519 #endif
 520                 char *dsk;
 521                 fd = open64(path, O_RDONLY);
 522                 if (fd == -1) {
 523                         err = errno;
 524                         free(realpath);
 525                         return (err);
 526                 }
 527                 if (fstat64(fd, &st) == -1) {
 528                         err = errno;
 529                         close(fd);
 530                         free(realpath);
 531                         return (err);
 532                 }
 533                 close(fd);
 534                 (void) sprintf(realpath, "%s", path);
 535                 dsk = strstr(path, "/dsk/");
 536                 if (dsk != NULL)
 537                         (void) sprintf(realpath + (dsk - path) + 1, "r%s",
 538                             dsk + 1);
 539         } else {
 540                 (void) sprintf(realpath, "%s", path);
 541                 if (!(flags & FCREAT) && stat64(realpath, &st) == -1) {
 542                         err = errno;
 543                         free(realpath);
 544                         return (err);
 545                 }
 546         }
 547
 548         if (!(flags & FCREAT) && S_ISBLK(st.st_mode)) {
 549 #ifdef __linux__
 550                 flags |= O_DIRECT;
 551 #endif
 552         }
 553
 554         if (flags & FCREAT)
 555                 old_umask = umask(0);
 556
 557         /*
 558          * The construct 'flags - FREAD' conveniently maps combinations of
 559          * FREAD and FWRITE to the corresponding O_RDONLY, O_WRONLY, and O_RDWR.
 560          */
 561         fd = open64(realpath, flags - FREAD, mode);
 562         if (fd == -1) {
 563                 err = errno;
 564                 free(realpath);
 565                 return (err);
 566         }
 567
 568         if (flags & FCREAT)
 569                 (void) umask(old_umask);
 570
 571         if (vn_dumpdir != NULL) {
 572                 char *dumppath = umem_zalloc(MAXPATHLEN, UMEM_NOFAIL);
 573                 (void) snprintf(dumppath, MAXPATHLEN,
 574                     "%s/%s", vn_dumpdir, basename(realpath));
 575                 dump_fd = open64(dumppath, O_CREAT | O_WRONLY, 0666);
 576                 umem_free(dumppath, MAXPATHLEN);
 577                 if (dump_fd == -1) {
 578                         err = errno;
 579                         free(realpath);
 580                         close(fd);
 581                         return (err);
 582                 }
 583         } else {
 584                 dump_fd = -1;
 585         }
 586
 587         free(realpath);
 588
 589         if (fstat64_blk(fd, &st) == -1) {
 590                 err = errno;
 591                 close(fd);
 592                 if (dump_fd != -1)
 593                         close(dump_fd);
 594                 return (err);
 595         }
 596
 597         (void) fcntl(fd, F_SETFD, FD_CLOEXEC);
 598
 599         *vpp = vp = umem_zalloc(sizeof (vnode_t), UMEM_NOFAIL);
 600
 601         vp->v_fd = fd;
 602         vp->v_size = st.st_size;
 603         vp->v_path = spa_strdup(path);
 604         vp->v_dump_fd = dump_fd;
 605
 606         return (0);
 607 }
 608
 609 /*ARGSUSED*/
 610 int
 611 vn_openat(char *path, int x1, int flags, int mode, vnode_t **vpp, int x2,
 612     int x3, vnode_t *startvp, int fd)
 613 {
 614         char *realpath = umem_alloc(strlen(path) + 2, UMEM_NOFAIL);
 615         int ret;
 616
 617         ASSERT(startvp == rootdir);
 618         (void) sprintf(realpath, "/%s", path);
 619
 620         /* fd ignored for now, need if want to simulate nbmand support */
 621         ret = vn_open(realpath, x1, flags, mode, vpp, x2, x3);
 622
 623         umem_free(realpath, strlen(path) + 2);
 624
 625         return (ret);
 626 }
 627
 628 /*ARGSUSED*/
 629 int
 630 vn_rdwr(int uio, vnode_t *vp, void *addr, ssize_t len, offset_t offset,
 631     int x1, int x2, rlim64_t x3, void *x4, ssize_t *residp)
 632 {
 633         ssize_t rc, done = 0, split;
 634
 635         if (uio == UIO_READ) {
 636                 rc = pread64(vp->v_fd, addr, len, offset);
 637                 if (vp->v_dump_fd != -1 && rc != -1) {
 638                         int status;
 639                         status = pwrite64(vp->v_dump_fd, addr, rc, offset);
 640                         ASSERT(status != -1);
 641                 }
 642         } else {
 643                 /*
 644                  * To simulate partial disk writes, we split writes into two
 645                  * system calls so that the process can be killed in between.
 646                  */
 647                 int sectors = len >> SPA_MINBLOCKSHIFT;
 648                 split = (sectors > 0 ? rand() % sectors : 0) <<
 649                     SPA_MINBLOCKSHIFT;
 650                 rc = pwrite64(vp->v_fd, addr, split, offset);
 651                 if (rc != -1) {
 652                         done = rc;
 653                         rc = pwrite64(vp->v_fd, (char *)addr + split,
 654                             len - split, offset + split);
 655                 }
 656         }
 657
 658 #ifdef __linux__
 659         if (rc == -1 && errno == EINVAL) {
 660                 /*
 661                  * Under Linux, this most likely means an alignment issue
 662                  * (memory or disk) due to O_DIRECT, so we abort() in order to
 663                  * catch the offender.
 664                  */
 665                 abort();
 666         }
 667 #endif
 668         if (rc == -1)
 669                 return (errno);
 670
 671         done += rc;
 672
 673         if (residp)
 674                 *residp = len - done;
 675         else if (done != len)
 676                 return (EIO);
 677         return (0);
 678 }
 679
 680 void
 681 vn_close(vnode_t *vp)
 682 {
 683         close(vp->v_fd);
 684         if (vp->v_dump_fd != -1)
 685                 close(vp->v_dump_fd);
 686         spa_strfree(vp->v_path);
 687         umem_free(vp, sizeof (vnode_t));
 688 }
 689
 690 /*
 691  * At a minimum we need to update the size since vdev_reopen()
 692  * will no longer call vn_openat().
 693  */
 694 int
 695 fop_getattr(vnode_t *vp, vattr_t *vap)
 696 {
 697         struct stat64 st;
 698         int err;
 699
 700         if (fstat64_blk(vp->v_fd, &st) == -1) {
 701                 err = errno;
 702                 close(vp->v_fd);
 703                 return (err);
 704         }
 705
 706         vap->va_size = st.st_size;
 707         return (0);
 708 }
 709
 710 /*
 711  * =========================================================================
 712  * Figure out which debugging statements to print
 713  * =========================================================================
 714  */
 715
 716 static char *dprintf_string;
 717 static int dprintf_print_all;
 718
 719 int
 720 dprintf_find_string(const char *string)
 721 {
 722         char *tmp_str = dprintf_string;
 723         int len = strlen(string);
 724
 725         /*
 726          * Find out if this is a string we want to print.
 727          * String format: file1.c,function_name1,file2.c,file3.c
 728          */
 729
 730         while (tmp_str != NULL) {
 731                 if (strncmp(tmp_str, string, len) == 0 &&
 732                     (tmp_str[len] == ',' || tmp_str[len] == '\0'))
 733                         return (1);
 734                 tmp_str = strchr(tmp_str, ',');
 735                 if (tmp_str != NULL)
 736                         tmp_str++; /* Get rid of , */
 737         }
 738         return (0);
 739 }
 740
 741 void
 742 dprintf_setup(int *argc, char **argv)
 743 {
 744         int i, j;
 745
 746         /*
 747          * Debugging can be specified two ways: by setting the
 748          * environment variable ZFS_DEBUG, or by including a
 749          * "debug=..."  argument on the command line.  The command
 750          * line setting overrides the environment variable.
 751          */
 752
 753         for (i = 1; i < *argc; i++) {
 754                 int len = strlen("debug=");
 755                 /* First look for a command line argument */
 756                 if (strncmp("debug=", argv[i], len) == 0) {
 757                         dprintf_string = argv[i] + len;
 758                         /* Remove from args */
 759                         for (j = i; j < *argc; j++)
 760                                 argv[j] = argv[j+1];
 761                         argv[j] = NULL;
 762                         (*argc)--;
 763                 }
 764         }
 765
 766         if (dprintf_string == NULL) {
 767                 /* Look for ZFS_DEBUG environment variable */
 768                 dprintf_string = getenv("ZFS_DEBUG");
 769         }
 770
 771         /*
 772          * Are we just turning on all debugging?
 773          */
 774         if (dprintf_find_string("on"))
 775                 dprintf_print_all = 1;
 776
 777         if (dprintf_string != NULL)
 778                 zfs_flags |= ZFS_DEBUG_DPRINTF;
 779 }
 780
 781 /*
 782  * =========================================================================
 783  * debug printfs
 784  * =========================================================================
 785  */
 786 void
 787 __dprintf(boolean_t dprint, const char *file, const char *func,
 788     int line, const char *fmt, ...)
 789 {
 790         const char *newfile;
 791         va_list adx;
 792
 793         /*
 794          * Get rid of annoying "../common/" prefix to filename.
 795          */
 796         newfile = strrchr(file, '/');
 797         if (newfile != NULL) {
 798                 newfile = newfile + 1; /* Get rid of leading / */
 799         } else {
 800                 newfile = file;
 801         }
 802
 803         if (dprint) {
 804                 /* dprintf messages are printed immediately */
 805
 806                 if (!dprintf_print_all &&
 807                     !dprintf_find_string(newfile) &&
 808                     !dprintf_find_string(func))
 809                         return;
 810
 811                 /* Print out just the function name if requested */
 812                 flockfile(stdout);
 813                 if (dprintf_find_string("pid"))
 814                         (void) printf("%d ", getpid());
 815                 if (dprintf_find_string("tid"))
 816                         (void) printf("%u ", (uint_t)pthread_self());
 817                 if (dprintf_find_string("cpu"))
 818                         (void) printf("%u ", getcpuid());
 819                 if (dprintf_find_string("time"))
 820                         (void) printf("%llu ", gethrtime());
 821                 if (dprintf_find_string("long"))
 822                         (void) printf("%s, line %d: ", newfile, line);
 823                 (void) printf("dprintf: %s: ", func);
 824                 va_start(adx, fmt);
 825                 (void) vprintf(fmt, adx);
 826                 va_end(adx);
 827                 funlockfile(stdout);
 828         } else {
 829                 /* zfs_dbgmsg is logged for dumping later */
 830                 size_t size;
 831                 char *buf;
 832                 int i;
 833
 834                 size = 1024;
 835                 buf = umem_alloc(size, UMEM_NOFAIL);
 836                 i = snprintf(buf, size, "%s:%d:%s(): ", newfile, line, func);
 837
 838                 if (i < size) {
 839                         va_start(adx, fmt);
 840                         (void) vsnprintf(buf + i, size - i, fmt, adx);
 841                         va_end(adx);
 842                 }
 843
 844                 __zfs_dbgmsg(buf);
 845
 846                 umem_free(buf, size);
 847         }
 848 }
 849
 850 /*
 851  * =========================================================================
 852  * cmn_err() and panic()
 853  * =========================================================================
 854  */
 855 static char ce_prefix[CE_IGNORE][10] = { "", "NOTICE: ", "WARNING: ", "" };
 856 static char ce_suffix[CE_IGNORE][2] = { "", "\n", "\n", "" };
 857
 858 void
 859 vpanic(const char *fmt, va_list adx)
 860 {
 861         (void) fprintf(stderr, "error: ");
 862         (void) vfprintf(stderr, fmt, adx);
 863         (void) fprintf(stderr, "\n");
 864
 865         abort();        /* think of it as a "user-level crash dump" */
 866 }
 867
 868 void
 869 panic(const char *fmt, ...)
 870 {
 871         va_list adx;
 872
 873         va_start(adx, fmt);
 874         vpanic(fmt, adx);
 875         va_end(adx);
 876 }
 877
 878 void
 879 vcmn_err(int ce, const char *fmt, va_list adx)
 880 {
 881         if (ce == CE_PANIC)
 882                 vpanic(fmt, adx);
 883         if (ce != CE_NOTE) {    /* suppress noise in userland stress testing */
 884                 (void) fprintf(stderr, "%s", ce_prefix[ce]);
 885                 (void) vfprintf(stderr, fmt, adx);
 886                 (void) fprintf(stderr, "%s", ce_suffix[ce]);
 887         }
 888 }
 889
 890 /*PRINTFLIKE2*/
 891 void
 892 cmn_err(int ce, const char *fmt, ...)
 893 {
 894         va_list adx;
 895
 896         va_start(adx, fmt);
 897         vcmn_err(ce, fmt, adx);
 898         va_end(adx);
 899 }
 900
 901 /*
 902  * =========================================================================
 903  * kobj interfaces
 904  * =========================================================================
 905  */
 906 struct _buf *
 907 kobj_open_file(char *name)
 908 {
 909         struct _buf *file;
 910         vnode_t *vp;
 911
 912         /* set vp as the _fd field of the file */
 913         if (vn_openat(name, UIO_SYSSPACE, FREAD, 0, &vp, 0, 0, rootdir,
 914             -1) != 0)
 915                 return ((void *)-1UL);
 916
 917         file = umem_zalloc(sizeof (struct _buf), UMEM_NOFAIL);
 918         file->_fd = (intptr_t)vp;
 919         return (file);
 920 }
 921
 922 int
 923 kobj_read_file(struct _buf *file, char *buf, unsigned size, unsigned off)
 924 {
 925         ssize_t resid = 0;
 926
 927         if (vn_rdwr(UIO_READ, (vnode_t *)file->_fd, buf, size, (offset_t)off,
 928             UIO_SYSSPACE, 0, 0, 0, &resid) != 0)
 929                 return (-1);
 930
 931         return (size - resid);
 932 }
 933
 934 void
 935 kobj_close_file(struct _buf *file)
 936 {
 937         vn_close((vnode_t *)file->_fd);
 938         umem_free(file, sizeof (struct _buf));
 939 }
 940
 941 int
 942 kobj_get_filesize(struct _buf *file, uint64_t *size)
 943 {
 944         struct stat64 st;
 945         vnode_t *vp = (vnode_t *)file->_fd;
 946
 947         if (fstat64(vp->v_fd, &st) == -1) {
 948                 vn_close(vp);
 949                 return (errno);
 950         }
 951         *size = st.st_size;
 952         return (0);
 953 }
 954
 955 /*
 956  * =========================================================================
 957  * misc routines
 958  * =========================================================================
 959  */
 960
 961 void
 962 delay(clock_t ticks)
 963 {
 964         (void) poll(0, 0, ticks * (1000 / hz));
 965 }
 966
 967 /*
 968  * Find highest one bit set.
 969  * Returns bit number + 1 of highest bit that is set, otherwise returns 0.
 970  * The __builtin_clzll() function is supported by both GCC and Clang.
 971  */
 972 int
 973 highbit64(uint64_t i)
 974 {
 975         if (i == 0)
 976         return (0);
 977
 978         return (NBBY * sizeof (uint64_t) - __builtin_clzll(i));
 979 }
 980
 981 /*
 982  * Find lowest one bit set.
 983  * Returns bit number + 1 of lowest bit that is set, otherwise returns 0.
 984  * The __builtin_ffsll() function is supported by both GCC and Clang.
 985  */
 986 int
 987 lowbit64(uint64_t i)
 988 {
 989         if (i == 0)
 990                 return (0);
 991
 992         return (__builtin_ffsll(i));
 993 }
 994
 995 char *random_path = "/dev/random";
 996 char *urandom_path = "/dev/urandom";
 997 static int random_fd = -1, urandom_fd = -1;
 998
 999 void
1000 random_init(void)
1001 {
1002         VERIFY((random_fd = open(random_path, O_RDONLY)) != -1);
1003         VERIFY((urandom_fd = open(urandom_path, O_RDONLY)) != -1);
1004 }
1005
1006 void
1007 random_fini(void)
1008 {
1009         close(random_fd);
1010         close(urandom_fd);
1011
1012         random_fd = -1;
1013         urandom_fd = -1;
1014 }
1015
1016 static int
1017 random_get_bytes_common(uint8_t *ptr, size_t len, int fd)
1018 {
1019         size_t resid = len;
1020         ssize_t bytes;
1021
1022         ASSERT(fd != -1);
1023
1024         while (resid != 0) {
1025                 bytes = read(fd, ptr, resid);
1026                 ASSERT3S(bytes, >=, 0);
1027                 ptr += bytes;
1028                 resid -= bytes;
1029         }
1030
1031         return (0);
1032 }
1033
1034 int
1035 random_get_bytes(uint8_t *ptr, size_t len)
1036 {
1037         return (random_get_bytes_common(ptr, len, random_fd));
1038 }
1039
1040 int
1041 random_get_pseudo_bytes(uint8_t *ptr, size_t len)
1042 {
1043         return (random_get_bytes_common(ptr, len, urandom_fd));
1044 }
1045
1046 int
1047 ddi_strtoul(const char *hw_serial, char **nptr, int base, unsigned long *result)
1048 {
1049         char *end;
1050
1051         *result = strtoul(hw_serial, &end, base);
1052         if (*result == 0)
1053                 return (errno);
1054         return (0);
1055 }
1056
1057 int
1058 ddi_strtoull(const char *str, char **nptr, int base, u_longlong_t *result)
1059 {
1060         char *end;
1061
1062         *result = strtoull(str, &end, base);
1063         if (*result == 0)
1064                 return (errno);
1065         return (0);
1066 }
1067
1068 utsname_t *
1069 utsname(void)
1070 {
1071         return (&hw_utsname);
1072 }
1073
1074 /*
1075  * =========================================================================
1076  * kernel emulation setup & teardown
1077  * =========================================================================
1078  */
1079 static int
1080 umem_out_of_memory(void)
1081 {
1082         char errmsg[] = "out of memory -- generating core dump\n";
1083
1084         (void) fprintf(stderr, "%s", errmsg);
1085         abort();
1086         return (0);
1087 }
1088
1089 void
1090 kernel_init(int mode)
1091 {
1092         extern uint_t rrw_tsd_key;
1093
1094         umem_nofail_callback(umem_out_of_memory);
1095
1096         physmem = sysconf(_SC_PHYS_PAGES);
1097
1098         dprintf("physmem = %llu pages (%.2f GB)\n", physmem,
1099             (double)physmem * sysconf(_SC_PAGE_SIZE) / (1ULL << 30));
1100
1101         (void) snprintf(hw_serial, sizeof (hw_serial), "%ld",
1102             (mode & FWRITE) ? get_system_hostid() : 0);
1103
1104         random_init();
1105
1106         VERIFY0(uname(&hw_utsname));
1107
1108         system_taskq_init();
1109         icp_init();
1110
1111         spa_init(mode);
1112
1113         fletcher_4_init();
1114
1115         tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
1116 }
1117
1118 void
1119 kernel_fini(void)
1120 {
1121         fletcher_4_fini();
1122         spa_fini();
1123
1124         icp_fini();
1125         system_taskq_fini();
1126
1127         random_fini();
1128 }
1129
1130 uid_t
1131 crgetuid(cred_t *cr)
1132 {
1133         return (0);
1134 }
1135
1136 uid_t
1137 crgetruid(cred_t *cr)
1138 {
1139         return (0);
1140 }
1141
1142 gid_t
1143 crgetgid(cred_t *cr)
1144 {
1145         return (0);
1146 }
1147
1148 int
1149 crgetngroups(cred_t *cr)
1150 {
1151         return (0);
1152 }
1153
1154 gid_t *
1155 crgetgroups(cred_t *cr)
1156 {
1157         return (NULL);
1158 }
1159
1160 int
1161 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
1162 {
1163         return (0);
1164 }
1165
1166 int
1167 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
1168 {
1169         return (0);
1170 }
1171
1172 int
1173 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
1174 {
1175         return (0);
1176 }
1177
1178 int
1179 secpolicy_zfs(const cred_t *cr)
1180 {
1181         return (0);
1182 }
1183
1184 ksiddomain_t *
1185 ksid_lookupdomain(const char *dom)
1186 {
1187         ksiddomain_t *kd;
1188
1189         kd = umem_zalloc(sizeof (ksiddomain_t), UMEM_NOFAIL);
1190         kd->kd_name = spa_strdup(dom);
1191         return (kd);
1192 }
1193
1194 void
1195 ksiddomain_rele(ksiddomain_t *ksid)
1196 {
1197         spa_strfree(ksid->kd_name);
1198         umem_free(ksid, sizeof (ksiddomain_t));
1199 }
1200
1201 char *
1202 kmem_vasprintf(const char *fmt, va_list adx)
1203 {
1204         char *buf = NULL;
1205         va_list adx_copy;
1206
1207         va_copy(adx_copy, adx);
1208         VERIFY(vasprintf(&buf, fmt, adx_copy) != -1);
1209         va_end(adx_copy);
1210
1211         return (buf);
1212 }
1213
1214 char *
1215 kmem_asprintf(const char *fmt, ...)
1216 {
1217         char *buf = NULL;
1218         va_list adx;
1219
1220         va_start(adx, fmt);
1221         VERIFY(vasprintf(&buf, fmt, adx) != -1);
1222         va_end(adx);
1223
1224         return (buf);
1225 }
1226
1227 /* ARGSUSED */
1228 int
1229 zfs_onexit_fd_hold(int fd, minor_t *minorp)
1230 {
1231         *minorp = 0;
1232         return (0);
1233 }
1234
1235 /* ARGSUSED */
1236 void
1237 zfs_onexit_fd_rele(int fd)
1238 {
1239 }
1240
1241 /* ARGSUSED */
1242 int
1243 zfs_onexit_add_cb(minor_t minor, void (*func)(void *), void *data,
1244     uint64_t *action_handle)
1245 {
1246         return (0);
1247 }
1248
1249 /* ARGSUSED */
1250 int
1251 zfs_onexit_del_cb(minor_t minor, uint64_t action_handle, boolean_t fire)
1252 {
1253         return (0);
1254 }
1255
1256 /* ARGSUSED */
1257 int
1258 zfs_onexit_cb_data(minor_t minor, uint64_t action_handle, void **data)
1259 {
1260         return (0);
1261 }
1262
1263 fstrans_cookie_t
1264 spl_fstrans_mark(void)
1265 {
1266         return ((fstrans_cookie_t)0);
1267 }
1268
1269 void
1270 spl_fstrans_unmark(fstrans_cookie_t cookie)
1271 {
1272 }
1273
1274 int
1275 __spl_pf_fstrans_check(void)
1276 {
1277         return (0);
1278 }
1279
1280 int
1281 kmem_cache_reap_active(void)
1282 {
1283         return (0);
1284 }
1285
1286 void *zvol_tag = "zvol_tag";
1287
1288 void
1289 zvol_create_minors(spa_t *spa, const char *name, boolean_t async)
1290 {
1291 }
1292
1293 void
1294 zvol_remove_minor(spa_t *spa, const char *name, boolean_t async)
1295 {
1296 }
1297
1298 void
1299 zvol_remove_minors(spa_t *spa, const char *name, boolean_t async)
1300 {
1301 }
1302
1303 void
1304 zvol_rename_minors(spa_t *spa, const char *oldname, const char *newname,
1305     boolean_t async)
1306 {
1307 }