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  */
  24
  25 #include <assert.h>
  26 #include <fcntl.h>
  27 #include <poll.h>
  28 #include <stdio.h>
  29 #include <stdlib.h>
  30 #include <string.h>
  31 #include <zlib.h>
  32 #include <sys/signal.h>
  33 #include <sys/spa.h>
  34 #include <sys/stat.h>
  35 #include <sys/processor.h>
  36 #include <sys/zfs_context.h>
  37 #include <sys/utsname.h>
  38 #include <sys/time.h>
  39 #include <sys/mount.h> /* for BLKGETSIZE64 */
  40 #include <sys/systeminfo.h>
  41
  42 /*
  43  * Emulation of kernel services in userland.
  44  */
  45
  46 int aok;
  47 uint64_t physmem;
  48 vnode_t *rootdir = (vnode_t *)0xabcd1234;
  49 char hw_serial[HW_HOSTID_LEN];
  50
  51 struct utsname utsname = {
  52         "userland", "libzpool", "1", "1", "na"
  53 };
  54
  55 /* this only exists to have its address taken */
  56 struct proc p0;
  57
  58 /*
  59  * =========================================================================
  60  * threads
  61  * =========================================================================
  62  */
  63
  64 pthread_cond_t kthread_cond = PTHREAD_COND_INITIALIZER;
  65 pthread_mutex_t kthread_lock = PTHREAD_MUTEX_INITIALIZER;
  66 pthread_key_t kthread_key;
  67 int kthread_nr = 0;
  68
  69 static void
  70 thread_init(void)
  71 {
  72         kthread_t *kt;
  73
  74         VERIFY3S(pthread_key_create(&kthread_key, NULL), ==, 0);
  75
  76         /* Create entry for primary kthread */
  77         kt = umem_zalloc(sizeof(kthread_t), UMEM_NOFAIL);
  78         kt->t_tid = pthread_self();
  79         kt->t_func = NULL;
  80
  81         VERIFY3S(pthread_setspecific(kthread_key, kt), ==, 0);
  82
  83         /* Only the main thread should be running at the moment */
  84         ASSERT3S(kthread_nr, ==, 0);
  85         kthread_nr = 1;
  86 }
  87
  88 static void
  89 thread_fini(void)
  90 {
  91         kthread_t *kt = curthread;
  92
  93         ASSERT(pthread_equal(kt->t_tid, pthread_self()));
  94         ASSERT3P(kt->t_func, ==, NULL);
  95
  96         umem_free(kt, sizeof(kthread_t));
  97
  98         /* Wait for all threads to exit via thread_exit() */
  99         VERIFY3S(pthread_mutex_lock(&kthread_lock), ==, 0);
 100
 101         kthread_nr--; /* Main thread is exiting */
 102
 103         while (kthread_nr > 0)
 104                 VERIFY3S(pthread_cond_wait(&kthread_cond, &kthread_lock), ==,
 105                     0);
 106
 107         ASSERT3S(kthread_nr, ==, 0);
 108         VERIFY3S(pthread_mutex_unlock(&kthread_lock), ==, 0);
 109
 110         VERIFY3S(pthread_key_delete(kthread_key), ==, 0);
 111 }
 112
 113 kthread_t *
 114 zk_thread_current(void)
 115 {
 116         kthread_t *kt = pthread_getspecific(kthread_key);
 117
 118         ASSERT3P(kt, !=, NULL);
 119
 120         return kt;
 121 }
 122
 123 void *
 124 zk_thread_helper(void *arg)
 125 {
 126         kthread_t *kt = (kthread_t *) arg;
 127
 128         VERIFY3S(pthread_setspecific(kthread_key, kt), ==, 0);
 129
 130         VERIFY3S(pthread_mutex_lock(&kthread_lock), ==, 0);
 131         kthread_nr++;
 132         VERIFY3S(pthread_mutex_unlock(&kthread_lock), ==, 0);
 133
 134         kt->t_tid = pthread_self();
 135         ((thread_func_arg_t) kt->t_func)(kt->t_arg);
 136
 137         /* Unreachable, thread must exit with thread_exit() */
 138         abort();
 139
 140         return NULL;
 141 }
 142
 143 kthread_t *
 144 zk_thread_create(caddr_t stk, size_t stksize, thread_func_t func, void *arg,
 145               size_t len, proc_t *pp, int state, pri_t pri)
 146 {
 147         kthread_t *kt;
 148         pthread_attr_t attr;
 149         size_t stack;
 150
 151         ASSERT3S(state & ~TS_RUN, ==, 0);
 152
 153         kt = umem_zalloc(sizeof(kthread_t), UMEM_NOFAIL);
 154         kt->t_func = func;
 155         kt->t_arg = arg;
 156
 157         /*
 158          * The Solaris kernel stack size is 24k for x86/x86_64.
 159          * The Linux kernel stack size is 8k for x86/x86_64.
 160          *
 161          * We reduce the default stack size in userspace, to ensure
 162          * we observe stack overruns in user space as well as in
 163          * kernel space.  PTHREAD_STACK_MIN is the minimum stack
 164          * required for a NULL procedure in user space and is added
 165          * in to the stack requirements.
 166          *
 167          * Some buggy NPTL threading implementations include the
 168          * guard area within the stack size allocations.  In
 169          * this case we allocate an extra page to account for the
 170          * guard area since we only have two pages of usable stack
 171          * on Linux.
 172          */
 173
 174         stack = PTHREAD_STACK_MIN + MAX(stksize, STACK_SIZE) +
 175                         EXTRA_GUARD_BYTES;
 176
 177         VERIFY3S(pthread_attr_init(&attr), ==, 0);
 178         VERIFY3S(pthread_attr_setstacksize(&attr, stack), ==, 0);
 179         VERIFY3S(pthread_attr_setguardsize(&attr, PAGESIZE), ==, 0);
 180
 181         VERIFY3S(pthread_create(&kt->t_tid, &attr, &zk_thread_helper, kt),
 182             ==, 0);
 183
 184         VERIFY3S(pthread_attr_destroy(&attr), ==, 0);
 185
 186         return kt;
 187 }
 188
 189 void
 190 zk_thread_exit(void)
 191 {
 192         kthread_t *kt = curthread;
 193
 194         ASSERT(pthread_equal(kt->t_tid, pthread_self()));
 195
 196         umem_free(kt, sizeof(kthread_t));
 197
 198         pthread_mutex_lock(&kthread_lock);
 199         kthread_nr--;
 200         pthread_mutex_unlock(&kthread_lock);
 201
 202         pthread_cond_broadcast(&kthread_cond);
 203         pthread_exit((void *)TS_MAGIC);
 204 }
 205
 206 void
 207 zk_thread_join(kt_did_t tid)
 208 {
 209         void *ret;
 210
 211         pthread_join((pthread_t)tid, &ret);
 212         VERIFY3P(ret, ==, (void *)TS_MAGIC);
 213 }
 214
 215 /*
 216  * =========================================================================
 217  * kstats
 218  * =========================================================================
 219  */
 220 /*ARGSUSED*/
 221 kstat_t *
 222 kstat_create(char *module, int instance, char *name, char *class,
 223     uchar_t type, ulong_t ndata, uchar_t ks_flag)
 224 {
 225         return (NULL);
 226 }
 227
 228 /*ARGSUSED*/
 229 void
 230 kstat_install(kstat_t *ksp)
 231 {}
 232
 233 /*ARGSUSED*/
 234 void
 235 kstat_delete(kstat_t *ksp)
 236 {}
 237
 238 /*
 239  * =========================================================================
 240  * mutexes
 241  * =========================================================================
 242  */
 243
 244 void
 245 mutex_init(kmutex_t *mp, char *name, int type, void *cookie)
 246 {
 247         ASSERT3S(type, ==, MUTEX_DEFAULT);
 248         ASSERT3P(cookie, ==, NULL);
 249         mp->m_owner = MTX_INIT;
 250         mp->m_magic = MTX_MAGIC;
 251         VERIFY3S(pthread_mutex_init(&mp->m_lock, NULL), ==, 0);
 252 }
 253
 254 void
 255 mutex_destroy(kmutex_t *mp)
 256 {
 257         ASSERT3U(mp->m_magic, ==, MTX_MAGIC);
 258         ASSERT3P(mp->m_owner, ==, MTX_INIT);
 259         VERIFY3S(pthread_mutex_destroy(&(mp)->m_lock), ==, 0);
 260         mp->m_owner = MTX_DEST;
 261         mp->m_magic = 0;
 262 }
 263
 264 void
 265 mutex_enter(kmutex_t *mp)
 266 {
 267         ASSERT3U(mp->m_magic, ==, MTX_MAGIC);
 268         ASSERT3P(mp->m_owner, !=, MTX_DEST);
 269         ASSERT3P(mp->m_owner, !=, curthread);
 270         VERIFY3S(pthread_mutex_lock(&mp->m_lock), ==, 0);
 271         ASSERT3P(mp->m_owner, ==, MTX_INIT);
 272         mp->m_owner = curthread;
 273 }
 274
 275 int
 276 mutex_tryenter(kmutex_t *mp)
 277 {
 278         ASSERT3U(mp->m_magic, ==, MTX_MAGIC);
 279         ASSERT3P(mp->m_owner, !=, MTX_DEST);
 280         if (0 == pthread_mutex_trylock(&mp->m_lock)) {
 281                 ASSERT3P(mp->m_owner, ==, MTX_INIT);
 282                 mp->m_owner = curthread;
 283                 return (1);
 284         } else {
 285                 return (0);
 286         }
 287 }
 288
 289 void
 290 mutex_exit(kmutex_t *mp)
 291 {
 292         ASSERT3U(mp->m_magic, ==, MTX_MAGIC);
 293         ASSERT3P(mutex_owner(mp), ==, curthread);
 294         mp->m_owner = MTX_INIT;
 295         VERIFY3S(pthread_mutex_unlock(&mp->m_lock), ==, 0);
 296 }
 297
 298 void *
 299 mutex_owner(kmutex_t *mp)
 300 {
 301         ASSERT3U(mp->m_magic, ==, MTX_MAGIC);
 302         return (mp->m_owner);
 303 }
 304
 305 int
 306 mutex_held(kmutex_t *mp)
 307 {
 308         return (mp->m_owner == curthread);
 309 }
 310
 311 /*
 312  * =========================================================================
 313  * rwlocks
 314  * =========================================================================
 315  */
 316
 317 void
 318 rw_init(krwlock_t *rwlp, char *name, int type, void *arg)
 319 {
 320         ASSERT3S(type, ==, RW_DEFAULT);
 321         ASSERT3P(arg, ==, NULL);
 322         VERIFY3S(pthread_rwlock_init(&rwlp->rw_lock, NULL), ==, 0);
 323         rwlp->rw_owner = RW_INIT;
 324         rwlp->rw_wr_owner = RW_INIT;
 325         rwlp->rw_readers = 0;
 326         rwlp->rw_magic = RW_MAGIC;
 327 }
 328
 329 void
 330 rw_destroy(krwlock_t *rwlp)
 331 {
 332         ASSERT3U(rwlp->rw_magic, ==, RW_MAGIC);
 333
 334         VERIFY3S(pthread_rwlock_destroy(&rwlp->rw_lock), ==, 0);
 335         rwlp->rw_magic = 0;
 336 }
 337
 338 void
 339 rw_enter(krwlock_t *rwlp, krw_t rw)
 340 {
 341         ASSERT3U(rwlp->rw_magic, ==, RW_MAGIC);
 342         ASSERT3P(rwlp->rw_owner, !=, curthread);
 343         ASSERT3P(rwlp->rw_wr_owner, !=, curthread);
 344
 345         if (rw == RW_READER) {
 346                 VERIFY3S(pthread_rwlock_rdlock(&rwlp->rw_lock), ==, 0);
 347                 ASSERT3P(rwlp->rw_wr_owner, ==, RW_INIT);
 348
 349                 atomic_inc_uint(&rwlp->rw_readers);
 350         } else {
 351                 VERIFY3S(pthread_rwlock_wrlock(&rwlp->rw_lock), ==, 0);
 352                 ASSERT3P(rwlp->rw_wr_owner, ==, RW_INIT);
 353                 ASSERT3U(rwlp->rw_readers, ==, 0);
 354
 355                 rwlp->rw_wr_owner = curthread;
 356         }
 357
 358         rwlp->rw_owner = curthread;
 359 }
 360
 361 void
 362 rw_exit(krwlock_t *rwlp)
 363 {
 364         ASSERT3U(rwlp->rw_magic, ==, RW_MAGIC);
 365         ASSERT(RW_LOCK_HELD(rwlp));
 366
 367         if (RW_READ_HELD(rwlp))
 368                 atomic_dec_uint(&rwlp->rw_readers);
 369         else
 370                 rwlp->rw_wr_owner = RW_INIT;
 371
 372         rwlp->rw_owner = RW_INIT;
 373         VERIFY3S(pthread_rwlock_unlock(&rwlp->rw_lock), ==, 0);
 374 }
 375
 376 int
 377 rw_tryenter(krwlock_t *rwlp, krw_t rw)
 378 {
 379         int rv;
 380
 381         ASSERT3U(rwlp->rw_magic, ==, RW_MAGIC);
 382
 383         if (rw == RW_READER)
 384                 rv = pthread_rwlock_tryrdlock(&rwlp->rw_lock);
 385         else
 386                 rv = pthread_rwlock_trywrlock(&rwlp->rw_lock);
 387
 388         if (rv == 0) {
 389                 ASSERT3P(rwlp->rw_wr_owner, ==, RW_INIT);
 390
 391                 if (rw == RW_READER)
 392                         atomic_inc_uint(&rwlp->rw_readers);
 393                 else {
 394                         ASSERT3U(rwlp->rw_readers, ==, 0);
 395                         rwlp->rw_wr_owner = curthread;
 396                 }
 397
 398                 rwlp->rw_owner = curthread;
 399                 return (1);
 400         }
 401
 402         VERIFY3S(rv, ==, EBUSY);
 403
 404         return (0);
 405 }
 406
 407 int
 408 rw_tryupgrade(krwlock_t *rwlp)
 409 {
 410         ASSERT3U(rwlp->rw_magic, ==, RW_MAGIC);
 411
 412         return (0);
 413 }
 414
 415 /*
 416  * =========================================================================
 417  * condition variables
 418  * =========================================================================
 419  */
 420
 421 void
 422 cv_init(kcondvar_t *cv, char *name, int type, void *arg)
 423 {
 424         ASSERT3S(type, ==, CV_DEFAULT);
 425         cv->cv_magic = CV_MAGIC;
 426         VERIFY3S(pthread_cond_init(&cv->cv, NULL), ==, 0);
 427 }
 428
 429 void
 430 cv_destroy(kcondvar_t *cv)
 431 {
 432         ASSERT3U(cv->cv_magic, ==, CV_MAGIC);
 433         VERIFY3S(pthread_cond_destroy(&cv->cv), ==, 0);
 434         cv->cv_magic = 0;
 435 }
 436
 437 void
 438 cv_wait(kcondvar_t *cv, kmutex_t *mp)
 439 {
 440         ASSERT3U(cv->cv_magic, ==, CV_MAGIC);
 441         ASSERT3P(mutex_owner(mp), ==, curthread);
 442         mp->m_owner = MTX_INIT;
 443         int ret = pthread_cond_wait(&cv->cv, &mp->m_lock);
 444         if (ret != 0)
 445                 VERIFY3S(ret, ==, EINTR);
 446         mp->m_owner = curthread;
 447 }
 448
 449 clock_t
 450 cv_timedwait(kcondvar_t *cv, kmutex_t *mp, clock_t abstime)
 451 {
 452         int error;
 453         struct timeval tv;
 454         timestruc_t ts;
 455         clock_t delta;
 456
 457         ASSERT3U(cv->cv_magic, ==, CV_MAGIC);
 458
 459 top:
 460         delta = abstime - ddi_get_lbolt();
 461         if (delta <= 0)
 462                 return (-1);
 463
 464         VERIFY(gettimeofday(&tv, NULL) == 0);
 465
 466         ts.tv_sec = tv.tv_sec + delta / hz;
 467         ts.tv_nsec = tv.tv_usec * 1000 + (delta % hz) * (NANOSEC / hz);
 468         if (ts.tv_nsec >= NANOSEC) {
 469                 ts.tv_sec++;
 470                 ts.tv_nsec -= NANOSEC;
 471         }
 472
 473         ASSERT3P(mutex_owner(mp), ==, curthread);
 474         mp->m_owner = MTX_INIT;
 475         error = pthread_cond_timedwait(&cv->cv, &mp->m_lock, &ts);
 476         mp->m_owner = curthread;
 477
 478         if (error == ETIMEDOUT)
 479                 return (-1);
 480
 481         if (error == EINTR)
 482                 goto top;
 483
 484         VERIFY3S(error, ==, 0);
 485
 486         return (1);
 487 }
 488
 489 void
 490 cv_signal(kcondvar_t *cv)
 491 {
 492         ASSERT3U(cv->cv_magic, ==, CV_MAGIC);
 493         VERIFY3S(pthread_cond_signal(&cv->cv), ==, 0);
 494 }
 495
 496 void
 497 cv_broadcast(kcondvar_t *cv)
 498 {
 499         ASSERT3U(cv->cv_magic, ==, CV_MAGIC);
 500         VERIFY3S(pthread_cond_broadcast(&cv->cv), ==, 0);
 501 }
 502
 503 /*
 504  * =========================================================================
 505  * vnode operations
 506  * =========================================================================
 507  */
 508 /*
 509  * Note: for the xxxat() versions of these functions, we assume that the
 510  * starting vp is always rootdir (which is true for spa_directory.c, the only
 511  * ZFS consumer of these interfaces).  We assert this is true, and then emulate
 512  * them by adding '/' in front of the path.
 513  */
 514
 515 /*ARGSUSED*/
 516 int
 517 vn_open(char *path, int x1, int flags, int mode, vnode_t **vpp, int x2, int x3)
 518 {
 519         int fd;
 520         vnode_t *vp;
 521         int old_umask;
 522         char *realpath;
 523         struct stat64 st;
 524         int err;
 525
 526         realpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL);
 527
 528         /*
 529          * If we're accessing a real disk from userland, we need to use
 530          * the character interface to avoid caching.  This is particularly
 531          * important if we're trying to look at a real in-kernel storage
 532          * pool from userland, e.g. via zdb, because otherwise we won't
 533          * see the changes occurring under the segmap cache.
 534          * On the other hand, the stupid character device returns zero
 535          * for its size.  So -- gag -- we open the block device to get
 536          * its size, and remember it for subsequent VOP_GETATTR().
 537          */
 538 #if defined(__sun__) || defined(__sun)
 539         if (strncmp(path, "/dev/", 5) == 0) {
 540 #else
 541         if (0) {
 542 #endif
 543                 char *dsk;
 544                 fd = open64(path, O_RDONLY);
 545                 if (fd == -1) {
 546                         err = errno;
 547                         free(realpath);
 548                         return (err);
 549                 }
 550                 if (fstat64(fd, &st) == -1) {
 551                         err = errno;
 552                         close(fd);
 553                         free(realpath);
 554                         return (err);
 555                 }
 556                 close(fd);
 557                 (void) sprintf(realpath, "%s", path);
 558                 dsk = strstr(path, "/dsk/");
 559                 if (dsk != NULL)
 560                         (void) sprintf(realpath + (dsk - path) + 1, "r%s",
 561                             dsk + 1);
 562         } else {
 563                 (void) sprintf(realpath, "%s", path);
 564                 if (!(flags & FCREAT) && stat64(realpath, &st) == -1) {
 565                         err = errno;
 566                         free(realpath);
 567                         return (err);
 568                 }
 569         }
 570
 571         if (!(flags & FCREAT) && S_ISBLK(st.st_mode)) {
 572 #ifdef __linux__
 573                 flags |= O_DIRECT;
 574 #endif
 575                 /* We shouldn't be writing to block devices in userspace */
 576                 VERIFY(!(flags & FWRITE));
 577         }
 578
 579         if (flags & FCREAT)
 580                 old_umask = umask(0);
 581
 582         /*
 583          * The construct 'flags - FREAD' conveniently maps combinations of
 584          * FREAD and FWRITE to the corresponding O_RDONLY, O_WRONLY, and O_RDWR.
 585          */
 586         fd = open64(realpath, flags - FREAD, mode);
 587         free(realpath);
 588
 589         if (flags & FCREAT)
 590                 (void) umask(old_umask);
 591
 592         if (fd == -1)
 593                 return (errno);
 594
 595         if (fstat64(fd, &st) == -1) {
 596                 err = errno;
 597                 close(fd);
 598                 return (err);
 599         }
 600
 601 #ifdef __linux__
 602         /* In Linux, use an ioctl to get the size of a block device. */
 603         if (S_ISBLK(st.st_mode)) {
 604                 if (ioctl(fd, BLKGETSIZE64, &st.st_size) != 0) {
 605                         err = errno;
 606                         close(fd);
 607                         return (err);
 608                 }
 609         }
 610 #endif
 611         (void) fcntl(fd, F_SETFD, FD_CLOEXEC);
 612
 613         *vpp = vp = umem_zalloc(sizeof (vnode_t), UMEM_NOFAIL);
 614
 615         vp->v_fd = fd;
 616         vp->v_size = st.st_size;
 617         vp->v_path = spa_strdup(path);
 618
 619         return (0);
 620 }
 621
 622 /*ARGSUSED*/
 623 int
 624 vn_openat(char *path, int x1, int flags, int mode, vnode_t **vpp, int x2,
 625     int x3, vnode_t *startvp, int fd)
 626 {
 627         char *realpath = umem_alloc(strlen(path) + 2, UMEM_NOFAIL);
 628         int ret;
 629
 630         ASSERT(startvp == rootdir);
 631         (void) sprintf(realpath, "/%s", path);
 632
 633         /* fd ignored for now, need if want to simulate nbmand support */
 634         ret = vn_open(realpath, x1, flags, mode, vpp, x2, x3);
 635
 636         umem_free(realpath, strlen(path) + 2);
 637
 638         return (ret);
 639 }
 640
 641 /*ARGSUSED*/
 642 int
 643 vn_rdwr(int uio, vnode_t *vp, void *addr, ssize_t len, offset_t offset,
 644         int x1, int x2, rlim64_t x3, void *x4, ssize_t *residp)
 645 {
 646         ssize_t rc, done = 0, split;
 647
 648         if (uio == UIO_READ) {
 649                 rc = pread64(vp->v_fd, addr, len, offset);
 650         } else {
 651                 /*
 652                  * To simulate partial disk writes, we split writes into two
 653                  * system calls so that the process can be killed in between.
 654                  */
 655                 split = (len > 0 ? rand() % len : 0);
 656                 rc = pwrite64(vp->v_fd, addr, split, offset);
 657                 if (rc != -1) {
 658                         done = rc;
 659                         rc = pwrite64(vp->v_fd, (char *)addr + split,
 660                             len - split, offset + split);
 661                 }
 662         }
 663
 664 #ifdef __linux__
 665         if (rc == -1 && errno == EINVAL) {
 666                 /*
 667                  * Under Linux, this most likely means an alignment issue
 668                  * (memory or disk) due to O_DIRECT, so we abort() in order to
 669                  * catch the offender.
 670                  */
 671                  abort();
 672         }
 673 #endif
 674         if (rc == -1)
 675                 return (errno);
 676
 677         done += rc;
 678
 679         if (residp)
 680                 *residp = len - done;
 681         else if (done != len)
 682                 return (EIO);
 683         return (0);
 684 }
 685
 686 void
 687 vn_close(vnode_t *vp)
 688 {
 689         close(vp->v_fd);
 690         spa_strfree(vp->v_path);
 691         umem_free(vp, sizeof (vnode_t));
 692 }
 693
 694 /*
 695  * At a minimum we need to update the size since vdev_reopen()
 696  * will no longer call vn_openat().
 697  */
 698 int
 699 fop_getattr(vnode_t *vp, vattr_t *vap)
 700 {
 701         struct stat64 st;
 702
 703         if (fstat64(vp->v_fd, &st) == -1) {
 704                 close(vp->v_fd);
 705                 return (errno);
 706         }
 707
 708         vap->va_size = st.st_size;
 709         return (0);
 710 }
 711
 712 #ifdef ZFS_DEBUG
 713
 714 /*
 715  * =========================================================================
 716  * Figure out which debugging statements to print
 717  * =========================================================================
 718  */
 719
 720 static char *dprintf_string;
 721 static int dprintf_print_all;
 722
 723 int
 724 dprintf_find_string(const char *string)
 725 {
 726         char *tmp_str = dprintf_string;
 727         int len = strlen(string);
 728
 729         /*
 730          * Find out if this is a string we want to print.
 731          * String format: file1.c,function_name1,file2.c,file3.c
 732          */
 733
 734         while (tmp_str != NULL) {
 735                 if (strncmp(tmp_str, string, len) == 0 &&
 736                     (tmp_str[len] == ',' || tmp_str[len] == '\0'))
 737                         return (1);
 738                 tmp_str = strchr(tmp_str, ',');
 739                 if (tmp_str != NULL)
 740                         tmp_str++; /* Get rid of , */
 741         }
 742         return (0);
 743 }
 744
 745 void
 746 dprintf_setup(int *argc, char **argv)
 747 {
 748         int i, j;
 749
 750         /*
 751          * Debugging can be specified two ways: by setting the
 752          * environment variable ZFS_DEBUG, or by including a
 753          * "debug=..."  argument on the command line.  The command
 754          * line setting overrides the environment variable.
 755          */
 756
 757         for (i = 1; i < *argc; i++) {
 758                 int len = strlen("debug=");
 759                 /* First look for a command line argument */
 760                 if (strncmp("debug=", argv[i], len) == 0) {
 761                         dprintf_string = argv[i] + len;
 762                         /* Remove from args */
 763                         for (j = i; j < *argc; j++)
 764                                 argv[j] = argv[j+1];
 765                         argv[j] = NULL;
 766                         (*argc)--;
 767                 }
 768         }
 769
 770         if (dprintf_string == NULL) {
 771                 /* Look for ZFS_DEBUG environment variable */
 772                 dprintf_string = getenv("ZFS_DEBUG");
 773         }
 774
 775         /*
 776          * Are we just turning on all debugging?
 777          */
 778         if (dprintf_find_string("on"))
 779                 dprintf_print_all = 1;
 780 }
 781
 782 /*
 783  * =========================================================================
 784  * debug printfs
 785  * =========================================================================
 786  */
 787 void
 788 __dprintf(const char *file, const char *func, 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 (dprintf_print_all ||
 804             dprintf_find_string(newfile) ||
 805             dprintf_find_string(func)) {
 806                 /* Print out just the function name if requested */
 807                 flockfile(stdout);
 808                 if (dprintf_find_string("pid"))
 809                         (void) printf("%d ", getpid());
 810                 if (dprintf_find_string("tid"))
 811                         (void) printf("%u ", (uint_t) pthread_self());
 812                 if (dprintf_find_string("cpu"))
 813                         (void) printf("%u ", getcpuid());
 814                 if (dprintf_find_string("time"))
 815                         (void) printf("%llu ", gethrtime());
 816                 if (dprintf_find_string("long"))
 817                         (void) printf("%s, line %d: ", newfile, line);
 818                 (void) printf("%s: ", func);
 819                 va_start(adx, fmt);
 820                 (void) vprintf(fmt, adx);
 821                 va_end(adx);
 822                 funlockfile(stdout);
 823         }
 824 }
 825
 826 #endif /* ZFS_DEBUG */
 827
 828 /*
 829  * =========================================================================
 830  * cmn_err() and panic()
 831  * =========================================================================
 832  */
 833 static char ce_prefix[CE_IGNORE][10] = { "", "NOTICE: ", "WARNING: ", "" };
 834 static char ce_suffix[CE_IGNORE][2] = { "", "\n", "\n", "" };
 835
 836 void
 837 vpanic(const char *fmt, va_list adx)
 838 {
 839         (void) fprintf(stderr, "error: ");
 840         (void) vfprintf(stderr, fmt, adx);
 841         (void) fprintf(stderr, "\n");
 842
 843         abort();        /* think of it as a "user-level crash dump" */
 844 }
 845
 846 void
 847 panic(const char *fmt, ...)
 848 {
 849         va_list adx;
 850
 851         va_start(adx, fmt);
 852         vpanic(fmt, adx);
 853         va_end(adx);
 854 }
 855
 856 void
 857 vcmn_err(int ce, const char *fmt, va_list adx)
 858 {
 859         if (ce == CE_PANIC)
 860                 vpanic(fmt, adx);
 861         if (ce != CE_NOTE) {    /* suppress noise in userland stress testing */
 862                 (void) fprintf(stderr, "%s", ce_prefix[ce]);
 863                 (void) vfprintf(stderr, fmt, adx);
 864                 (void) fprintf(stderr, "%s", ce_suffix[ce]);
 865         }
 866 }
 867
 868 /*PRINTFLIKE2*/
 869 void
 870 cmn_err(int ce, const char *fmt, ...)
 871 {
 872         va_list adx;
 873
 874         va_start(adx, fmt);
 875         vcmn_err(ce, fmt, adx);
 876         va_end(adx);
 877 }
 878
 879 /*
 880  * =========================================================================
 881  * kobj interfaces
 882  * =========================================================================
 883  */
 884 struct _buf *
 885 kobj_open_file(char *name)
 886 {
 887         struct _buf *file;
 888         vnode_t *vp;
 889
 890         /* set vp as the _fd field of the file */
 891         if (vn_openat(name, UIO_SYSSPACE, FREAD, 0, &vp, 0, 0, rootdir,
 892             -1) != 0)
 893                 return ((void *)-1UL);
 894
 895         file = umem_zalloc(sizeof (struct _buf), UMEM_NOFAIL);
 896         file->_fd = (intptr_t)vp;
 897         return (file);
 898 }
 899
 900 int
 901 kobj_read_file(struct _buf *file, char *buf, unsigned size, unsigned off)
 902 {
 903         ssize_t resid;
 904
 905         vn_rdwr(UIO_READ, (vnode_t *)file->_fd, buf, size, (offset_t)off,
 906             UIO_SYSSPACE, 0, 0, 0, &resid);
 907
 908         return (size - resid);
 909 }
 910
 911 void
 912 kobj_close_file(struct _buf *file)
 913 {
 914         vn_close((vnode_t *)file->_fd);
 915         umem_free(file, sizeof (struct _buf));
 916 }
 917
 918 int
 919 kobj_get_filesize(struct _buf *file, uint64_t *size)
 920 {
 921         struct stat64 st;
 922         vnode_t *vp = (vnode_t *)file->_fd;
 923
 924         if (fstat64(vp->v_fd, &st) == -1) {
 925                 vn_close(vp);
 926                 return (errno);
 927         }
 928         *size = st.st_size;
 929         return (0);
 930 }
 931
 932 /*
 933  * =========================================================================
 934  * misc routines
 935  * =========================================================================
 936  */
 937
 938 void
 939 delay(clock_t ticks)
 940 {
 941         poll(0, 0, ticks * (1000 / hz));
 942 }
 943
 944 /*
 945  * Find highest one bit set.
 946  *      Returns bit number + 1 of highest bit that is set, otherwise returns 0.
 947  * High order bit is 31 (or 63 in _LP64 kernel).
 948  */
 949 int
 950 highbit(ulong_t i)
 951 {
 952         register int h = 1;
 953
 954         if (i == 0)
 955                 return (0);
 956 #ifdef _LP64
 957         if (i & 0xffffffff00000000ul) {
 958                 h += 32; i >>= 32;
 959         }
 960 #endif
 961         if (i & 0xffff0000) {
 962                 h += 16; i >>= 16;
 963         }
 964         if (i & 0xff00) {
 965                 h += 8; i >>= 8;
 966         }
 967         if (i & 0xf0) {
 968                 h += 4; i >>= 4;
 969         }
 970         if (i & 0xc) {
 971                 h += 2; i >>= 2;
 972         }
 973         if (i & 0x2) {
 974                 h += 1;
 975         }
 976         return (h);
 977 }
 978
 979 static int random_fd = -1, urandom_fd = -1;
 980
 981 static int
 982 random_get_bytes_common(uint8_t *ptr, size_t len, int fd)
 983 {
 984         size_t resid = len;
 985         ssize_t bytes;
 986
 987         ASSERT(fd != -1);
 988
 989         while (resid != 0) {
 990                 bytes = read(fd, ptr, resid);
 991                 ASSERT3S(bytes, >=, 0);
 992                 ptr += bytes;
 993                 resid -= bytes;
 994         }
 995
 996         return (0);
 997 }
 998
 999 int
1000 random_get_bytes(uint8_t *ptr, size_t len)
1001 {
1002         return (random_get_bytes_common(ptr, len, random_fd));
1003 }
1004
1005 int
1006 random_get_pseudo_bytes(uint8_t *ptr, size_t len)
1007 {
1008         return (random_get_bytes_common(ptr, len, urandom_fd));
1009 }
1010
1011 int
1012 ddi_strtoul(const char *hw_serial, char **nptr, int base, unsigned long *result)
1013 {
1014         char *end;
1015
1016         *result = strtoul(hw_serial, &end, base);
1017         if (*result == 0)
1018                 return (errno);
1019         return (0);
1020 }
1021
1022 int
1023 ddi_strtoull(const char *str, char **nptr, int base, u_longlong_t *result)
1024 {
1025         char *end;
1026
1027         *result = strtoull(str, &end, base);
1028         if (*result == 0)
1029                 return (errno);
1030         return (0);
1031 }
1032
1033 /*
1034  * =========================================================================
1035  * kernel emulation setup & teardown
1036  * =========================================================================
1037  */
1038 static int
1039 umem_out_of_memory(void)
1040 {
1041         char errmsg[] = "out of memory -- generating core dump\n";
1042
1043         (void) fprintf(stderr, "%s", errmsg);
1044         abort();
1045         return (0);
1046 }
1047
1048 void
1049 kernel_init(int mode)
1050 {
1051         umem_nofail_callback(umem_out_of_memory);
1052
1053         physmem = sysconf(_SC_PHYS_PAGES);
1054
1055         dprintf("physmem = %llu pages (%.2f GB)\n", physmem,
1056             (double)physmem * sysconf(_SC_PAGE_SIZE) / (1ULL << 30));
1057
1058         (void) snprintf(hw_serial, sizeof (hw_serial), "%ld",
1059             (mode & FWRITE) ? gethostid() : 0);
1060
1061         VERIFY((random_fd = open("/dev/random", O_RDONLY)) != -1);
1062         VERIFY((urandom_fd = open("/dev/urandom", O_RDONLY)) != -1);
1063
1064         thread_init();
1065         system_taskq_init();
1066
1067         spa_init(mode);
1068 }
1069
1070 void
1071 kernel_fini(void)
1072 {
1073         spa_fini();
1074
1075         system_taskq_fini();
1076         thread_fini();
1077
1078         close(random_fd);
1079         close(urandom_fd);
1080
1081         random_fd = -1;
1082         urandom_fd = -1;
1083 }
1084
1085 uid_t
1086 crgetuid(cred_t *cr)
1087 {
1088         return (0);
1089 }
1090
1091 gid_t
1092 crgetgid(cred_t *cr)
1093 {
1094         return (0);
1095 }
1096
1097 int
1098 crgetngroups(cred_t *cr)
1099 {
1100         return (0);
1101 }
1102
1103 gid_t *
1104 crgetgroups(cred_t *cr)
1105 {
1106         return (NULL);
1107 }
1108
1109 int
1110 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
1111 {
1112         return (0);
1113 }
1114
1115 int
1116 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
1117 {
1118         return (0);
1119 }
1120
1121 int
1122 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
1123 {
1124         return (0);
1125 }
1126
1127 ksiddomain_t *
1128 ksid_lookupdomain(const char *dom)
1129 {
1130         ksiddomain_t *kd;
1131
1132         kd = umem_zalloc(sizeof (ksiddomain_t), UMEM_NOFAIL);
1133         kd->kd_name = spa_strdup(dom);
1134         return (kd);
1135 }
1136
1137 void
1138 ksiddomain_rele(ksiddomain_t *ksid)
1139 {
1140         spa_strfree(ksid->kd_name);
1141         umem_free(ksid, sizeof (ksiddomain_t));
1142 }
1143
1144 char *
1145 kmem_vasprintf(const char *fmt, va_list adx)
1146 {
1147         char *buf = NULL;
1148         va_list adx_copy;
1149
1150         va_copy(adx_copy, adx);
1151         VERIFY(vasprintf(&buf, fmt, adx_copy) != -1);
1152         va_end(adx_copy);
1153
1154         return (buf);
1155 }
1156
1157 char *
1158 kmem_asprintf(const char *fmt, ...)
1159 {
1160         char *buf = NULL;
1161         va_list adx;
1162
1163         va_start(adx, fmt);
1164         VERIFY(vasprintf(&buf, fmt, adx) != -1);
1165         va_end(adx);
1166
1167         return (buf);
1168 }
1169
1170 /* ARGSUSED */
1171 int
1172 zfs_onexit_fd_hold(int fd, minor_t *minorp)
1173 {
1174         *minorp = 0;
1175         return (0);
1176 }
1177
1178 /* ARGSUSED */
1179 void
1180 zfs_onexit_fd_rele(int fd)
1181 {
1182 }
1183
1184 /* ARGSUSED */
1185 int
1186 zfs_onexit_add_cb(minor_t minor, void (*func)(void *), void *data,
1187     uint64_t *action_handle)
1188 {
1189         return (0);
1190 }
1191
1192 /* ARGSUSED */
1193 int
1194 zfs_onexit_del_cb(minor_t minor, uint64_t action_handle, boolean_t fire)
1195 {
1196         return (0);
1197 }
1198
1199 /* ARGSUSED */
1200 int
1201 zfs_onexit_cb_data(minor_t minor, uint64_t action_handle, void **data)
1202 {
1203         return (0);
1204 }