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