]>
git.proxmox.com Git - mirror_zfs.git/blob - lib/libzpool/kernel.c
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.
34 #include <sys/processor.h>
35 #include <sys/zfs_context.h>
37 #include <sys/utsname.h>
38 #include <sys/systeminfo.h>
41 * Emulation of kernel services in userland.
46 vnode_t
*rootdir
= (vnode_t
*)0xabcd1234;
47 char hw_serial
[HW_HOSTID_LEN
];
49 struct utsname utsname
= {
50 "userland", "libzpool", "1", "1", "na"
53 /* this only exists to have its address taken */
57 * =========================================================================
59 * =========================================================================
63 zk_thread_create(void (*func
)(), void *arg
)
67 VERIFY(thr_create(0, 0, (void *(*)(void *))func
, arg
, THR_DETACHED
,
70 return ((void *)(uintptr_t)tid
);
74 * =========================================================================
76 * =========================================================================
80 kstat_create(char *module
, int instance
, char *name
, char *class,
81 uchar_t type
, ulong_t ndata
, uchar_t ks_flag
)
88 kstat_install(kstat_t
*ksp
)
93 kstat_delete(kstat_t
*ksp
)
97 * =========================================================================
99 * =========================================================================
102 zmutex_init(kmutex_t
*mp
)
105 mp
->initialized
= B_TRUE
;
106 (void) _mutex_init(&mp
->m_lock
, USYNC_THREAD
, NULL
);
110 zmutex_destroy(kmutex_t
*mp
)
112 ASSERT(mp
->initialized
== B_TRUE
);
113 ASSERT(mp
->m_owner
== NULL
);
114 (void) _mutex_destroy(&(mp
)->m_lock
);
115 mp
->m_owner
= (void *)-1UL;
116 mp
->initialized
= B_FALSE
;
120 mutex_enter(kmutex_t
*mp
)
122 ASSERT(mp
->initialized
== B_TRUE
);
123 ASSERT(mp
->m_owner
!= (void *)-1UL);
124 ASSERT(mp
->m_owner
!= curthread
);
125 VERIFY(mutex_lock(&mp
->m_lock
) == 0);
126 ASSERT(mp
->m_owner
== NULL
);
127 mp
->m_owner
= curthread
;
131 mutex_tryenter(kmutex_t
*mp
)
133 ASSERT(mp
->initialized
== B_TRUE
);
134 ASSERT(mp
->m_owner
!= (void *)-1UL);
135 if (0 == mutex_trylock(&mp
->m_lock
)) {
136 ASSERT(mp
->m_owner
== NULL
);
137 mp
->m_owner
= curthread
;
145 mutex_exit(kmutex_t
*mp
)
147 ASSERT(mp
->initialized
== B_TRUE
);
148 ASSERT(mutex_owner(mp
) == curthread
);
150 VERIFY(mutex_unlock(&mp
->m_lock
) == 0);
154 mutex_owner(kmutex_t
*mp
)
156 ASSERT(mp
->initialized
== B_TRUE
);
157 return (mp
->m_owner
);
161 * =========================================================================
163 * =========================================================================
167 rw_init(krwlock_t
*rwlp
, char *name
, int type
, void *arg
)
169 rwlock_init(&rwlp
->rw_lock
, USYNC_THREAD
, NULL
);
170 rwlp
->rw_owner
= NULL
;
171 rwlp
->initialized
= B_TRUE
;
175 rw_destroy(krwlock_t
*rwlp
)
177 rwlock_destroy(&rwlp
->rw_lock
);
178 rwlp
->rw_owner
= (void *)-1UL;
179 rwlp
->initialized
= B_FALSE
;
183 rw_enter(krwlock_t
*rwlp
, krw_t rw
)
185 ASSERT(!RW_LOCK_HELD(rwlp
));
186 ASSERT(rwlp
->initialized
== B_TRUE
);
187 ASSERT(rwlp
->rw_owner
!= (void *)-1UL);
188 ASSERT(rwlp
->rw_owner
!= curthread
);
191 VERIFY(rw_rdlock(&rwlp
->rw_lock
) == 0);
193 VERIFY(rw_wrlock(&rwlp
->rw_lock
) == 0);
195 rwlp
->rw_owner
= curthread
;
199 rw_exit(krwlock_t
*rwlp
)
201 ASSERT(rwlp
->initialized
== B_TRUE
);
202 ASSERT(rwlp
->rw_owner
!= (void *)-1UL);
204 rwlp
->rw_owner
= NULL
;
205 VERIFY(rw_unlock(&rwlp
->rw_lock
) == 0);
209 rw_tryenter(krwlock_t
*rwlp
, krw_t rw
)
213 ASSERT(rwlp
->initialized
== B_TRUE
);
214 ASSERT(rwlp
->rw_owner
!= (void *)-1UL);
217 rv
= rw_tryrdlock(&rwlp
->rw_lock
);
219 rv
= rw_trywrlock(&rwlp
->rw_lock
);
222 rwlp
->rw_owner
= curthread
;
231 rw_tryupgrade(krwlock_t
*rwlp
)
233 ASSERT(rwlp
->initialized
== B_TRUE
);
234 ASSERT(rwlp
->rw_owner
!= (void *)-1UL);
240 * =========================================================================
241 * condition variables
242 * =========================================================================
246 cv_init(kcondvar_t
*cv
, char *name
, int type
, void *arg
)
248 VERIFY(cond_init(cv
, type
, NULL
) == 0);
252 cv_destroy(kcondvar_t
*cv
)
254 VERIFY(cond_destroy(cv
) == 0);
258 cv_wait(kcondvar_t
*cv
, kmutex_t
*mp
)
260 ASSERT(mutex_owner(mp
) == curthread
);
262 int ret
= cond_wait(cv
, &mp
->m_lock
);
263 VERIFY(ret
== 0 || ret
== EINTR
);
264 mp
->m_owner
= curthread
;
268 cv_timedwait(kcondvar_t
*cv
, kmutex_t
*mp
, clock_t abstime
)
275 delta
= abstime
- ddi_get_lbolt();
279 ts
.tv_sec
= delta
/ hz
;
280 ts
.tv_nsec
= (delta
% hz
) * (NANOSEC
/ hz
);
282 ASSERT(mutex_owner(mp
) == curthread
);
284 error
= cond_reltimedwait(cv
, &mp
->m_lock
, &ts
);
285 mp
->m_owner
= curthread
;
299 cv_signal(kcondvar_t
*cv
)
301 VERIFY(cond_signal(cv
) == 0);
305 cv_broadcast(kcondvar_t
*cv
)
307 VERIFY(cond_broadcast(cv
) == 0);
311 * =========================================================================
313 * =========================================================================
316 * Note: for the xxxat() versions of these functions, we assume that the
317 * starting vp is always rootdir (which is true for spa_directory.c, the only
318 * ZFS consumer of these interfaces). We assert this is true, and then emulate
319 * them by adding '/' in front of the path.
324 vn_open(char *path
, int x1
, int flags
, int mode
, vnode_t
**vpp
, int x2
, int x3
)
329 char realpath
[MAXPATHLEN
];
334 * If we're accessing a real disk from userland, we need to use
335 * the character interface to avoid caching. This is particularly
336 * important if we're trying to look at a real in-kernel storage
337 * pool from userland, e.g. via zdb, because otherwise we won't
338 * see the changes occurring under the segmap cache.
339 * On the other hand, the stupid character device returns zero
340 * for its size. So -- gag -- we open the block device to get
341 * its size, and remember it for subsequent VOP_GETATTR().
343 if (strncmp(path
, "/dev/", 5) == 0) {
345 fd
= open64(path
, O_RDONLY
);
348 if (fstat64(fd
, &st
) == -1) {
353 (void) sprintf(realpath
, "%s", path
);
354 dsk
= strstr(path
, "/dsk/");
356 (void) sprintf(realpath
+ (dsk
- path
) + 1, "r%s",
359 (void) sprintf(realpath
, "%s", path
);
360 if (!(flags
& FCREAT
) && stat64(realpath
, &st
) == -1)
365 old_umask
= umask(0);
368 * The construct 'flags - FREAD' conveniently maps combinations of
369 * FREAD and FWRITE to the corresponding O_RDONLY, O_WRONLY, and O_RDWR.
371 fd
= open64(realpath
, flags
- FREAD
, mode
);
374 (void) umask(old_umask
);
379 if (fstat64(fd
, &st
) == -1) {
385 (void) fcntl(fd
, F_SETFD
, FD_CLOEXEC
);
387 *vpp
= vp
= umem_zalloc(sizeof (vnode_t
), UMEM_NOFAIL
);
390 vp
->v_size
= st
.st_size
;
391 vp
->v_path
= spa_strdup(path
);
398 vn_openat(char *path
, int x1
, int flags
, int mode
, vnode_t
**vpp
, int x2
,
399 int x3
, vnode_t
*startvp
, int fd
)
401 char *realpath
= umem_alloc(strlen(path
) + 2, UMEM_NOFAIL
);
404 ASSERT(startvp
== rootdir
);
405 (void) sprintf(realpath
, "/%s", path
);
407 /* fd ignored for now, need if want to simulate nbmand support */
408 ret
= vn_open(realpath
, x1
, flags
, mode
, vpp
, x2
, x3
);
410 umem_free(realpath
, strlen(path
) + 2);
417 vn_rdwr(int uio
, vnode_t
*vp
, void *addr
, ssize_t len
, offset_t offset
,
418 int x1
, int x2
, rlim64_t x3
, void *x4
, ssize_t
*residp
)
420 ssize_t rc
, done
= 0, split
;
422 if (uio
== UIO_READ
) {
423 rc
= pread64(vp
->v_fd
, addr
, len
, offset
);
426 * To simulate partial disk writes, we split writes into two
427 * system calls so that the process can be killed in between.
429 split
= (len
> 0 ? rand() % len
: 0);
430 rc
= pwrite64(vp
->v_fd
, addr
, split
, offset
);
433 rc
= pwrite64(vp
->v_fd
, (char *)addr
+ split
,
434 len
- split
, offset
+ split
);
444 *residp
= len
- done
;
445 else if (done
!= len
)
451 vn_close(vnode_t
*vp
)
454 spa_strfree(vp
->v_path
);
455 umem_free(vp
, sizeof (vnode_t
));
459 * At a minimum we need to update the size since vdev_reopen()
460 * will no longer call vn_openat().
463 fop_getattr(vnode_t
*vp
, vattr_t
*vap
)
467 if (fstat64(vp
->v_fd
, &st
) == -1) {
472 vap
->va_size
= st
.st_size
;
479 * =========================================================================
480 * Figure out which debugging statements to print
481 * =========================================================================
484 static char *dprintf_string
;
485 static int dprintf_print_all
;
488 dprintf_find_string(const char *string
)
490 char *tmp_str
= dprintf_string
;
491 int len
= strlen(string
);
494 * Find out if this is a string we want to print.
495 * String format: file1.c,function_name1,file2.c,file3.c
498 while (tmp_str
!= NULL
) {
499 if (strncmp(tmp_str
, string
, len
) == 0 &&
500 (tmp_str
[len
] == ',' || tmp_str
[len
] == '\0'))
502 tmp_str
= strchr(tmp_str
, ',');
504 tmp_str
++; /* Get rid of , */
510 dprintf_setup(int *argc
, char **argv
)
515 * Debugging can be specified two ways: by setting the
516 * environment variable ZFS_DEBUG, or by including a
517 * "debug=..." argument on the command line. The command
518 * line setting overrides the environment variable.
521 for (i
= 1; i
< *argc
; i
++) {
522 int len
= strlen("debug=");
523 /* First look for a command line argument */
524 if (strncmp("debug=", argv
[i
], len
) == 0) {
525 dprintf_string
= argv
[i
] + len
;
526 /* Remove from args */
527 for (j
= i
; j
< *argc
; j
++)
534 if (dprintf_string
== NULL
) {
535 /* Look for ZFS_DEBUG environment variable */
536 dprintf_string
= getenv("ZFS_DEBUG");
540 * Are we just turning on all debugging?
542 if (dprintf_find_string("on"))
543 dprintf_print_all
= 1;
547 * =========================================================================
549 * =========================================================================
552 __dprintf(const char *file
, const char *func
, int line
, const char *fmt
, ...)
558 * Get rid of annoying "../common/" prefix to filename.
560 newfile
= strrchr(file
, '/');
561 if (newfile
!= NULL
) {
562 newfile
= newfile
+ 1; /* Get rid of leading / */
567 if (dprintf_print_all
||
568 dprintf_find_string(newfile
) ||
569 dprintf_find_string(func
)) {
570 /* Print out just the function name if requested */
572 if (dprintf_find_string("pid"))
573 (void) printf("%d ", getpid());
574 if (dprintf_find_string("tid"))
575 (void) printf("%u ", thr_self());
576 if (dprintf_find_string("cpu"))
577 (void) printf("%u ", getcpuid());
578 if (dprintf_find_string("time"))
579 (void) printf("%llu ", gethrtime());
580 if (dprintf_find_string("long"))
581 (void) printf("%s, line %d: ", newfile
, line
);
582 (void) printf("%s: ", func
);
584 (void) vprintf(fmt
, adx
);
590 #endif /* ZFS_DEBUG */
593 * =========================================================================
594 * cmn_err() and panic()
595 * =========================================================================
597 static char ce_prefix
[CE_IGNORE
][10] = { "", "NOTICE: ", "WARNING: ", "" };
598 static char ce_suffix
[CE_IGNORE
][2] = { "", "\n", "\n", "" };
601 vpanic(const char *fmt
, va_list adx
)
603 (void) fprintf(stderr
, "error: ");
604 (void) vfprintf(stderr
, fmt
, adx
);
605 (void) fprintf(stderr
, "\n");
607 abort(); /* think of it as a "user-level crash dump" */
611 panic(const char *fmt
, ...)
621 vcmn_err(int ce
, const char *fmt
, va_list adx
)
625 if (ce
!= CE_NOTE
) { /* suppress noise in userland stress testing */
626 (void) fprintf(stderr
, "%s", ce_prefix
[ce
]);
627 (void) vfprintf(stderr
, fmt
, adx
);
628 (void) fprintf(stderr
, "%s", ce_suffix
[ce
]);
634 cmn_err(int ce
, const char *fmt
, ...)
639 vcmn_err(ce
, fmt
, adx
);
644 * =========================================================================
646 * =========================================================================
649 kobj_open_file(char *name
)
654 /* set vp as the _fd field of the file */
655 if (vn_openat(name
, UIO_SYSSPACE
, FREAD
, 0, &vp
, 0, 0, rootdir
,
657 return ((void *)-1UL);
659 file
= umem_zalloc(sizeof (struct _buf
), UMEM_NOFAIL
);
660 file
->_fd
= (intptr_t)vp
;
665 kobj_read_file(struct _buf
*file
, char *buf
, unsigned size
, unsigned off
)
669 vn_rdwr(UIO_READ
, (vnode_t
*)file
->_fd
, buf
, size
, (offset_t
)off
,
670 UIO_SYSSPACE
, 0, 0, 0, &resid
);
672 return (size
- resid
);
676 kobj_close_file(struct _buf
*file
)
678 vn_close((vnode_t
*)file
->_fd
);
679 umem_free(file
, sizeof (struct _buf
));
683 kobj_get_filesize(struct _buf
*file
, uint64_t *size
)
686 vnode_t
*vp
= (vnode_t
*)file
->_fd
;
688 if (fstat64(vp
->v_fd
, &st
) == -1) {
697 * =========================================================================
699 * =========================================================================
705 poll(0, 0, ticks
* (1000 / hz
));
709 * Find highest one bit set.
710 * Returns bit number + 1 of highest bit that is set, otherwise returns 0.
711 * High order bit is 31 (or 63 in _LP64 kernel).
721 if (i
& 0xffffffff00000000ul
) {
725 if (i
& 0xffff0000) {
743 static int random_fd
= -1, urandom_fd
= -1;
746 random_get_bytes_common(uint8_t *ptr
, size_t len
, int fd
)
754 bytes
= read(fd
, ptr
, resid
);
755 ASSERT3S(bytes
, >=, 0);
764 random_get_bytes(uint8_t *ptr
, size_t len
)
766 return (random_get_bytes_common(ptr
, len
, random_fd
));
770 random_get_pseudo_bytes(uint8_t *ptr
, size_t len
)
772 return (random_get_bytes_common(ptr
, len
, urandom_fd
));
776 ddi_strtoul(const char *hw_serial
, char **nptr
, int base
, unsigned long *result
)
780 *result
= strtoul(hw_serial
, &end
, base
);
787 ddi_strtoull(const char *str
, char **nptr
, int base
, u_longlong_t
*result
)
791 *result
= strtoull(str
, &end
, base
);
798 * =========================================================================
799 * kernel emulation setup & teardown
800 * =========================================================================
803 umem_out_of_memory(void)
805 char errmsg
[] = "out of memory -- generating core dump\n";
807 (void) fprintf(stderr
, "%s", errmsg
);
813 kernel_init(int mode
)
815 umem_nofail_callback(umem_out_of_memory
);
817 physmem
= sysconf(_SC_PHYS_PAGES
);
819 dprintf("physmem = %llu pages (%.2f GB)\n", physmem
,
820 (double)physmem
* sysconf(_SC_PAGE_SIZE
) / (1ULL << 30));
822 (void) snprintf(hw_serial
, sizeof (hw_serial
), "%ld",
823 (mode
& FWRITE
) ? gethostid() : 0);
825 VERIFY((random_fd
= open("/dev/random", O_RDONLY
)) != -1);
826 VERIFY((urandom_fd
= open("/dev/urandom", O_RDONLY
)) != -1);
848 z_uncompress(void *dst
, size_t *dstlen
, const void *src
, size_t srclen
)
851 uLongf len
= *dstlen
;
853 if ((ret
= uncompress(dst
, &len
, src
, srclen
)) == Z_OK
)
854 *dstlen
= (size_t)len
;
860 z_compress_level(void *dst
, size_t *dstlen
, const void *src
, size_t srclen
,
864 uLongf len
= *dstlen
;
866 if ((ret
= compress2(dst
, &len
, src
, srclen
, level
)) == Z_OK
)
867 *dstlen
= (size_t)len
;
885 crgetngroups(cred_t
*cr
)
891 crgetgroups(cred_t
*cr
)
897 zfs_secpolicy_snapshot_perms(const char *name
, cred_t
*cr
)
903 zfs_secpolicy_rename_perms(const char *from
, const char *to
, cred_t
*cr
)
909 zfs_secpolicy_destroy_perms(const char *name
, cred_t
*cr
)
915 ksid_lookupdomain(const char *dom
)
919 kd
= umem_zalloc(sizeof (ksiddomain_t
), UMEM_NOFAIL
);
920 kd
->kd_name
= spa_strdup(dom
);
925 ksiddomain_rele(ksiddomain_t
*ksid
)
927 spa_strfree(ksid
->kd_name
);
928 umem_free(ksid
, sizeof (ksiddomain_t
));
932 * Do not change the length of the returned string; it must be freed
936 kmem_asprintf(const char *fmt
, ...)
943 size
= vsnprintf(NULL
, 0, fmt
, adx
) + 1;
946 buf
= kmem_alloc(size
, KM_SLEEP
);
949 size
= vsnprintf(buf
, size
, fmt
, adx
);
957 zfs_onexit_fd_hold(int fd
, minor_t
*minorp
)
965 zfs_onexit_fd_rele(int fd
)
971 zfs_onexit_add_cb(minor_t minor
, void (*func
)(void *), void *data
,
972 uint64_t *action_handle
)
979 zfs_onexit_del_cb(minor_t minor
, uint64_t action_handle
, boolean_t fire
)
986 zfs_onexit_cb_data(minor_t minor
, uint64_t action_handle
, void **data
)