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>
46 * Emulation of kernel services in userland.
51 vnode_t
*rootdir
= (vnode_t
*)0xabcd1234;
52 char hw_serial
[HW_HOSTID_LEN
];
53 struct utsname hw_utsname
;
54 vmem_t
*zio_arena
= NULL
;
56 /* If set, all blocks read will be copied to the specified directory. */
57 char *vn_dumpdir
= NULL
;
59 /* this only exists to have its address taken */
63 * =========================================================================
65 * =========================================================================
68 pthread_cond_t kthread_cond
= PTHREAD_COND_INITIALIZER
;
69 pthread_mutex_t kthread_lock
= PTHREAD_MUTEX_INITIALIZER
;
70 pthread_key_t kthread_key
;
78 VERIFY3S(pthread_key_create(&kthread_key
, NULL
), ==, 0);
80 /* Create entry for primary kthread */
81 kt
= umem_zalloc(sizeof (kthread_t
), UMEM_NOFAIL
);
82 kt
->t_tid
= pthread_self();
85 VERIFY3S(pthread_setspecific(kthread_key
, kt
), ==, 0);
87 /* Only the main thread should be running at the moment */
88 ASSERT3S(kthread_nr
, ==, 0);
95 kthread_t
*kt
= curthread
;
97 ASSERT(pthread_equal(kt
->t_tid
, pthread_self()));
98 ASSERT3P(kt
->t_func
, ==, NULL
);
100 umem_free(kt
, sizeof (kthread_t
));
102 /* Wait for all threads to exit via thread_exit() */
103 VERIFY3S(pthread_mutex_lock(&kthread_lock
), ==, 0);
105 kthread_nr
--; /* Main thread is exiting */
107 while (kthread_nr
> 0)
108 VERIFY0(pthread_cond_wait(&kthread_cond
, &kthread_lock
));
110 ASSERT3S(kthread_nr
, ==, 0);
111 VERIFY3S(pthread_mutex_unlock(&kthread_lock
), ==, 0);
113 VERIFY3S(pthread_key_delete(kthread_key
), ==, 0);
117 zk_thread_current(void)
119 kthread_t
*kt
= pthread_getspecific(kthread_key
);
121 ASSERT3P(kt
, !=, NULL
);
127 zk_thread_helper(void *arg
)
129 kthread_t
*kt
= (kthread_t
*) arg
;
131 VERIFY3S(pthread_setspecific(kthread_key
, kt
), ==, 0);
133 VERIFY3S(pthread_mutex_lock(&kthread_lock
), ==, 0);
135 VERIFY3S(pthread_mutex_unlock(&kthread_lock
), ==, 0);
136 (void) setpriority(PRIO_PROCESS
, 0, kt
->t_pri
);
138 kt
->t_tid
= pthread_self();
139 ((thread_func_arg_t
) kt
->t_func
)(kt
->t_arg
);
141 /* Unreachable, thread must exit with thread_exit() */
148 zk_thread_create(caddr_t stk
, size_t stksize
, thread_func_t func
, void *arg
,
149 size_t len
, proc_t
*pp
, int state
, pri_t pri
, int detachstate
)
155 ASSERT0(state
& ~TS_RUN
);
157 kt
= umem_zalloc(sizeof (kthread_t
), UMEM_NOFAIL
);
162 VERIFY0(pthread_attr_init(&attr
));
163 VERIFY0(pthread_attr_setdetachstate(&attr
, detachstate
));
166 * We allow the default stack size in user space to be specified by
167 * setting the ZFS_STACK_SIZE environment variable. This allows us
168 * the convenience of observing and debugging stack overruns in
169 * user space. Explicitly specified stack sizes will be honored.
170 * The usage of ZFS_STACK_SIZE is discussed further in the
171 * ENVIRONMENT VARIABLES sections of the ztest(1) man page.
174 stkstr
= getenv("ZFS_STACK_SIZE");
177 stksize
= TS_STACK_MAX
;
179 stksize
= MAX(atoi(stkstr
), TS_STACK_MIN
);
182 VERIFY3S(stksize
, >, 0);
183 stksize
= P2ROUNDUP(MAX(stksize
, TS_STACK_MIN
), PAGESIZE
);
185 * If this ever fails, it may be because the stack size is not a
186 * multiple of system page size.
188 VERIFY0(pthread_attr_setstacksize(&attr
, stksize
));
189 VERIFY0(pthread_attr_setguardsize(&attr
, PAGESIZE
));
191 VERIFY0(pthread_create(&kt
->t_tid
, &attr
, &zk_thread_helper
, kt
));
192 VERIFY0(pthread_attr_destroy(&attr
));
200 kthread_t
*kt
= curthread
;
202 ASSERT(pthread_equal(kt
->t_tid
, pthread_self()));
204 umem_free(kt
, sizeof (kthread_t
));
206 VERIFY0(pthread_mutex_lock(&kthread_lock
));
208 VERIFY0(pthread_mutex_unlock(&kthread_lock
));
210 VERIFY0(pthread_cond_broadcast(&kthread_cond
));
211 pthread_exit((void *)TS_MAGIC
);
215 zk_thread_join(kt_did_t tid
)
219 pthread_join((pthread_t
)tid
, &ret
);
220 VERIFY3P(ret
, ==, (void *)TS_MAGIC
);
224 * =========================================================================
226 * =========================================================================
230 kstat_create(const char *module
, int instance
, const char *name
,
231 const char *class, uchar_t type
, ulong_t ndata
, uchar_t ks_flag
)
238 kstat_install(kstat_t
*ksp
)
243 kstat_delete(kstat_t
*ksp
)
248 kstat_waitq_enter(kstat_io_t
*kiop
)
253 kstat_waitq_exit(kstat_io_t
*kiop
)
258 kstat_runq_enter(kstat_io_t
*kiop
)
263 kstat_runq_exit(kstat_io_t
*kiop
)
268 kstat_waitq_to_runq(kstat_io_t
*kiop
)
273 kstat_runq_back_to_waitq(kstat_io_t
*kiop
)
277 kstat_set_raw_ops(kstat_t
*ksp
,
278 int (*headers
)(char *buf
, size_t size
),
279 int (*data
)(char *buf
, size_t size
, void *data
),
280 void *(*addr
)(kstat_t
*ksp
, loff_t index
))
284 * =========================================================================
286 * =========================================================================
290 mutex_init(kmutex_t
*mp
, char *name
, int type
, void *cookie
)
292 ASSERT3S(type
, ==, MUTEX_DEFAULT
);
293 ASSERT3P(cookie
, ==, NULL
);
294 mp
->m_owner
= MTX_INIT
;
295 mp
->m_magic
= MTX_MAGIC
;
296 VERIFY3S(pthread_mutex_init(&mp
->m_lock
, NULL
), ==, 0);
300 mutex_destroy(kmutex_t
*mp
)
302 ASSERT3U(mp
->m_magic
, ==, MTX_MAGIC
);
303 ASSERT3P(mp
->m_owner
, ==, MTX_INIT
);
304 ASSERT0(pthread_mutex_destroy(&(mp
)->m_lock
));
305 mp
->m_owner
= MTX_DEST
;
310 mutex_enter(kmutex_t
*mp
)
312 ASSERT3U(mp
->m_magic
, ==, MTX_MAGIC
);
313 ASSERT3P(mp
->m_owner
, !=, MTX_DEST
);
314 ASSERT3P(mp
->m_owner
, !=, curthread
);
315 VERIFY3S(pthread_mutex_lock(&mp
->m_lock
), ==, 0);
316 ASSERT3P(mp
->m_owner
, ==, MTX_INIT
);
317 mp
->m_owner
= curthread
;
321 mutex_tryenter(kmutex_t
*mp
)
324 ASSERT3U(mp
->m_magic
, ==, MTX_MAGIC
);
325 ASSERT3P(mp
->m_owner
, !=, MTX_DEST
);
326 if (0 == (err
= pthread_mutex_trylock(&mp
->m_lock
))) {
327 ASSERT3P(mp
->m_owner
, ==, MTX_INIT
);
328 mp
->m_owner
= curthread
;
331 VERIFY3S(err
, ==, EBUSY
);
337 mutex_exit(kmutex_t
*mp
)
339 ASSERT3U(mp
->m_magic
, ==, MTX_MAGIC
);
340 ASSERT3P(mutex_owner(mp
), ==, curthread
);
341 mp
->m_owner
= MTX_INIT
;
342 VERIFY3S(pthread_mutex_unlock(&mp
->m_lock
), ==, 0);
346 mutex_owner(kmutex_t
*mp
)
348 ASSERT3U(mp
->m_magic
, ==, MTX_MAGIC
);
349 return (mp
->m_owner
);
353 mutex_held(kmutex_t
*mp
)
355 return (mp
->m_owner
== curthread
);
359 * =========================================================================
361 * =========================================================================
365 rw_init(krwlock_t
*rwlp
, char *name
, int type
, void *arg
)
367 ASSERT3S(type
, ==, RW_DEFAULT
);
368 ASSERT3P(arg
, ==, NULL
);
369 VERIFY3S(pthread_rwlock_init(&rwlp
->rw_lock
, NULL
), ==, 0);
370 rwlp
->rw_owner
= RW_INIT
;
371 rwlp
->rw_wr_owner
= RW_INIT
;
372 rwlp
->rw_readers
= 0;
373 rwlp
->rw_magic
= RW_MAGIC
;
377 rw_destroy(krwlock_t
*rwlp
)
379 ASSERT3U(rwlp
->rw_magic
, ==, RW_MAGIC
);
380 ASSERT(rwlp
->rw_readers
== 0 && rwlp
->rw_wr_owner
== RW_INIT
);
381 VERIFY3S(pthread_rwlock_destroy(&rwlp
->rw_lock
), ==, 0);
386 rw_enter(krwlock_t
*rwlp
, krw_t rw
)
388 ASSERT3U(rwlp
->rw_magic
, ==, RW_MAGIC
);
389 ASSERT3P(rwlp
->rw_owner
, !=, curthread
);
390 ASSERT3P(rwlp
->rw_wr_owner
, !=, curthread
);
392 if (rw
== RW_READER
) {
393 VERIFY3S(pthread_rwlock_rdlock(&rwlp
->rw_lock
), ==, 0);
394 ASSERT3P(rwlp
->rw_wr_owner
, ==, RW_INIT
);
396 atomic_inc_uint(&rwlp
->rw_readers
);
398 VERIFY3S(pthread_rwlock_wrlock(&rwlp
->rw_lock
), ==, 0);
399 ASSERT3P(rwlp
->rw_wr_owner
, ==, RW_INIT
);
400 ASSERT3U(rwlp
->rw_readers
, ==, 0);
402 rwlp
->rw_wr_owner
= curthread
;
405 rwlp
->rw_owner
= curthread
;
409 rw_exit(krwlock_t
*rwlp
)
411 ASSERT3U(rwlp
->rw_magic
, ==, RW_MAGIC
);
412 ASSERT(RW_LOCK_HELD(rwlp
));
414 if (RW_READ_HELD(rwlp
))
415 atomic_dec_uint(&rwlp
->rw_readers
);
417 rwlp
->rw_wr_owner
= RW_INIT
;
419 rwlp
->rw_owner
= RW_INIT
;
420 VERIFY3S(pthread_rwlock_unlock(&rwlp
->rw_lock
), ==, 0);
424 rw_tryenter(krwlock_t
*rwlp
, krw_t rw
)
428 ASSERT3U(rwlp
->rw_magic
, ==, RW_MAGIC
);
431 rv
= pthread_rwlock_tryrdlock(&rwlp
->rw_lock
);
433 rv
= pthread_rwlock_trywrlock(&rwlp
->rw_lock
);
436 ASSERT3P(rwlp
->rw_wr_owner
, ==, RW_INIT
);
439 atomic_inc_uint(&rwlp
->rw_readers
);
441 ASSERT3U(rwlp
->rw_readers
, ==, 0);
442 rwlp
->rw_wr_owner
= curthread
;
445 rwlp
->rw_owner
= curthread
;
449 VERIFY3S(rv
, ==, EBUSY
);
455 rw_tryupgrade(krwlock_t
*rwlp
)
457 ASSERT3U(rwlp
->rw_magic
, ==, RW_MAGIC
);
463 * =========================================================================
464 * condition variables
465 * =========================================================================
469 cv_init(kcondvar_t
*cv
, char *name
, int type
, void *arg
)
471 ASSERT3S(type
, ==, CV_DEFAULT
);
472 cv
->cv_magic
= CV_MAGIC
;
473 VERIFY0(pthread_cond_init(&cv
->cv
, NULL
));
477 cv_destroy(kcondvar_t
*cv
)
479 ASSERT3U(cv
->cv_magic
, ==, CV_MAGIC
);
480 VERIFY0(pthread_cond_destroy(&cv
->cv
));
485 cv_wait(kcondvar_t
*cv
, kmutex_t
*mp
)
487 ASSERT3U(cv
->cv_magic
, ==, CV_MAGIC
);
488 ASSERT3P(mutex_owner(mp
), ==, curthread
);
489 mp
->m_owner
= MTX_INIT
;
490 VERIFY0(pthread_cond_wait(&cv
->cv
, &mp
->m_lock
));
491 mp
->m_owner
= curthread
;
495 cv_timedwait(kcondvar_t
*cv
, kmutex_t
*mp
, clock_t abstime
)
502 ASSERT3U(cv
->cv_magic
, ==, CV_MAGIC
);
504 delta
= abstime
- ddi_get_lbolt();
508 VERIFY(gettimeofday(&tv
, NULL
) == 0);
510 ts
.tv_sec
= tv
.tv_sec
+ delta
/ hz
;
511 ts
.tv_nsec
= tv
.tv_usec
* 1000 + (delta
% hz
) * (NANOSEC
/ hz
);
512 if (ts
.tv_nsec
>= NANOSEC
) {
514 ts
.tv_nsec
-= NANOSEC
;
517 ASSERT3P(mutex_owner(mp
), ==, curthread
);
518 mp
->m_owner
= MTX_INIT
;
519 error
= pthread_cond_timedwait(&cv
->cv
, &mp
->m_lock
, &ts
);
520 mp
->m_owner
= curthread
;
522 if (error
== ETIMEDOUT
)
532 cv_timedwait_hires(kcondvar_t
*cv
, kmutex_t
*mp
, hrtime_t tim
, hrtime_t res
,
539 ASSERT(flag
== 0 || flag
== CALLOUT_FLAG_ABSOLUTE
);
542 if (flag
& CALLOUT_FLAG_ABSOLUTE
)
543 delta
-= gethrtime();
548 ts
.tv_sec
= delta
/ NANOSEC
;
549 ts
.tv_nsec
= delta
% NANOSEC
;
551 ASSERT(mutex_owner(mp
) == curthread
);
553 error
= pthread_cond_timedwait(&cv
->cv
, &mp
->m_lock
, &ts
);
554 mp
->m_owner
= curthread
;
556 if (error
== ETIMEDOUT
)
565 cv_signal(kcondvar_t
*cv
)
567 ASSERT3U(cv
->cv_magic
, ==, CV_MAGIC
);
568 VERIFY0(pthread_cond_signal(&cv
->cv
));
572 cv_broadcast(kcondvar_t
*cv
)
574 ASSERT3U(cv
->cv_magic
, ==, CV_MAGIC
);
575 VERIFY0(pthread_cond_broadcast(&cv
->cv
));
579 * =========================================================================
581 * =========================================================================
584 * Note: for the xxxat() versions of these functions, we assume that the
585 * starting vp is always rootdir (which is true for spa_directory.c, the only
586 * ZFS consumer of these interfaces). We assert this is true, and then emulate
587 * them by adding '/' in front of the path.
592 vn_open(char *path
, int x1
, int flags
, int mode
, vnode_t
**vpp
, int x2
, int x3
)
602 realpath
= umem_alloc(MAXPATHLEN
, UMEM_NOFAIL
);
605 * If we're accessing a real disk from userland, we need to use
606 * the character interface to avoid caching. This is particularly
607 * important if we're trying to look at a real in-kernel storage
608 * pool from userland, e.g. via zdb, because otherwise we won't
609 * see the changes occurring under the segmap cache.
610 * On the other hand, the stupid character device returns zero
611 * for its size. So -- gag -- we open the block device to get
612 * its size, and remember it for subsequent VOP_GETATTR().
614 #if defined(__sun__) || defined(__sun)
615 if (strncmp(path
, "/dev/", 5) == 0) {
620 fd
= open64(path
, O_RDONLY
);
626 if (fstat64(fd
, &st
) == -1) {
633 (void) sprintf(realpath
, "%s", path
);
634 dsk
= strstr(path
, "/dsk/");
636 (void) sprintf(realpath
+ (dsk
- path
) + 1, "r%s",
639 (void) sprintf(realpath
, "%s", path
);
640 if (!(flags
& FCREAT
) && stat64(realpath
, &st
) == -1) {
647 if (!(flags
& FCREAT
) && S_ISBLK(st
.st_mode
)) {
651 /* We shouldn't be writing to block devices in userspace */
652 VERIFY(!(flags
& FWRITE
));
656 old_umask
= umask(0);
659 * The construct 'flags - FREAD' conveniently maps combinations of
660 * FREAD and FWRITE to the corresponding O_RDONLY, O_WRONLY, and O_RDWR.
662 fd
= open64(realpath
, flags
- FREAD
, mode
);
666 (void) umask(old_umask
);
668 if (vn_dumpdir
!= NULL
) {
669 char *dumppath
= umem_zalloc(MAXPATHLEN
, UMEM_NOFAIL
);
670 (void) snprintf(dumppath
, MAXPATHLEN
,
671 "%s/%s", vn_dumpdir
, basename(realpath
));
672 dump_fd
= open64(dumppath
, O_CREAT
| O_WRONLY
, 0666);
673 umem_free(dumppath
, MAXPATHLEN
);
689 if (fstat64_blk(fd
, &st
) == -1) {
695 (void) fcntl(fd
, F_SETFD
, FD_CLOEXEC
);
697 *vpp
= vp
= umem_zalloc(sizeof (vnode_t
), UMEM_NOFAIL
);
700 vp
->v_size
= st
.st_size
;
701 vp
->v_path
= spa_strdup(path
);
702 vp
->v_dump_fd
= dump_fd
;
709 vn_openat(char *path
, int x1
, int flags
, int mode
, vnode_t
**vpp
, int x2
,
710 int x3
, vnode_t
*startvp
, int fd
)
712 char *realpath
= umem_alloc(strlen(path
) + 2, UMEM_NOFAIL
);
715 ASSERT(startvp
== rootdir
);
716 (void) sprintf(realpath
, "/%s", path
);
718 /* fd ignored for now, need if want to simulate nbmand support */
719 ret
= vn_open(realpath
, x1
, flags
, mode
, vpp
, x2
, x3
);
721 umem_free(realpath
, strlen(path
) + 2);
728 vn_rdwr(int uio
, vnode_t
*vp
, void *addr
, ssize_t len
, offset_t offset
,
729 int x1
, int x2
, rlim64_t x3
, void *x4
, ssize_t
*residp
)
731 ssize_t rc
, done
= 0, split
;
733 if (uio
== UIO_READ
) {
734 rc
= pread64(vp
->v_fd
, addr
, len
, offset
);
735 if (vp
->v_dump_fd
!= -1) {
737 status
= pwrite64(vp
->v_dump_fd
, addr
, rc
, offset
);
738 ASSERT(status
!= -1);
742 * To simulate partial disk writes, we split writes into two
743 * system calls so that the process can be killed in between.
745 int sectors
= len
>> SPA_MINBLOCKSHIFT
;
746 split
= (sectors
> 0 ? rand() % sectors
: 0) <<
748 rc
= pwrite64(vp
->v_fd
, addr
, split
, offset
);
751 rc
= pwrite64(vp
->v_fd
, (char *)addr
+ split
,
752 len
- split
, offset
+ split
);
757 if (rc
== -1 && errno
== EINVAL
) {
759 * Under Linux, this most likely means an alignment issue
760 * (memory or disk) due to O_DIRECT, so we abort() in order to
761 * catch the offender.
772 *residp
= len
- done
;
773 else if (done
!= len
)
779 vn_close(vnode_t
*vp
)
782 if (vp
->v_dump_fd
!= -1)
783 close(vp
->v_dump_fd
);
784 spa_strfree(vp
->v_path
);
785 umem_free(vp
, sizeof (vnode_t
));
789 * At a minimum we need to update the size since vdev_reopen()
790 * will no longer call vn_openat().
793 fop_getattr(vnode_t
*vp
, vattr_t
*vap
)
798 if (fstat64_blk(vp
->v_fd
, &st
) == -1) {
804 vap
->va_size
= st
.st_size
;
809 * =========================================================================
810 * Figure out which debugging statements to print
811 * =========================================================================
814 static char *dprintf_string
;
815 static int dprintf_print_all
;
818 dprintf_find_string(const char *string
)
820 char *tmp_str
= dprintf_string
;
821 int len
= strlen(string
);
824 * Find out if this is a string we want to print.
825 * String format: file1.c,function_name1,file2.c,file3.c
828 while (tmp_str
!= NULL
) {
829 if (strncmp(tmp_str
, string
, len
) == 0 &&
830 (tmp_str
[len
] == ',' || tmp_str
[len
] == '\0'))
832 tmp_str
= strchr(tmp_str
, ',');
834 tmp_str
++; /* Get rid of , */
840 dprintf_setup(int *argc
, char **argv
)
845 * Debugging can be specified two ways: by setting the
846 * environment variable ZFS_DEBUG, or by including a
847 * "debug=..." argument on the command line. The command
848 * line setting overrides the environment variable.
851 for (i
= 1; i
< *argc
; i
++) {
852 int len
= strlen("debug=");
853 /* First look for a command line argument */
854 if (strncmp("debug=", argv
[i
], len
) == 0) {
855 dprintf_string
= argv
[i
] + len
;
856 /* Remove from args */
857 for (j
= i
; j
< *argc
; j
++)
864 if (dprintf_string
== NULL
) {
865 /* Look for ZFS_DEBUG environment variable */
866 dprintf_string
= getenv("ZFS_DEBUG");
870 * Are we just turning on all debugging?
872 if (dprintf_find_string("on"))
873 dprintf_print_all
= 1;
875 if (dprintf_string
!= NULL
)
876 zfs_flags
|= ZFS_DEBUG_DPRINTF
;
880 * =========================================================================
882 * =========================================================================
885 __dprintf(const char *file
, const char *func
, int line
, const char *fmt
, ...)
891 * Get rid of annoying "../common/" prefix to filename.
893 newfile
= strrchr(file
, '/');
894 if (newfile
!= NULL
) {
895 newfile
= newfile
+ 1; /* Get rid of leading / */
900 if (dprintf_print_all
||
901 dprintf_find_string(newfile
) ||
902 dprintf_find_string(func
)) {
903 /* Print out just the function name if requested */
905 if (dprintf_find_string("pid"))
906 (void) printf("%d ", getpid());
907 if (dprintf_find_string("tid"))
908 (void) printf("%u ", (uint_t
) pthread_self());
909 if (dprintf_find_string("cpu"))
910 (void) printf("%u ", getcpuid());
911 if (dprintf_find_string("time"))
912 (void) printf("%llu ", gethrtime());
913 if (dprintf_find_string("long"))
914 (void) printf("%s, line %d: ", newfile
, line
);
915 (void) printf("%s: ", func
);
917 (void) vprintf(fmt
, adx
);
924 * =========================================================================
925 * cmn_err() and panic()
926 * =========================================================================
928 static char ce_prefix
[CE_IGNORE
][10] = { "", "NOTICE: ", "WARNING: ", "" };
929 static char ce_suffix
[CE_IGNORE
][2] = { "", "\n", "\n", "" };
932 vpanic(const char *fmt
, va_list adx
)
934 (void) fprintf(stderr
, "error: ");
935 (void) vfprintf(stderr
, fmt
, adx
);
936 (void) fprintf(stderr
, "\n");
938 abort(); /* think of it as a "user-level crash dump" */
942 panic(const char *fmt
, ...)
952 vcmn_err(int ce
, const char *fmt
, va_list adx
)
956 if (ce
!= CE_NOTE
) { /* suppress noise in userland stress testing */
957 (void) fprintf(stderr
, "%s", ce_prefix
[ce
]);
958 (void) vfprintf(stderr
, fmt
, adx
);
959 (void) fprintf(stderr
, "%s", ce_suffix
[ce
]);
965 cmn_err(int ce
, const char *fmt
, ...)
970 vcmn_err(ce
, fmt
, adx
);
975 * =========================================================================
977 * =========================================================================
980 kobj_open_file(char *name
)
985 /* set vp as the _fd field of the file */
986 if (vn_openat(name
, UIO_SYSSPACE
, FREAD
, 0, &vp
, 0, 0, rootdir
,
988 return ((void *)-1UL);
990 file
= umem_zalloc(sizeof (struct _buf
), UMEM_NOFAIL
);
991 file
->_fd
= (intptr_t)vp
;
996 kobj_read_file(struct _buf
*file
, char *buf
, unsigned size
, unsigned off
)
1000 if (vn_rdwr(UIO_READ
, (vnode_t
*)file
->_fd
, buf
, size
, (offset_t
)off
,
1001 UIO_SYSSPACE
, 0, 0, 0, &resid
) != 0)
1004 return (size
- resid
);
1008 kobj_close_file(struct _buf
*file
)
1010 vn_close((vnode_t
*)file
->_fd
);
1011 umem_free(file
, sizeof (struct _buf
));
1015 kobj_get_filesize(struct _buf
*file
, uint64_t *size
)
1018 vnode_t
*vp
= (vnode_t
*)file
->_fd
;
1020 if (fstat64(vp
->v_fd
, &st
) == -1) {
1029 * =========================================================================
1031 * =========================================================================
1035 delay(clock_t ticks
)
1037 poll(0, 0, ticks
* (1000 / hz
));
1041 * Find highest one bit set.
1042 * Returns bit number + 1 of highest bit that is set, otherwise returns 0.
1043 * High order bit is 31 (or 63 in _LP64 kernel).
1046 highbit64(uint64_t i
)
1052 if (i
& 0xffffffff00000000ULL
) {
1055 if (i
& 0xffff0000) {
1074 * Find lowest one bit set.
1075 * Returns bit number + 1 of lowest bit that is set, otherwise returns 0.
1076 * This is basically a reimplementation of ffsll(), which is GNU specific.
1079 lowbit64(uint64_t i
)
1081 register int h
= 64;
1085 if (i
& 0x00000000ffffffffULL
)
1117 static int random_fd
= -1, urandom_fd
= -1;
1120 random_get_bytes_common(uint8_t *ptr
, size_t len
, int fd
)
1127 while (resid
!= 0) {
1128 bytes
= read(fd
, ptr
, resid
);
1129 ASSERT3S(bytes
, >=, 0);
1138 random_get_bytes(uint8_t *ptr
, size_t len
)
1140 return (random_get_bytes_common(ptr
, len
, random_fd
));
1144 random_get_pseudo_bytes(uint8_t *ptr
, size_t len
)
1146 return (random_get_bytes_common(ptr
, len
, urandom_fd
));
1150 ddi_strtoul(const char *hw_serial
, char **nptr
, int base
, unsigned long *result
)
1154 *result
= strtoul(hw_serial
, &end
, base
);
1161 ddi_strtoull(const char *str
, char **nptr
, int base
, u_longlong_t
*result
)
1165 *result
= strtoull(str
, &end
, base
);
1174 return (&hw_utsname
);
1178 * =========================================================================
1179 * kernel emulation setup & teardown
1180 * =========================================================================
1183 umem_out_of_memory(void)
1185 char errmsg
[] = "out of memory -- generating core dump\n";
1187 (void) fprintf(stderr
, "%s", errmsg
);
1192 static unsigned long
1193 get_spl_hostid(void)
1196 unsigned long hostid
;
1198 f
= fopen("/sys/module/spl/parameters/spl_hostid", "r");
1201 if (fscanf(f
, "%lu", &hostid
) != 1)
1204 return (hostid
& 0xffffffff);
1208 get_system_hostid(void)
1210 unsigned long system_hostid
= get_spl_hostid();
1211 if (system_hostid
== 0)
1212 system_hostid
= gethostid() & 0xffffffff;
1213 return (system_hostid
);
1217 kernel_init(int mode
)
1219 extern uint_t rrw_tsd_key
;
1221 umem_nofail_callback(umem_out_of_memory
);
1223 physmem
= sysconf(_SC_PHYS_PAGES
);
1225 dprintf("physmem = %llu pages (%.2f GB)\n", physmem
,
1226 (double)physmem
* sysconf(_SC_PAGE_SIZE
) / (1ULL << 30));
1228 (void) snprintf(hw_serial
, sizeof (hw_serial
), "%ld",
1229 (mode
& FWRITE
) ? get_system_hostid() : 0);
1231 VERIFY((random_fd
= open("/dev/random", O_RDONLY
)) != -1);
1232 VERIFY((urandom_fd
= open("/dev/urandom", O_RDONLY
)) != -1);
1233 VERIFY0(uname(&hw_utsname
));
1236 system_taskq_init();
1242 tsd_create(&rrw_tsd_key
, rrw_tsd_destroy
);
1251 system_taskq_fini();
1262 crgetuid(cred_t
*cr
)
1268 crgetruid(cred_t
*cr
)
1274 crgetgid(cred_t
*cr
)
1280 crgetngroups(cred_t
*cr
)
1286 crgetgroups(cred_t
*cr
)
1292 zfs_secpolicy_snapshot_perms(const char *name
, cred_t
*cr
)
1298 zfs_secpolicy_rename_perms(const char *from
, const char *to
, cred_t
*cr
)
1304 zfs_secpolicy_destroy_perms(const char *name
, cred_t
*cr
)
1310 secpolicy_zfs(const cred_t
*cr
)
1316 ksid_lookupdomain(const char *dom
)
1320 kd
= umem_zalloc(sizeof (ksiddomain_t
), UMEM_NOFAIL
);
1321 kd
->kd_name
= spa_strdup(dom
);
1326 ksiddomain_rele(ksiddomain_t
*ksid
)
1328 spa_strfree(ksid
->kd_name
);
1329 umem_free(ksid
, sizeof (ksiddomain_t
));
1333 kmem_vasprintf(const char *fmt
, va_list adx
)
1338 va_copy(adx_copy
, adx
);
1339 VERIFY(vasprintf(&buf
, fmt
, adx_copy
) != -1);
1346 kmem_asprintf(const char *fmt
, ...)
1352 VERIFY(vasprintf(&buf
, fmt
, adx
) != -1);
1360 zfs_onexit_fd_hold(int fd
, minor_t
*minorp
)
1368 zfs_onexit_fd_rele(int fd
)
1374 zfs_onexit_add_cb(minor_t minor
, void (*func
)(void *), void *data
,
1375 uint64_t *action_handle
)
1382 zfs_onexit_del_cb(minor_t minor
, uint64_t action_handle
, boolean_t fire
)
1389 zfs_onexit_cb_data(minor_t minor
, uint64_t action_handle
, void **data
)
1395 spl_fstrans_mark(void)
1397 return ((fstrans_cookie_t
) 0);
1401 spl_fstrans_unmark(fstrans_cookie_t cookie
)
1406 spl_fstrans_check(void)
1411 void *zvol_tag
= "zvol_tag";
1414 zvol_create_minors(spa_t
*spa
, const char *name
, boolean_t async
)
1419 zvol_remove_minor(spa_t
*spa
, const char *name
, boolean_t async
)
1424 zvol_remove_minors(spa_t
*spa
, const char *name
, boolean_t async
)
1429 zvol_rename_minors(spa_t
*spa
, const char *oldname
, const char *newname
,