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 int
 301 rw_tryupgrade(krwlock_t *rwlp)
 302 {
 303         return (0);
 304 }
 305
 306 /*
 307  * =========================================================================
 308  * condition variables
 309  * =========================================================================
 310  */
 311
 312 void
 313 cv_init(kcondvar_t *cv, char *name, int type, void *arg)
 314 {
 315         VERIFY0(pthread_cond_init(cv, NULL));
 316 }
 317
 318 void
 319 cv_destroy(kcondvar_t *cv)
 320 {
 321         VERIFY0(pthread_cond_destroy(cv));
 322 }
 323
 324 void
 325 cv_wait(kcondvar_t *cv, kmutex_t *mp)
 326 {
 327         memset(&mp->m_owner, 0, sizeof (pthread_t));
 328         VERIFY0(pthread_cond_wait(cv, &mp->m_lock));
 329         mp->m_owner = pthread_self();
 330 }
 331
 332 clock_t
 333 cv_timedwait(kcondvar_t *cv, kmutex_t *mp, clock_t abstime)
 334 {
 335         int error;
 336         struct timeval tv;
 337         timestruc_t ts;
 338         clock_t delta;
 339
 340         delta = abstime - ddi_get_lbolt();
 341         if (delta <= 0)
 342                 return (-1);
 343
 344         VERIFY(gettimeofday(&tv, NULL) == 0);
 345
 346         ts.tv_sec = tv.tv_sec + delta / hz;
 347         ts.tv_nsec = tv.tv_usec * NSEC_PER_USEC + (delta % hz) * (NANOSEC / hz);
 348         if (ts.tv_nsec >= NANOSEC) {
 349                 ts.tv_sec++;
 350                 ts.tv_nsec -= NANOSEC;
 351         }
 352
 353         memset(&mp->m_owner, 0, sizeof (pthread_t));
 354         error = pthread_cond_timedwait(cv, &mp->m_lock, &ts);
 355         mp->m_owner = pthread_self();
 356
 357         if (error == ETIMEDOUT)
 358                 return (-1);
 359
 360         VERIFY0(error);
 361
 362         return (1);
 363 }
 364
 365 /*ARGSUSED*/
 366 clock_t
 367 cv_timedwait_hires(kcondvar_t *cv, kmutex_t *mp, hrtime_t tim, hrtime_t res,
 368     int flag)
 369 {
 370         int error;
 371         struct timeval tv;
 372         timestruc_t ts;
 373         hrtime_t delta;
 374
 375         ASSERT(flag == 0 || flag == CALLOUT_FLAG_ABSOLUTE);
 376
 377         delta = tim;
 378         if (flag & CALLOUT_FLAG_ABSOLUTE)
 379                 delta -= gethrtime();
 380
 381         if (delta <= 0)
 382                 return (-1);
 383
 384         VERIFY0(gettimeofday(&tv, NULL));
 385
 386         ts.tv_sec = tv.tv_sec + delta / NANOSEC;
 387         ts.tv_nsec = tv.tv_usec * NSEC_PER_USEC + (delta % NANOSEC);
 388         if (ts.tv_nsec >= NANOSEC) {
 389                 ts.tv_sec++;
 390                 ts.tv_nsec -= NANOSEC;
 391         }
 392
 393         memset(&mp->m_owner, 0, sizeof (pthread_t));
 394         error = pthread_cond_timedwait(cv, &mp->m_lock, &ts);
 395         mp->m_owner = pthread_self();
 396
 397         if (error == ETIMEDOUT)
 398                 return (-1);
 399
 400         VERIFY0(error);
 401
 402         return (1);
 403 }
 404
 405 void
 406 cv_signal(kcondvar_t *cv)
 407 {
 408         VERIFY0(pthread_cond_signal(cv));
 409 }
 410
 411 void
 412 cv_broadcast(kcondvar_t *cv)
 413 {
 414         VERIFY0(pthread_cond_broadcast(cv));
 415 }
 416
 417 /*
 418  * =========================================================================
 419  * vnode operations
 420  * =========================================================================
 421  */
 422 /*
 423  * Note: for the xxxat() versions of these functions, we assume that the
 424  * starting vp is always rootdir (which is true for spa_directory.c, the only
 425  * ZFS consumer of these interfaces).  We assert this is true, and then emulate
 426  * them by adding '/' in front of the path.
 427  */
 428
 429 /*ARGSUSED*/
 430 int
 431 vn_open(char *path, int x1, int flags, int mode, vnode_t **vpp, int x2, int x3)
 432 {
 433         int fd = -1;
 434         int dump_fd = -1;
 435         vnode_t *vp;
 436         int old_umask = 0;
 437         char *realpath;
 438         struct stat64 st;
 439         int err;
 440
 441         realpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL);
 442
 443         /*
 444          * If we're accessing a real disk from userland, we need to use
 445          * the character interface to avoid caching.  This is particularly
 446          * important if we're trying to look at a real in-kernel storage
 447          * pool from userland, e.g. via zdb, because otherwise we won't
 448          * see the changes occurring under the segmap cache.
 449          * On the other hand, the stupid character device returns zero
 450          * for its size.  So -- gag -- we open the block device to get
 451          * its size, and remember it for subsequent VOP_GETATTR().
 452          */
 453 #if defined(__sun__) || defined(__sun)
 454         if (strncmp(path, "/dev/", 5) == 0) {
 455 #else
 456         if (0) {
 457 #endif
 458                 char *dsk;
 459                 fd = open64(path, O_RDONLY);
 460                 if (fd == -1) {
 461                         err = errno;
 462                         free(realpath);
 463                         return (err);
 464                 }
 465                 if (fstat64(fd, &st) == -1) {
 466                         err = errno;
 467                         close(fd);
 468                         free(realpath);
 469                         return (err);
 470                 }
 471                 close(fd);
 472                 (void) sprintf(realpath, "%s", path);
 473                 dsk = strstr(path, "/dsk/");
 474                 if (dsk != NULL)
 475                         (void) sprintf(realpath + (dsk - path) + 1, "r%s",
 476                             dsk + 1);
 477         } else {
 478                 (void) sprintf(realpath, "%s", path);
 479                 if (!(flags & FCREAT) && stat64(realpath, &st) == -1) {
 480                         err = errno;
 481                         free(realpath);
 482                         return (err);
 483                 }
 484         }
 485
 486         if (!(flags & FCREAT) && S_ISBLK(st.st_mode)) {
 487 #ifdef __linux__
 488                 flags |= O_DIRECT;
 489 #endif
 490                 /* We shouldn't be writing to block devices in userspace */
 491                 VERIFY(!(flags & FWRITE));
 492         }
 493
 494         if (flags & FCREAT)
 495                 old_umask = umask(0);
 496
 497         /*
 498          * The construct 'flags - FREAD' conveniently maps combinations of
 499          * FREAD and FWRITE to the corresponding O_RDONLY, O_WRONLY, and O_RDWR.
 500          */
 501         fd = open64(realpath, flags - FREAD, mode);
 502         if (fd == -1) {
 503                 err = errno;
 504                 free(realpath);
 505                 return (err);
 506         }
 507
 508         if (flags & FCREAT)
 509                 (void) umask(old_umask);
 510
 511         if (vn_dumpdir != NULL) {
 512                 char *dumppath = umem_zalloc(MAXPATHLEN, UMEM_NOFAIL);
 513                 (void) snprintf(dumppath, MAXPATHLEN,
 514                     "%s/%s", vn_dumpdir, basename(realpath));
 515                 dump_fd = open64(dumppath, O_CREAT | O_WRONLY, 0666);
 516                 umem_free(dumppath, MAXPATHLEN);
 517                 if (dump_fd == -1) {
 518                         err = errno;
 519                         free(realpath);
 520                         close(fd);
 521                         return (err);
 522                 }
 523         } else {
 524                 dump_fd = -1;
 525         }
 526
 527         free(realpath);
 528
 529         if (fstat64_blk(fd, &st) == -1) {
 530                 err = errno;
 531                 close(fd);
 532                 if (dump_fd != -1)
 533                         close(dump_fd);
 534                 return (err);
 535         }
 536
 537         (void) fcntl(fd, F_SETFD, FD_CLOEXEC);
 538
 539         *vpp = vp = umem_zalloc(sizeof (vnode_t), UMEM_NOFAIL);
 540
 541         vp->v_fd = fd;
 542         vp->v_size = st.st_size;
 543         vp->v_path = spa_strdup(path);
 544         vp->v_dump_fd = dump_fd;
 545
 546         return (0);
 547 }
 548
 549 /*ARGSUSED*/
 550 int
 551 vn_openat(char *path, int x1, int flags, int mode, vnode_t **vpp, int x2,
 552     int x3, vnode_t *startvp, int fd)
 553 {
 554         char *realpath = umem_alloc(strlen(path) + 2, UMEM_NOFAIL);
 555         int ret;
 556
 557         ASSERT(startvp == rootdir);
 558         (void) sprintf(realpath, "/%s", path);
 559
 560         /* fd ignored for now, need if want to simulate nbmand support */
 561         ret = vn_open(realpath, x1, flags, mode, vpp, x2, x3);
 562
 563         umem_free(realpath, strlen(path) + 2);
 564
 565         return (ret);
 566 }
 567
 568 /*ARGSUSED*/
 569 int
 570 vn_rdwr(int uio, vnode_t *vp, void *addr, ssize_t len, offset_t offset,
 571     int x1, int x2, rlim64_t x3, void *x4, ssize_t *residp)
 572 {
 573         ssize_t rc, done = 0, split;
 574
 575         if (uio == UIO_READ) {
 576                 rc = pread64(vp->v_fd, addr, len, offset);
 577                 if (vp->v_dump_fd != -1 && rc != -1) {
 578                         int status;
 579                         status = pwrite64(vp->v_dump_fd, addr, rc, offset);
 580                         ASSERT(status != -1);
 581                 }
 582         } else {
 583                 /*
 584                  * To simulate partial disk writes, we split writes into two
 585                  * system calls so that the process can be killed in between.
 586                  */
 587                 int sectors = len >> SPA_MINBLOCKSHIFT;
 588                 split = (sectors > 0 ? rand() % sectors : 0) <<
 589                     SPA_MINBLOCKSHIFT;
 590                 rc = pwrite64(vp->v_fd, addr, split, offset);
 591                 if (rc != -1) {
 592                         done = rc;
 593                         rc = pwrite64(vp->v_fd, (char *)addr + split,
 594                             len - split, offset + split);
 595                 }
 596         }
 597
 598 #ifdef __linux__
 599         if (rc == -1 && errno == EINVAL) {
 600                 /*
 601                  * Under Linux, this most likely means an alignment issue
 602                  * (memory or disk) due to O_DIRECT, so we abort() in order to
 603                  * catch the offender.
 604                  */
 605                 abort();
 606         }
 607 #endif
 608         if (rc == -1)
 609                 return (errno);
 610
 611         done += rc;
 612
 613         if (residp)
 614                 *residp = len - done;
 615         else if (done != len)
 616                 return (EIO);
 617         return (0);
 618 }
 619
 620 void
 621 vn_close(vnode_t *vp)
 622 {
 623         close(vp->v_fd);
 624         if (vp->v_dump_fd != -1)
 625                 close(vp->v_dump_fd);
 626         spa_strfree(vp->v_path);
 627         umem_free(vp, sizeof (vnode_t));
 628 }
 629
 630 /*
 631  * At a minimum we need to update the size since vdev_reopen()
 632  * will no longer call vn_openat().
 633  */
 634 int
 635 fop_getattr(vnode_t *vp, vattr_t *vap)
 636 {
 637         struct stat64 st;
 638         int err;
 639
 640         if (fstat64_blk(vp->v_fd, &st) == -1) {
 641                 err = errno;
 642                 close(vp->v_fd);
 643                 return (err);
 644         }
 645
 646         vap->va_size = st.st_size;
 647         return (0);
 648 }
 649
 650 /*
 651  * =========================================================================
 652  * Figure out which debugging statements to print
 653  * =========================================================================
 654  */
 655
 656 static char *dprintf_string;
 657 static int dprintf_print_all;
 658
 659 int
 660 dprintf_find_string(const char *string)
 661 {
 662         char *tmp_str = dprintf_string;
 663         int len = strlen(string);
 664
 665         /*
 666          * Find out if this is a string we want to print.
 667          * String format: file1.c,function_name1,file2.c,file3.c
 668          */
 669
 670         while (tmp_str != NULL) {
 671                 if (strncmp(tmp_str, string, len) == 0 &&
 672                     (tmp_str[len] == ',' || tmp_str[len] == '\0'))
 673                         return (1);
 674                 tmp_str = strchr(tmp_str, ',');
 675                 if (tmp_str != NULL)
 676                         tmp_str++; /* Get rid of , */
 677         }
 678         return (0);
 679 }
 680
 681 void
 682 dprintf_setup(int *argc, char **argv)
 683 {
 684         int i, j;
 685
 686         /*
 687          * Debugging can be specified two ways: by setting the
 688          * environment variable ZFS_DEBUG, or by including a
 689          * "debug=..."  argument on the command line.  The command
 690          * line setting overrides the environment variable.
 691          */
 692
 693         for (i = 1; i < *argc; i++) {
 694                 int len = strlen("debug=");
 695                 /* First look for a command line argument */
 696                 if (strncmp("debug=", argv[i], len) == 0) {
 697                         dprintf_string = argv[i] + len;
 698                         /* Remove from args */
 699                         for (j = i; j < *argc; j++)
 700                                 argv[j] = argv[j+1];
 701                         argv[j] = NULL;
 702                         (*argc)--;
 703                 }
 704         }
 705
 706         if (dprintf_string == NULL) {
 707                 /* Look for ZFS_DEBUG environment variable */
 708                 dprintf_string = getenv("ZFS_DEBUG");
 709         }
 710
 711         /*
 712          * Are we just turning on all debugging?
 713          */
 714         if (dprintf_find_string("on"))
 715                 dprintf_print_all = 1;
 716
 717         if (dprintf_string != NULL)
 718                 zfs_flags |= ZFS_DEBUG_DPRINTF;
 719 }
 720
 721 /*
 722  * =========================================================================
 723  * debug printfs
 724  * =========================================================================
 725  */
 726 void
 727 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
 728 {
 729         const char *newfile;
 730         va_list adx;
 731
 732         /*
 733          * Get rid of annoying "../common/" prefix to filename.
 734          */
 735         newfile = strrchr(file, '/');
 736         if (newfile != NULL) {
 737                 newfile = newfile + 1; /* Get rid of leading / */
 738         } else {
 739                 newfile = file;
 740         }
 741
 742         if (dprintf_print_all ||
 743             dprintf_find_string(newfile) ||
 744             dprintf_find_string(func)) {
 745                 /* Print out just the function name if requested */
 746                 flockfile(stdout);
 747                 if (dprintf_find_string("pid"))
 748                         (void) printf("%d ", getpid());
 749                 if (dprintf_find_string("tid"))
 750                         (void) printf("%u ", (uint_t)pthread_self());
 751                 if (dprintf_find_string("cpu"))
 752                         (void) printf("%u ", getcpuid());
 753                 if (dprintf_find_string("time"))
 754                         (void) printf("%llu ", gethrtime());
 755                 if (dprintf_find_string("long"))
 756                         (void) printf("%s, line %d: ", newfile, line);
 757                 (void) printf("%s: ", func);
 758                 va_start(adx, fmt);
 759                 (void) vprintf(fmt, adx);
 760                 va_end(adx);
 761                 funlockfile(stdout);
 762         }
 763 }
 764
 765 /*
 766  * =========================================================================
 767  * cmn_err() and panic()
 768  * =========================================================================
 769  */
 770 static char ce_prefix[CE_IGNORE][10] = { "", "NOTICE: ", "WARNING: ", "" };
 771 static char ce_suffix[CE_IGNORE][2] = { "", "\n", "\n", "" };
 772
 773 void
 774 vpanic(const char *fmt, va_list adx)
 775 {
 776         (void) fprintf(stderr, "error: ");
 777         (void) vfprintf(stderr, fmt, adx);
 778         (void) fprintf(stderr, "\n");
 779
 780         abort();        /* think of it as a "user-level crash dump" */
 781 }
 782
 783 void
 784 panic(const char *fmt, ...)
 785 {
 786         va_list adx;
 787
 788         va_start(adx, fmt);
 789         vpanic(fmt, adx);
 790         va_end(adx);
 791 }
 792
 793 void
 794 vcmn_err(int ce, const char *fmt, va_list adx)
 795 {
 796         if (ce == CE_PANIC)
 797                 vpanic(fmt, adx);
 798         if (ce != CE_NOTE) {    /* suppress noise in userland stress testing */
 799                 (void) fprintf(stderr, "%s", ce_prefix[ce]);
 800                 (void) vfprintf(stderr, fmt, adx);
 801                 (void) fprintf(stderr, "%s", ce_suffix[ce]);
 802         }
 803 }
 804
 805 /*PRINTFLIKE2*/
 806 void
 807 cmn_err(int ce, const char *fmt, ...)
 808 {
 809         va_list adx;
 810
 811         va_start(adx, fmt);
 812         vcmn_err(ce, fmt, adx);
 813         va_end(adx);
 814 }
 815
 816 /*
 817  * =========================================================================
 818  * kobj interfaces
 819  * =========================================================================
 820  */
 821 struct _buf *
 822 kobj_open_file(char *name)
 823 {
 824         struct _buf *file;
 825         vnode_t *vp;
 826
 827         /* set vp as the _fd field of the file */
 828         if (vn_openat(name, UIO_SYSSPACE, FREAD, 0, &vp, 0, 0, rootdir,
 829             -1) != 0)
 830                 return ((void *)-1UL);
 831
 832         file = umem_zalloc(sizeof (struct _buf), UMEM_NOFAIL);
 833         file->_fd = (intptr_t)vp;
 834         return (file);
 835 }
 836
 837 int
 838 kobj_read_file(struct _buf *file, char *buf, unsigned size, unsigned off)
 839 {
 840         ssize_t resid = 0;
 841
 842         if (vn_rdwr(UIO_READ, (vnode_t *)file->_fd, buf, size, (offset_t)off,
 843             UIO_SYSSPACE, 0, 0, 0, &resid) != 0)
 844                 return (-1);
 845
 846         return (size - resid);
 847 }
 848
 849 void
 850 kobj_close_file(struct _buf *file)
 851 {
 852         vn_close((vnode_t *)file->_fd);
 853         umem_free(file, sizeof (struct _buf));
 854 }
 855
 856 int
 857 kobj_get_filesize(struct _buf *file, uint64_t *size)
 858 {
 859         struct stat64 st;
 860         vnode_t *vp = (vnode_t *)file->_fd;
 861
 862         if (fstat64(vp->v_fd, &st) == -1) {
 863                 vn_close(vp);
 864                 return (errno);
 865         }
 866         *size = st.st_size;
 867         return (0);
 868 }
 869
 870 /*
 871  * =========================================================================
 872  * misc routines
 873  * =========================================================================
 874  */
 875
 876 void
 877 delay(clock_t ticks)
 878 {
 879         (void) poll(0, 0, ticks * (1000 / hz));
 880 }
 881
 882 /*
 883  * Find highest one bit set.
 884  * Returns bit number + 1 of highest bit that is set, otherwise returns 0.
 885  * The __builtin_clzll() function is supported by both GCC and Clang.
 886  */
 887 int
 888 highbit64(uint64_t i)
 889 {
 890         if (i == 0)
 891         return (0);
 892
 893         return (NBBY * sizeof (uint64_t) - __builtin_clzll(i));
 894 }
 895
 896 /*
 897  * Find lowest one bit set.
 898  * Returns bit number + 1 of lowest bit that is set, otherwise returns 0.
 899  * The __builtin_ffsll() function is supported by both GCC and Clang.
 900  */
 901 int
 902 lowbit64(uint64_t i)
 903 {
 904         if (i == 0)
 905                 return (0);
 906
 907         return (__builtin_ffsll(i));
 908 }
 909
 910 static int random_fd = -1, urandom_fd = -1;
 911
 912 void
 913 random_init(void)
 914 {
 915         VERIFY((random_fd = open("/dev/random", O_RDONLY)) != -1);
 916         VERIFY((urandom_fd = open("/dev/urandom", O_RDONLY)) != -1);
 917 }
 918
 919 void
 920 random_fini(void)
 921 {
 922         close(random_fd);
 923         close(urandom_fd);
 924
 925         random_fd = -1;
 926         urandom_fd = -1;
 927 }
 928
 929 static int
 930 random_get_bytes_common(uint8_t *ptr, size_t len, int fd)
 931 {
 932         size_t resid = len;
 933         ssize_t bytes;
 934
 935         ASSERT(fd != -1);
 936
 937         while (resid != 0) {
 938                 bytes = read(fd, ptr, resid);
 939                 ASSERT3S(bytes, >=, 0);
 940                 ptr += bytes;
 941                 resid -= bytes;
 942         }
 943
 944         return (0);
 945 }
 946
 947 int
 948 random_get_bytes(uint8_t *ptr, size_t len)
 949 {
 950         return (random_get_bytes_common(ptr, len, random_fd));
 951 }
 952
 953 int
 954 random_get_pseudo_bytes(uint8_t *ptr, size_t len)
 955 {
 956         return (random_get_bytes_common(ptr, len, urandom_fd));
 957 }
 958
 959 int
 960 ddi_strtoul(const char *hw_serial, char **nptr, int base, unsigned long *result)
 961 {
 962         char *end;
 963
 964         *result = strtoul(hw_serial, &end, base);
 965         if (*result == 0)
 966                 return (errno);
 967         return (0);
 968 }
 969
 970 int
 971 ddi_strtoull(const char *str, char **nptr, int base, u_longlong_t *result)
 972 {
 973         char *end;
 974
 975         *result = strtoull(str, &end, base);
 976         if (*result == 0)
 977                 return (errno);
 978         return (0);
 979 }
 980
 981 utsname_t *
 982 utsname(void)
 983 {
 984         return (&hw_utsname);
 985 }
 986
 987 /*
 988  * =========================================================================
 989  * kernel emulation setup & teardown
 990  * =========================================================================
 991  */
 992 static int
 993 umem_out_of_memory(void)
 994 {
 995         char errmsg[] = "out of memory -- generating core dump\n";
 996
 997         (void) fprintf(stderr, "%s", errmsg);
 998         abort();
 999         return (0);
1000 }
1001
1002 void
1003 kernel_init(int mode)
1004 {
1005         extern uint_t rrw_tsd_key;
1006
1007         umem_nofail_callback(umem_out_of_memory);
1008
1009         physmem = sysconf(_SC_PHYS_PAGES);
1010
1011         dprintf("physmem = %llu pages (%.2f GB)\n", physmem,
1012             (double)physmem * sysconf(_SC_PAGE_SIZE) / (1ULL << 30));
1013
1014         (void) snprintf(hw_serial, sizeof (hw_serial), "%ld",
1015             (mode & FWRITE) ? get_system_hostid() : 0);
1016
1017         random_init();
1018
1019         VERIFY0(uname(&hw_utsname));
1020
1021         system_taskq_init();
1022         icp_init();
1023
1024         spa_init(mode);
1025
1026         fletcher_4_init();
1027
1028         tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
1029 }
1030
1031 void
1032 kernel_fini(void)
1033 {
1034         fletcher_4_fini();
1035         spa_fini();
1036
1037         icp_fini();
1038         system_taskq_fini();
1039
1040         random_fini();
1041 }
1042
1043 uid_t
1044 crgetuid(cred_t *cr)
1045 {
1046         return (0);
1047 }
1048
1049 uid_t
1050 crgetruid(cred_t *cr)
1051 {
1052         return (0);
1053 }
1054
1055 gid_t
1056 crgetgid(cred_t *cr)
1057 {
1058         return (0);
1059 }
1060
1061 int
1062 crgetngroups(cred_t *cr)
1063 {
1064         return (0);
1065 }
1066
1067 gid_t *
1068 crgetgroups(cred_t *cr)
1069 {
1070         return (NULL);
1071 }
1072
1073 int
1074 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
1075 {
1076         return (0);
1077 }
1078
1079 int
1080 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
1081 {
1082         return (0);
1083 }
1084
1085 int
1086 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
1087 {
1088         return (0);
1089 }
1090
1091 int
1092 secpolicy_zfs(const cred_t *cr)
1093 {
1094         return (0);
1095 }
1096
1097 ksiddomain_t *
1098 ksid_lookupdomain(const char *dom)
1099 {
1100         ksiddomain_t *kd;
1101
1102         kd = umem_zalloc(sizeof (ksiddomain_t), UMEM_NOFAIL);
1103         kd->kd_name = spa_strdup(dom);
1104         return (kd);
1105 }
1106
1107 void
1108 ksiddomain_rele(ksiddomain_t *ksid)
1109 {
1110         spa_strfree(ksid->kd_name);
1111         umem_free(ksid, sizeof (ksiddomain_t));
1112 }
1113
1114 char *
1115 kmem_vasprintf(const char *fmt, va_list adx)
1116 {
1117         char *buf = NULL;
1118         va_list adx_copy;
1119
1120         va_copy(adx_copy, adx);
1121         VERIFY(vasprintf(&buf, fmt, adx_copy) != -1);
1122         va_end(adx_copy);
1123
1124         return (buf);
1125 }
1126
1127 char *
1128 kmem_asprintf(const char *fmt, ...)
1129 {
1130         char *buf = NULL;
1131         va_list adx;
1132
1133         va_start(adx, fmt);
1134         VERIFY(vasprintf(&buf, fmt, adx) != -1);
1135         va_end(adx);
1136
1137         return (buf);
1138 }
1139
1140 /* ARGSUSED */
1141 int
1142 zfs_onexit_fd_hold(int fd, minor_t *minorp)
1143 {
1144         *minorp = 0;
1145         return (0);
1146 }
1147
1148 /* ARGSUSED */
1149 void
1150 zfs_onexit_fd_rele(int fd)
1151 {
1152 }
1153
1154 /* ARGSUSED */
1155 int
1156 zfs_onexit_add_cb(minor_t minor, void (*func)(void *), void *data,
1157     uint64_t *action_handle)
1158 {
1159         return (0);
1160 }
1161
1162 /* ARGSUSED */
1163 int
1164 zfs_onexit_del_cb(minor_t minor, uint64_t action_handle, boolean_t fire)
1165 {
1166         return (0);
1167 }
1168
1169 /* ARGSUSED */
1170 int
1171 zfs_onexit_cb_data(minor_t minor, uint64_t action_handle, void **data)
1172 {
1173         return (0);
1174 }
1175
1176 fstrans_cookie_t
1177 spl_fstrans_mark(void)
1178 {
1179         return ((fstrans_cookie_t)0);
1180 }
1181
1182 void
1183 spl_fstrans_unmark(fstrans_cookie_t cookie)
1184 {
1185 }
1186
1187 int
1188 __spl_pf_fstrans_check(void)
1189 {
1190         return (0);
1191 }
1192
1193 void *zvol_tag = "zvol_tag";
1194
1195 void
1196 zvol_create_minors(spa_t *spa, const char *name, boolean_t async)
1197 {
1198 }
1199
1200 void
1201 zvol_remove_minor(spa_t *spa, const char *name, boolean_t async)
1202 {
1203 }
1204
1205 void
1206 zvol_remove_minors(spa_t *spa, const char *name, boolean_t async)
1207 {
1208 }
1209
1210 void
1211 zvol_rename_minors(spa_t *spa, const char *oldname, const char *newname,
1212     boolean_t async)
1213 {
1214 }