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.
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.
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]
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2016 Actifio, Inc. All rights reserved.
34 #include <sys/signal.h>
37 #include <sys/processor.h>
38 #include <sys/zfs_context.h>
39 #include <sys/rrwlock.h>
40 #include <sys/utsname.h>
42 #include <sys/systeminfo.h>
43 #include <zfs_fletcher.h>
44 #include <sys/crypto/icp.h>
47 * Emulation of kernel services in userland.
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;
57 /* If set, all blocks read will be copied to the specified directory. */
58 char *vn_dumpdir = NULL;
60 /* this only exists to have its address taken */
64 * =========================================================================
66 * =========================================================================
69 pthread_cond_t kthread_cond = PTHREAD_COND_INITIALIZER;
70 pthread_mutex_t kthread_lock = PTHREAD_MUTEX_INITIALIZER;
71 pthread_key_t kthread_key;
79 VERIFY3S(pthread_key_create(&kthread_key, NULL), ==, 0);
81 /* Create entry for primary kthread */
82 kt = umem_zalloc(sizeof (kthread_t), UMEM_NOFAIL);
83 kt->t_tid = pthread_self();
86 VERIFY3S(pthread_setspecific(kthread_key, kt), ==, 0);
88 /* Only the main thread should be running at the moment */
89 ASSERT3S(kthread_nr, ==, 0);
96 kthread_t *kt = curthread;
98 ASSERT(pthread_equal(kt->t_tid, pthread_self()));
99 ASSERT3P(kt->t_func, ==, NULL);
101 umem_free(kt, sizeof (kthread_t));
103 /* Wait for all threads to exit via thread_exit() */
104 VERIFY3S(pthread_mutex_lock(&kthread_lock), ==, 0);
106 kthread_nr--; /* Main thread is exiting */
108 while (kthread_nr > 0)
109 VERIFY0(pthread_cond_wait(&kthread_cond, &kthread_lock));
111 ASSERT3S(kthread_nr, ==, 0);
112 VERIFY3S(pthread_mutex_unlock(&kthread_lock), ==, 0);
114 VERIFY3S(pthread_key_delete(kthread_key), ==, 0);
118 zk_thread_current(void)
120 kthread_t *kt = pthread_getspecific(kthread_key);
122 ASSERT3P(kt, !=, NULL);
128 zk_thread_helper(void *arg)
130 kthread_t *kt = (kthread_t *) arg;
132 VERIFY3S(pthread_setspecific(kthread_key, kt), ==, 0);
134 VERIFY3S(pthread_mutex_lock(&kthread_lock), ==, 0);
136 VERIFY3S(pthread_mutex_unlock(&kthread_lock), ==, 0);
137 (void) setpriority(PRIO_PROCESS, 0, kt->t_pri);
139 kt->t_tid = pthread_self();
140 ((thread_func_arg_t) kt->t_func)(kt->t_arg);
142 /* Unreachable, thread must exit with thread_exit() */
149 zk_thread_create(caddr_t stk, size_t stksize, thread_func_t func, void *arg,
150 size_t len, proc_t *pp, int state, pri_t pri, int detachstate)
156 ASSERT0(state & ~TS_RUN);
158 kt = umem_zalloc(sizeof (kthread_t), UMEM_NOFAIL);
163 VERIFY0(pthread_attr_init(&attr));
164 VERIFY0(pthread_attr_setdetachstate(&attr, detachstate));
167 * We allow the default stack size in user space to be specified by
168 * setting the ZFS_STACK_SIZE environment variable. This allows us
169 * the convenience of observing and debugging stack overruns in
170 * user space. Explicitly specified stack sizes will be honored.
171 * The usage of ZFS_STACK_SIZE is discussed further in the
172 * ENVIRONMENT VARIABLES sections of the ztest(1) man page.
175 stkstr = getenv("ZFS_STACK_SIZE");
178 stksize = TS_STACK_MAX;
180 stksize = MAX(atoi(stkstr), TS_STACK_MIN);
183 VERIFY3S(stksize, >, 0);
184 stksize = P2ROUNDUP(MAX(stksize, TS_STACK_MIN), PAGESIZE);
186 * If this ever fails, it may be because the stack size is not a
187 * multiple of system page size.
189 VERIFY0(pthread_attr_setstacksize(&attr, stksize));
190 VERIFY0(pthread_attr_setguardsize(&attr, PAGESIZE));
192 VERIFY0(pthread_create(&kt->t_tid, &attr, &zk_thread_helper, kt));
193 VERIFY0(pthread_attr_destroy(&attr));
201 kthread_t *kt = curthread;
203 ASSERT(pthread_equal(kt->t_tid, pthread_self()));
205 umem_free(kt, sizeof (kthread_t));
207 VERIFY0(pthread_mutex_lock(&kthread_lock));
209 VERIFY0(pthread_mutex_unlock(&kthread_lock));
211 VERIFY0(pthread_cond_broadcast(&kthread_cond));
212 pthread_exit((void *)TS_MAGIC);
216 zk_thread_join(kt_did_t tid)
220 pthread_join((pthread_t)tid, &ret);
221 VERIFY3P(ret, ==, (void *)TS_MAGIC);
225 * =========================================================================
227 * =========================================================================
231 kstat_create(const char *module, int instance, const char *name,
232 const char *class, uchar_t type, ulong_t ndata, uchar_t ks_flag)
239 kstat_install(kstat_t *ksp)
244 kstat_delete(kstat_t *ksp)
249 kstat_waitq_enter(kstat_io_t *kiop)
254 kstat_waitq_exit(kstat_io_t *kiop)
259 kstat_runq_enter(kstat_io_t *kiop)
264 kstat_runq_exit(kstat_io_t *kiop)
269 kstat_waitq_to_runq(kstat_io_t *kiop)
274 kstat_runq_back_to_waitq(kstat_io_t *kiop)
278 kstat_set_raw_ops(kstat_t *ksp,
279 int (*headers)(char *buf, size_t size),
280 int (*data)(char *buf, size_t size, void *data),
281 void *(*addr)(kstat_t *ksp, loff_t index))
285 * =========================================================================
287 * =========================================================================
291 mutex_init(kmutex_t *mp, char *name, int type, void *cookie)
293 ASSERT3S(type, ==, MUTEX_DEFAULT);
294 ASSERT3P(cookie, ==, NULL);
295 mp->m_owner = MTX_INIT;
296 mp->m_magic = MTX_MAGIC;
297 VERIFY3S(pthread_mutex_init(&mp->m_lock, NULL), ==, 0);
301 mutex_destroy(kmutex_t *mp)
303 ASSERT3U(mp->m_magic, ==, MTX_MAGIC);
304 ASSERT3P(mp->m_owner, ==, MTX_INIT);
305 ASSERT0(pthread_mutex_destroy(&(mp)->m_lock));
306 mp->m_owner = MTX_DEST;
311 mutex_enter(kmutex_t *mp)
313 ASSERT3U(mp->m_magic, ==, MTX_MAGIC);
314 ASSERT3P(mp->m_owner, !=, MTX_DEST);
315 ASSERT3P(mp->m_owner, !=, curthread);
316 VERIFY3S(pthread_mutex_lock(&mp->m_lock), ==, 0);
317 ASSERT3P(mp->m_owner, ==, MTX_INIT);
318 mp->m_owner = curthread;
322 mutex_tryenter(kmutex_t *mp)
325 ASSERT3U(mp->m_magic, ==, MTX_MAGIC);
326 ASSERT3P(mp->m_owner, !=, MTX_DEST);
327 if (0 == (err = pthread_mutex_trylock(&mp->m_lock))) {
328 ASSERT3P(mp->m_owner, ==, MTX_INIT);
329 mp->m_owner = curthread;
332 VERIFY3S(err, ==, EBUSY);
338 mutex_exit(kmutex_t *mp)
340 ASSERT3U(mp->m_magic, ==, MTX_MAGIC);
341 ASSERT3P(mutex_owner(mp), ==, curthread);
342 mp->m_owner = MTX_INIT;
343 VERIFY3S(pthread_mutex_unlock(&mp->m_lock), ==, 0);
347 mutex_owner(kmutex_t *mp)
349 ASSERT3U(mp->m_magic, ==, MTX_MAGIC);
350 return (mp->m_owner);
354 mutex_held(kmutex_t *mp)
356 return (mp->m_owner == curthread);
360 * =========================================================================
362 * =========================================================================
366 rw_init(krwlock_t *rwlp, char *name, int type, void *arg)
368 ASSERT3S(type, ==, RW_DEFAULT);
369 ASSERT3P(arg, ==, NULL);
370 VERIFY3S(pthread_rwlock_init(&rwlp->rw_lock, NULL), ==, 0);
371 rwlp->rw_owner = RW_INIT;
372 rwlp->rw_wr_owner = RW_INIT;
373 rwlp->rw_readers = 0;
374 rwlp->rw_magic = RW_MAGIC;
378 rw_destroy(krwlock_t *rwlp)
380 ASSERT3U(rwlp->rw_magic, ==, RW_MAGIC);
381 ASSERT(rwlp->rw_readers == 0 && rwlp->rw_wr_owner == RW_INIT);
382 VERIFY3S(pthread_rwlock_destroy(&rwlp->rw_lock), ==, 0);
387 rw_enter(krwlock_t *rwlp, krw_t rw)
389 ASSERT3U(rwlp->rw_magic, ==, RW_MAGIC);
390 ASSERT3P(rwlp->rw_owner, !=, curthread);
391 ASSERT3P(rwlp->rw_wr_owner, !=, curthread);
393 if (rw == RW_READER) {
394 VERIFY3S(pthread_rwlock_rdlock(&rwlp->rw_lock), ==, 0);
395 ASSERT3P(rwlp->rw_wr_owner, ==, RW_INIT);
397 atomic_inc_uint(&rwlp->rw_readers);
399 VERIFY3S(pthread_rwlock_wrlock(&rwlp->rw_lock), ==, 0);
400 ASSERT3P(rwlp->rw_wr_owner, ==, RW_INIT);
401 ASSERT3U(rwlp->rw_readers, ==, 0);
403 rwlp->rw_wr_owner = curthread;
406 rwlp->rw_owner = curthread;
410 rw_exit(krwlock_t *rwlp)
412 ASSERT3U(rwlp->rw_magic, ==, RW_MAGIC);
413 ASSERT(RW_LOCK_HELD(rwlp));
415 if (RW_READ_HELD(rwlp))
416 atomic_dec_uint(&rwlp->rw_readers);
418 rwlp->rw_wr_owner = RW_INIT;
420 rwlp->rw_owner = RW_INIT;
421 VERIFY3S(pthread_rwlock_unlock(&rwlp->rw_lock), ==, 0);
425 rw_tryenter(krwlock_t *rwlp, krw_t rw)
429 ASSERT3U(rwlp->rw_magic, ==, RW_MAGIC);
432 rv = pthread_rwlock_tryrdlock(&rwlp->rw_lock);
434 rv = pthread_rwlock_trywrlock(&rwlp->rw_lock);
437 ASSERT3P(rwlp->rw_wr_owner, ==, RW_INIT);
440 atomic_inc_uint(&rwlp->rw_readers);
442 ASSERT3U(rwlp->rw_readers, ==, 0);
443 rwlp->rw_wr_owner = curthread;
446 rwlp->rw_owner = curthread;
450 VERIFY3S(rv, ==, EBUSY);
456 rw_tryupgrade(krwlock_t *rwlp)
458 ASSERT3U(rwlp->rw_magic, ==, RW_MAGIC);
464 * =========================================================================
465 * condition variables
466 * =========================================================================
470 cv_init(kcondvar_t *cv, char *name, int type, void *arg)
472 ASSERT3S(type, ==, CV_DEFAULT);
473 cv->cv_magic = CV_MAGIC;
474 VERIFY0(pthread_cond_init(&cv->cv, NULL));
478 cv_destroy(kcondvar_t *cv)
480 ASSERT3U(cv->cv_magic, ==, CV_MAGIC);
481 VERIFY0(pthread_cond_destroy(&cv->cv));
486 cv_wait(kcondvar_t *cv, kmutex_t *mp)
488 ASSERT3U(cv->cv_magic, ==, CV_MAGIC);
489 ASSERT3P(mutex_owner(mp), ==, curthread);
490 mp->m_owner = MTX_INIT;
491 VERIFY0(pthread_cond_wait(&cv->cv, &mp->m_lock));
492 mp->m_owner = curthread;
496 cv_timedwait(kcondvar_t *cv, kmutex_t *mp, clock_t abstime)
503 ASSERT3U(cv->cv_magic, ==, CV_MAGIC);
505 delta = abstime - ddi_get_lbolt();
509 VERIFY(gettimeofday(&tv, NULL) == 0);
511 ts.tv_sec = tv.tv_sec + delta / hz;
512 ts.tv_nsec = tv.tv_usec * 1000 + (delta % hz) * (NANOSEC / hz);
513 if (ts.tv_nsec >= NANOSEC) {
515 ts.tv_nsec -= NANOSEC;
518 ASSERT3P(mutex_owner(mp), ==, curthread);
519 mp->m_owner = MTX_INIT;
520 error = pthread_cond_timedwait(&cv->cv, &mp->m_lock, &ts);
521 mp->m_owner = curthread;
523 if (error == ETIMEDOUT)
533 cv_timedwait_hires(kcondvar_t *cv, kmutex_t *mp, hrtime_t tim, hrtime_t res,
540 ASSERT(flag == 0 || flag == CALLOUT_FLAG_ABSOLUTE);
543 if (flag & CALLOUT_FLAG_ABSOLUTE)
544 delta -= gethrtime();
549 ts.tv_sec = delta / NANOSEC;
550 ts.tv_nsec = delta % NANOSEC;
552 ASSERT(mutex_owner(mp) == curthread);
554 error = pthread_cond_timedwait(&cv->cv, &mp->m_lock, &ts);
555 mp->m_owner = curthread;
557 if (error == ETIMEDOUT)
566 cv_signal(kcondvar_t *cv)
568 ASSERT3U(cv->cv_magic, ==, CV_MAGIC);
569 VERIFY0(pthread_cond_signal(&cv->cv));
573 cv_broadcast(kcondvar_t *cv)
575 ASSERT3U(cv->cv_magic, ==, CV_MAGIC);
576 VERIFY0(pthread_cond_broadcast(&cv->cv));
580 * =========================================================================
582 * =========================================================================
585 * Note: for the xxxat() versions of these functions, we assume that the
586 * starting vp is always rootdir (which is true for spa_directory.c, the only
587 * ZFS consumer of these interfaces). We assert this is true, and then emulate
588 * them by adding '/' in front of the path.
593 vn_open(char *path, int x1, int flags, int mode, vnode_t **vpp, int x2, int x3)
603 realpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL);
606 * If we're accessing a real disk from userland, we need to use
607 * the character interface to avoid caching. This is particularly
608 * important if we're trying to look at a real in-kernel storage
609 * pool from userland, e.g. via zdb, because otherwise we won't
610 * see the changes occurring under the segmap cache.
611 * On the other hand, the stupid character device returns zero
612 * for its size. So -- gag -- we open the block device to get
613 * its size, and remember it for subsequent VOP_GETATTR().
615 #if defined(__sun__) || defined(__sun)
616 if (strncmp(path, "/dev/", 5) == 0) {
621 fd = open64(path, O_RDONLY);
627 if (fstat64(fd, &st) == -1) {
634 (void) sprintf(realpath, "%s", path);
635 dsk = strstr(path, "/dsk/");
637 (void) sprintf(realpath + (dsk - path) + 1, "r%s",
640 (void) sprintf(realpath, "%s", path);
641 if (!(flags & FCREAT) && stat64(realpath, &st) == -1) {
648 if (!(flags & FCREAT) && S_ISBLK(st.st_mode)) {
652 /* We shouldn't be writing to block devices in userspace */
653 VERIFY(!(flags & FWRITE));
657 old_umask = umask(0);
660 * The construct 'flags - FREAD' conveniently maps combinations of
661 * FREAD and FWRITE to the corresponding O_RDONLY, O_WRONLY, and O_RDWR.
663 fd = open64(realpath, flags - FREAD, mode);
667 (void) umask(old_umask);
669 if (vn_dumpdir != NULL) {
670 char *dumppath = umem_zalloc(MAXPATHLEN, UMEM_NOFAIL);
671 (void) snprintf(dumppath, MAXPATHLEN,
672 "%s/%s", vn_dumpdir, basename(realpath));
673 dump_fd = open64(dumppath, O_CREAT | O_WRONLY, 0666);
674 umem_free(dumppath, MAXPATHLEN);
690 if (fstat64_blk(fd, &st) == -1) {
696 (void) fcntl(fd, F_SETFD, FD_CLOEXEC);
698 *vpp = vp = umem_zalloc(sizeof (vnode_t), UMEM_NOFAIL);
701 vp->v_size = st.st_size;
702 vp->v_path = spa_strdup(path);
703 vp->v_dump_fd = dump_fd;
710 vn_openat(char *path, int x1, int flags, int mode, vnode_t **vpp, int x2,
711 int x3, vnode_t *startvp, int fd)
713 char *realpath = umem_alloc(strlen(path) + 2, UMEM_NOFAIL);
716 ASSERT(startvp == rootdir);
717 (void) sprintf(realpath, "/%s", path);
719 /* fd ignored for now, need if want to simulate nbmand support */
720 ret = vn_open(realpath, x1, flags, mode, vpp, x2, x3);
722 umem_free(realpath, strlen(path) + 2);
729 vn_rdwr(int uio, vnode_t *vp, void *addr, ssize_t len, offset_t offset,
730 int x1, int x2, rlim64_t x3, void *x4, ssize_t *residp)
732 ssize_t rc, done = 0, split;
734 if (uio == UIO_READ) {
735 rc = pread64(vp->v_fd, addr, len, offset);
736 if (vp->v_dump_fd != -1) {
738 status = pwrite64(vp->v_dump_fd, addr, rc, offset);
739 ASSERT(status != -1);
743 * To simulate partial disk writes, we split writes into two
744 * system calls so that the process can be killed in between.
746 int sectors = len >> SPA_MINBLOCKSHIFT;
747 split = (sectors > 0 ? rand() % sectors : 0) <<
749 rc = pwrite64(vp->v_fd, addr, split, offset);
752 rc = pwrite64(vp->v_fd, (char *)addr + split,
753 len - split, offset + split);
758 if (rc == -1 && errno == EINVAL) {
760 * Under Linux, this most likely means an alignment issue
761 * (memory or disk) due to O_DIRECT, so we abort() in order to
762 * catch the offender.
773 *residp = len - done;
774 else if (done != len)
780 vn_close(vnode_t *vp)
783 if (vp->v_dump_fd != -1)
784 close(vp->v_dump_fd);
785 spa_strfree(vp->v_path);
786 umem_free(vp, sizeof (vnode_t));
790 * At a minimum we need to update the size since vdev_reopen()
791 * will no longer call vn_openat().
794 fop_getattr(vnode_t *vp, vattr_t *vap)
799 if (fstat64_blk(vp->v_fd, &st) == -1) {
805 vap->va_size = st.st_size;
810 * =========================================================================
811 * Figure out which debugging statements to print
812 * =========================================================================
815 static char *dprintf_string;
816 static int dprintf_print_all;
819 dprintf_find_string(const char *string)
821 char *tmp_str = dprintf_string;
822 int len = strlen(string);
825 * Find out if this is a string we want to print.
826 * String format: file1.c,function_name1,file2.c,file3.c
829 while (tmp_str != NULL) {
830 if (strncmp(tmp_str, string, len) == 0 &&
831 (tmp_str[len] == ',' || tmp_str[len] == '\0'))
833 tmp_str = strchr(tmp_str, ',');
835 tmp_str++; /* Get rid of , */
841 dprintf_setup(int *argc, char **argv)
846 * Debugging can be specified two ways: by setting the
847 * environment variable ZFS_DEBUG, or by including a
848 * "debug=..." argument on the command line. The command
849 * line setting overrides the environment variable.
852 for (i = 1; i < *argc; i++) {
853 int len = strlen("debug=");
854 /* First look for a command line argument */
855 if (strncmp("debug=", argv[i], len) == 0) {
856 dprintf_string = argv[i] + len;
857 /* Remove from args */
858 for (j = i; j < *argc; j++)
865 if (dprintf_string == NULL) {
866 /* Look for ZFS_DEBUG environment variable */
867 dprintf_string = getenv("ZFS_DEBUG");
871 * Are we just turning on all debugging?
873 if (dprintf_find_string("on"))
874 dprintf_print_all = 1;
876 if (dprintf_string != NULL)
877 zfs_flags |= ZFS_DEBUG_DPRINTF;
881 * =========================================================================
883 * =========================================================================
886 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
892 * Get rid of annoying "../common/" prefix to filename.
894 newfile = strrchr(file, '/');
895 if (newfile != NULL) {
896 newfile = newfile + 1; /* Get rid of leading / */
901 if (dprintf_print_all ||
902 dprintf_find_string(newfile) ||
903 dprintf_find_string(func)) {
904 /* Print out just the function name if requested */
906 if (dprintf_find_string("pid"))
907 (void) printf("%d ", getpid());
908 if (dprintf_find_string("tid"))
909 (void) printf("%u ", (uint_t) pthread_self());
910 if (dprintf_find_string("cpu"))
911 (void) printf("%u ", getcpuid());
912 if (dprintf_find_string("time"))
913 (void) printf("%llu ", gethrtime());
914 if (dprintf_find_string("long"))
915 (void) printf("%s, line %d: ", newfile, line);
916 (void) printf("%s: ", func);
918 (void) vprintf(fmt, adx);
925 * =========================================================================
926 * cmn_err() and panic()
927 * =========================================================================
929 static char ce_prefix[CE_IGNORE][10] = { "", "NOTICE: ", "WARNING: ", "" };
930 static char ce_suffix[CE_IGNORE][2] = { "", "\n", "\n", "" };
933 vpanic(const char *fmt, va_list adx)
935 (void) fprintf(stderr, "error: ");
936 (void) vfprintf(stderr, fmt, adx);
937 (void) fprintf(stderr, "\n");
939 abort(); /* think of it as a "user-level crash dump" */
943 panic(const char *fmt, ...)
953 vcmn_err(int ce, const char *fmt, va_list adx)
957 if (ce != CE_NOTE) { /* suppress noise in userland stress testing */
958 (void) fprintf(stderr, "%s", ce_prefix[ce]);
959 (void) vfprintf(stderr, fmt, adx);
960 (void) fprintf(stderr, "%s", ce_suffix[ce]);
966 cmn_err(int ce, const char *fmt, ...)
971 vcmn_err(ce, fmt, adx);
976 * =========================================================================
978 * =========================================================================
981 kobj_open_file(char *name)
986 /* set vp as the _fd field of the file */
987 if (vn_openat(name, UIO_SYSSPACE, FREAD, 0, &vp, 0, 0, rootdir,
989 return ((void *)-1UL);
991 file = umem_zalloc(sizeof (struct _buf), UMEM_NOFAIL);
992 file->_fd = (intptr_t)vp;
997 kobj_read_file(struct _buf *file, char *buf, unsigned size, unsigned off)
1001 if (vn_rdwr(UIO_READ, (vnode_t *)file->_fd, buf, size, (offset_t)off,
1002 UIO_SYSSPACE, 0, 0, 0, &resid) != 0)
1005 return (size - resid);
1009 kobj_close_file(struct _buf *file)
1011 vn_close((vnode_t *)file->_fd);
1012 umem_free(file, sizeof (struct _buf));
1016 kobj_get_filesize(struct _buf *file, uint64_t *size)
1019 vnode_t *vp = (vnode_t *)file->_fd;
1021 if (fstat64(vp->v_fd, &st) == -1) {
1030 * =========================================================================
1032 * =========================================================================
1036 delay(clock_t ticks)
1038 poll(0, 0, ticks * (1000 / hz));
1042 * Find highest one bit set.
1043 * Returns bit number + 1 of highest bit that is set, otherwise returns 0.
1044 * High order bit is 31 (or 63 in _LP64 kernel).
1047 highbit64(uint64_t i)
1053 if (i & 0xffffffff00000000ULL) {
1056 if (i & 0xffff0000) {
1075 * Find lowest one bit set.
1076 * Returns bit number + 1 of lowest bit that is set, otherwise returns 0.
1077 * This is basically a reimplementation of ffsll(), which is GNU specific.
1080 lowbit64(uint64_t i)
1082 register int h = 64;
1086 if (i & 0x00000000ffffffffULL)
1118 * Find highest one bit set.
1119 * Returns bit number + 1 of highest bit that is set, otherwise returns 0.
1120 * High order bit is 31 (or 63 in _LP64 kernel).
1130 if (i & 0xffffffff00000000ul) {
1134 if (i & 0xffff0000) {
1153 * Find lowest one bit set.
1154 * Returns bit number + 1 of lowest bit that is set, otherwise returns 0.
1155 * Low order bit is 0.
1166 if (!(i & 0xffffffff)) {
1170 if (!(i & 0xffff)) {
1188 static int random_fd = -1, urandom_fd = -1;
1193 VERIFY((random_fd = open("/dev/random", O_RDONLY)) != -1);
1194 VERIFY((urandom_fd = open("/dev/urandom", O_RDONLY)) != -1);
1208 random_get_bytes_common(uint8_t *ptr, size_t len, int fd)
1215 while (resid != 0) {
1216 bytes = read(fd, ptr, resid);
1217 ASSERT3S(bytes, >=, 0);
1226 random_get_bytes(uint8_t *ptr, size_t len)
1228 return (random_get_bytes_common(ptr, len, random_fd));
1232 random_get_pseudo_bytes(uint8_t *ptr, size_t len)
1234 return (random_get_bytes_common(ptr, len, urandom_fd));
1238 ddi_strtoul(const char *hw_serial, char **nptr, int base, unsigned long *result)
1242 *result = strtoul(hw_serial, &end, base);
1249 ddi_strtoull(const char *str, char **nptr, int base, u_longlong_t *result)
1253 *result = strtoull(str, &end, base);
1262 return (&hw_utsname);
1266 * =========================================================================
1267 * kernel emulation setup & teardown
1268 * =========================================================================
1271 umem_out_of_memory(void)
1273 char errmsg[] = "out of memory -- generating core dump\n";
1275 (void) fprintf(stderr, "%s", errmsg);
1280 static unsigned long
1281 get_spl_hostid(void)
1284 unsigned long hostid;
1286 f = fopen("/sys/module/spl/parameters/spl_hostid", "r");
1289 if (fscanf(f, "%lu", &hostid) != 1)
1292 return (hostid & 0xffffffff);
1296 get_system_hostid(void)
1298 unsigned long system_hostid = get_spl_hostid();
1299 if (system_hostid == 0)
1300 system_hostid = gethostid() & 0xffffffff;
1301 return (system_hostid);
1305 kernel_init(int mode)
1307 extern uint_t rrw_tsd_key;
1309 umem_nofail_callback(umem_out_of_memory);
1311 physmem = sysconf(_SC_PHYS_PAGES);
1313 dprintf("physmem = %llu pages (%.2f GB)\n", physmem,
1314 (double)physmem * sysconf(_SC_PAGE_SIZE) / (1ULL << 30));
1316 (void) snprintf(hw_serial, sizeof (hw_serial), "%ld",
1317 (mode & FWRITE) ? get_system_hostid() : 0);
1321 VERIFY0(uname(&hw_utsname));
1324 system_taskq_init();
1331 tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
1341 system_taskq_fini();
1348 crgetuid(cred_t *cr)
1354 crgetruid(cred_t *cr)
1360 crgetgid(cred_t *cr)
1366 crgetngroups(cred_t *cr)
1372 crgetgroups(cred_t *cr)
1378 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
1384 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
1390 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
1396 secpolicy_zfs(const cred_t *cr)
1402 ksid_lookupdomain(const char *dom)
1406 kd = umem_zalloc(sizeof (ksiddomain_t), UMEM_NOFAIL);
1407 kd->kd_name = spa_strdup(dom);
1412 ksiddomain_rele(ksiddomain_t *ksid)
1414 spa_strfree(ksid->kd_name);
1415 umem_free(ksid, sizeof (ksiddomain_t));
1419 kmem_vasprintf(const char *fmt, va_list adx)
1424 va_copy(adx_copy, adx);
1425 VERIFY(vasprintf(&buf, fmt, adx_copy) != -1);
1432 kmem_asprintf(const char *fmt, ...)
1438 VERIFY(vasprintf(&buf, fmt, adx) != -1);
1446 zfs_onexit_fd_hold(int fd, minor_t *minorp)
1454 zfs_onexit_fd_rele(int fd)
1460 zfs_onexit_add_cb(minor_t minor, void (*func)(void *), void *data,
1461 uint64_t *action_handle)
1468 zfs_onexit_del_cb(minor_t minor, uint64_t action_handle, boolean_t fire)
1475 zfs_onexit_cb_data(minor_t minor, uint64_t action_handle, void **data)
1481 spl_fstrans_mark(void)
1483 return ((fstrans_cookie_t) 0);
1487 spl_fstrans_unmark(fstrans_cookie_t cookie)
1492 spl_fstrans_check(void)
1497 void *zvol_tag = "zvol_tag";
1500 zvol_create_minors(spa_t *spa, const char *name, boolean_t async)
1505 zvol_remove_minor(spa_t *spa, const char *name, boolean_t async)
1510 zvol_remove_minors(spa_t *spa, const char *name, boolean_t async)
1515 zvol_rename_minors(spa_t *spa, const char *oldname, const char *newname,