]>
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 2009 Sun Microsystems, Inc. All rights reserved.
23 * Use is subject to license terms.
35 #include <sys/processor.h>
36 #include <sys/zfs_context.h>
38 #include <sys/utsname.h>
39 #include <sys/systeminfo.h>
42 * 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"
54 * =========================================================================
56 * =========================================================================
60 zk_thread_create(void (*func
)(), void *arg
)
64 VERIFY(thr_create(0, 0, (void *(*)(void *))func
, arg
, THR_DETACHED
,
67 return ((void *)(uintptr_t)tid
);
71 * =========================================================================
73 * =========================================================================
77 kstat_create(char *module
, int instance
, char *name
, char *class,
78 uchar_t type
, ulong_t ndata
, uchar_t ks_flag
)
85 kstat_install(kstat_t
*ksp
)
90 kstat_delete(kstat_t
*ksp
)
94 * =========================================================================
96 * =========================================================================
99 zmutex_init(kmutex_t
*mp
)
102 mp
->initialized
= B_TRUE
;
103 (void) _mutex_init(&mp
->m_lock
, USYNC_THREAD
, NULL
);
107 zmutex_destroy(kmutex_t
*mp
)
109 ASSERT(mp
->initialized
== B_TRUE
);
110 ASSERT(mp
->m_owner
== NULL
);
111 (void) _mutex_destroy(&(mp
)->m_lock
);
112 mp
->m_owner
= (void *)-1UL;
113 mp
->initialized
= B_FALSE
;
117 mutex_enter(kmutex_t
*mp
)
119 ASSERT(mp
->initialized
== B_TRUE
);
120 ASSERT(mp
->m_owner
!= (void *)-1UL);
121 ASSERT(mp
->m_owner
!= curthread
);
122 VERIFY(mutex_lock(&mp
->m_lock
) == 0);
123 ASSERT(mp
->m_owner
== NULL
);
124 mp
->m_owner
= curthread
;
128 mutex_tryenter(kmutex_t
*mp
)
130 ASSERT(mp
->initialized
== B_TRUE
);
131 ASSERT(mp
->m_owner
!= (void *)-1UL);
132 if (0 == mutex_trylock(&mp
->m_lock
)) {
133 ASSERT(mp
->m_owner
== NULL
);
134 mp
->m_owner
= curthread
;
142 mutex_exit(kmutex_t
*mp
)
144 ASSERT(mp
->initialized
== B_TRUE
);
145 ASSERT(mutex_owner(mp
) == curthread
);
147 VERIFY(mutex_unlock(&mp
->m_lock
) == 0);
151 mutex_owner(kmutex_t
*mp
)
153 ASSERT(mp
->initialized
== B_TRUE
);
154 return (mp
->m_owner
);
158 * =========================================================================
160 * =========================================================================
164 rw_init(krwlock_t
*rwlp
, char *name
, int type
, void *arg
)
166 rwlock_init(&rwlp
->rw_lock
, USYNC_THREAD
, NULL
);
167 rwlp
->rw_owner
= NULL
;
168 rwlp
->initialized
= B_TRUE
;
172 rw_destroy(krwlock_t
*rwlp
)
174 rwlock_destroy(&rwlp
->rw_lock
);
175 rwlp
->rw_owner
= (void *)-1UL;
176 rwlp
->initialized
= B_FALSE
;
180 rw_enter(krwlock_t
*rwlp
, krw_t rw
)
182 ASSERT(!RW_LOCK_HELD(rwlp
));
183 ASSERT(rwlp
->initialized
== B_TRUE
);
184 ASSERT(rwlp
->rw_owner
!= (void *)-1UL);
185 ASSERT(rwlp
->rw_owner
!= curthread
);
188 VERIFY(rw_rdlock(&rwlp
->rw_lock
) == 0);
190 VERIFY(rw_wrlock(&rwlp
->rw_lock
) == 0);
192 rwlp
->rw_owner
= curthread
;
196 rw_exit(krwlock_t
*rwlp
)
198 ASSERT(rwlp
->initialized
== B_TRUE
);
199 ASSERT(rwlp
->rw_owner
!= (void *)-1UL);
201 rwlp
->rw_owner
= NULL
;
202 VERIFY(rw_unlock(&rwlp
->rw_lock
) == 0);
206 rw_tryenter(krwlock_t
*rwlp
, krw_t rw
)
210 ASSERT(rwlp
->initialized
== B_TRUE
);
211 ASSERT(rwlp
->rw_owner
!= (void *)-1UL);
214 rv
= rw_tryrdlock(&rwlp
->rw_lock
);
216 rv
= rw_trywrlock(&rwlp
->rw_lock
);
219 rwlp
->rw_owner
= curthread
;
228 rw_tryupgrade(krwlock_t
*rwlp
)
230 ASSERT(rwlp
->initialized
== B_TRUE
);
231 ASSERT(rwlp
->rw_owner
!= (void *)-1UL);
237 * =========================================================================
238 * condition variables
239 * =========================================================================
243 cv_init(kcondvar_t
*cv
, char *name
, int type
, void *arg
)
245 VERIFY(cond_init(cv
, type
, NULL
) == 0);
249 cv_destroy(kcondvar_t
*cv
)
251 VERIFY(cond_destroy(cv
) == 0);
255 cv_wait(kcondvar_t
*cv
, kmutex_t
*mp
)
257 ASSERT(mutex_owner(mp
) == curthread
);
259 int ret
= cond_wait(cv
, &mp
->m_lock
);
260 VERIFY(ret
== 0 || ret
== EINTR
);
261 mp
->m_owner
= curthread
;
265 cv_timedwait(kcondvar_t
*cv
, kmutex_t
*mp
, clock_t abstime
)
272 delta
= abstime
- lbolt
;
276 ts
.tv_sec
= delta
/ hz
;
277 ts
.tv_nsec
= (delta
% hz
) * (NANOSEC
/ hz
);
279 ASSERT(mutex_owner(mp
) == curthread
);
281 error
= cond_reltimedwait(cv
, &mp
->m_lock
, &ts
);
282 mp
->m_owner
= curthread
;
296 cv_signal(kcondvar_t
*cv
)
298 VERIFY(cond_signal(cv
) == 0);
302 cv_broadcast(kcondvar_t
*cv
)
304 VERIFY(cond_broadcast(cv
) == 0);
308 * =========================================================================
310 * =========================================================================
313 * Note: for the xxxat() versions of these functions, we assume that the
314 * starting vp is always rootdir (which is true for spa_directory.c, the only
315 * ZFS consumer of these interfaces). We assert this is true, and then emulate
316 * them by adding '/' in front of the path.
321 vn_open(char *path
, int x1
, int flags
, int mode
, vnode_t
**vpp
, int x2
, int x3
)
326 char realpath
[MAXPATHLEN
];
330 * If we're accessing a real disk from userland, we need to use
331 * the character interface to avoid caching. This is particularly
332 * important if we're trying to look at a real in-kernel storage
333 * pool from userland, e.g. via zdb, because otherwise we won't
334 * see the changes occurring under the segmap cache.
335 * On the other hand, the stupid character device returns zero
336 * for its size. So -- gag -- we open the block device to get
337 * its size, and remember it for subsequent VOP_GETATTR().
339 if (strncmp(path
, "/dev/", 5) == 0) {
341 fd
= open64(path
, O_RDONLY
);
344 if (fstat64(fd
, &st
) == -1) {
349 (void) sprintf(realpath
, "%s", path
);
350 dsk
= strstr(path
, "/dsk/");
352 (void) sprintf(realpath
+ (dsk
- path
) + 1, "r%s",
355 (void) sprintf(realpath
, "%s", path
);
356 if (!(flags
& FCREAT
) && stat64(realpath
, &st
) == -1)
361 old_umask
= umask(0);
364 * The construct 'flags - FREAD' conveniently maps combinations of
365 * FREAD and FWRITE to the corresponding O_RDONLY, O_WRONLY, and O_RDWR.
367 fd
= open64(realpath
, flags
- FREAD
, mode
);
370 (void) umask(old_umask
);
375 if (fstat64(fd
, &st
) == -1) {
380 (void) fcntl(fd
, F_SETFD
, FD_CLOEXEC
);
382 *vpp
= vp
= umem_zalloc(sizeof (vnode_t
), UMEM_NOFAIL
);
385 vp
->v_size
= st
.st_size
;
386 vp
->v_path
= spa_strdup(path
);
393 vn_openat(char *path
, int x1
, int flags
, int mode
, vnode_t
**vpp
, int x2
,
394 int x3
, vnode_t
*startvp
, int fd
)
396 char *realpath
= umem_alloc(strlen(path
) + 2, UMEM_NOFAIL
);
399 ASSERT(startvp
== rootdir
);
400 (void) sprintf(realpath
, "/%s", path
);
402 /* fd ignored for now, need if want to simulate nbmand support */
403 ret
= vn_open(realpath
, x1
, flags
, mode
, vpp
, x2
, x3
);
405 umem_free(realpath
, strlen(path
) + 2);
412 vn_rdwr(int uio
, vnode_t
*vp
, void *addr
, ssize_t len
, offset_t offset
,
413 int x1
, int x2
, rlim64_t x3
, void *x4
, ssize_t
*residp
)
415 ssize_t iolen
, split
;
417 if (uio
== UIO_READ
) {
418 iolen
= pread64(vp
->v_fd
, addr
, len
, offset
);
421 * To simulate partial disk writes, we split writes into two
422 * system calls so that the process can be killed in between.
424 split
= (len
> 0 ? rand() % len
: 0);
425 iolen
= pwrite64(vp
->v_fd
, addr
, split
, offset
);
426 iolen
+= pwrite64(vp
->v_fd
, (char *)addr
+ split
,
427 len
- split
, offset
+ split
);
433 *residp
= len
- iolen
;
434 else if (iolen
!= len
)
440 vn_close(vnode_t
*vp
)
443 spa_strfree(vp
->v_path
);
444 umem_free(vp
, sizeof (vnode_t
));
450 * =========================================================================
451 * Figure out which debugging statements to print
452 * =========================================================================
455 static char *dprintf_string
;
456 static int dprintf_print_all
;
459 dprintf_find_string(const char *string
)
461 char *tmp_str
= dprintf_string
;
462 int len
= strlen(string
);
465 * Find out if this is a string we want to print.
466 * String format: file1.c,function_name1,file2.c,file3.c
469 while (tmp_str
!= NULL
) {
470 if (strncmp(tmp_str
, string
, len
) == 0 &&
471 (tmp_str
[len
] == ',' || tmp_str
[len
] == '\0'))
473 tmp_str
= strchr(tmp_str
, ',');
475 tmp_str
++; /* Get rid of , */
481 dprintf_setup(int *argc
, char **argv
)
486 * Debugging can be specified two ways: by setting the
487 * environment variable ZFS_DEBUG, or by including a
488 * "debug=..." argument on the command line. The command
489 * line setting overrides the environment variable.
492 for (i
= 1; i
< *argc
; i
++) {
493 int len
= strlen("debug=");
494 /* First look for a command line argument */
495 if (strncmp("debug=", argv
[i
], len
) == 0) {
496 dprintf_string
= argv
[i
] + len
;
497 /* Remove from args */
498 for (j
= i
; j
< *argc
; j
++)
505 if (dprintf_string
== NULL
) {
506 /* Look for ZFS_DEBUG environment variable */
507 dprintf_string
= getenv("ZFS_DEBUG");
511 * Are we just turning on all debugging?
513 if (dprintf_find_string("on"))
514 dprintf_print_all
= 1;
518 * =========================================================================
520 * =========================================================================
523 __dprintf(const char *file
, const char *func
, int line
, const char *fmt
, ...)
529 * Get rid of annoying "../common/" prefix to filename.
531 newfile
= strrchr(file
, '/');
532 if (newfile
!= NULL
) {
533 newfile
= newfile
+ 1; /* Get rid of leading / */
538 if (dprintf_print_all
||
539 dprintf_find_string(newfile
) ||
540 dprintf_find_string(func
)) {
541 /* Print out just the function name if requested */
543 if (dprintf_find_string("pid"))
544 (void) printf("%d ", getpid());
545 if (dprintf_find_string("tid"))
546 (void) printf("%u ", thr_self());
547 if (dprintf_find_string("cpu"))
548 (void) printf("%u ", getcpuid());
549 if (dprintf_find_string("time"))
550 (void) printf("%llu ", gethrtime());
551 if (dprintf_find_string("long"))
552 (void) printf("%s, line %d: ", newfile
, line
);
553 (void) printf("%s: ", func
);
555 (void) vprintf(fmt
, adx
);
561 #endif /* ZFS_DEBUG */
564 * =========================================================================
565 * cmn_err() and panic()
566 * =========================================================================
568 static char ce_prefix
[CE_IGNORE
][10] = { "", "NOTICE: ", "WARNING: ", "" };
569 static char ce_suffix
[CE_IGNORE
][2] = { "", "\n", "\n", "" };
572 vpanic(const char *fmt
, va_list adx
)
574 (void) fprintf(stderr
, "error: ");
575 (void) vfprintf(stderr
, fmt
, adx
);
576 (void) fprintf(stderr
, "\n");
578 abort(); /* think of it as a "user-level crash dump" */
582 panic(const char *fmt
, ...)
592 vcmn_err(int ce
, const char *fmt
, va_list adx
)
596 if (ce
!= CE_NOTE
) { /* suppress noise in userland stress testing */
597 (void) fprintf(stderr
, "%s", ce_prefix
[ce
]);
598 (void) vfprintf(stderr
, fmt
, adx
);
599 (void) fprintf(stderr
, "%s", ce_suffix
[ce
]);
605 cmn_err(int ce
, const char *fmt
, ...)
610 vcmn_err(ce
, fmt
, adx
);
615 * =========================================================================
617 * =========================================================================
620 kobj_open_file(char *name
)
625 /* set vp as the _fd field of the file */
626 if (vn_openat(name
, UIO_SYSSPACE
, FREAD
, 0, &vp
, 0, 0, rootdir
,
628 return ((void *)-1UL);
630 file
= umem_zalloc(sizeof (struct _buf
), UMEM_NOFAIL
);
631 file
->_fd
= (intptr_t)vp
;
636 kobj_read_file(struct _buf
*file
, char *buf
, unsigned size
, unsigned off
)
640 vn_rdwr(UIO_READ
, (vnode_t
*)file
->_fd
, buf
, size
, (offset_t
)off
,
641 UIO_SYSSPACE
, 0, 0, 0, &resid
);
643 return (size
- resid
);
647 kobj_close_file(struct _buf
*file
)
649 vn_close((vnode_t
*)file
->_fd
);
650 umem_free(file
, sizeof (struct _buf
));
654 kobj_get_filesize(struct _buf
*file
, uint64_t *size
)
657 vnode_t
*vp
= (vnode_t
*)file
->_fd
;
659 if (fstat64(vp
->v_fd
, &st
) == -1) {
668 * =========================================================================
670 * =========================================================================
676 poll(0, 0, ticks
* (1000 / hz
));
680 * Find highest one bit set.
681 * Returns bit number + 1 of highest bit that is set, otherwise returns 0.
682 * High order bit is 31 (or 63 in _LP64 kernel).
692 if (i
& 0xffffffff00000000ul
) {
696 if (i
& 0xffff0000) {
714 static int random_fd
= -1, urandom_fd
= -1;
717 random_get_bytes_common(uint8_t *ptr
, size_t len
, int fd
)
725 bytes
= read(fd
, ptr
, resid
);
726 ASSERT3S(bytes
, >=, 0);
735 random_get_bytes(uint8_t *ptr
, size_t len
)
737 return (random_get_bytes_common(ptr
, len
, random_fd
));
741 random_get_pseudo_bytes(uint8_t *ptr
, size_t len
)
743 return (random_get_bytes_common(ptr
, len
, urandom_fd
));
747 ddi_strtoul(const char *hw_serial
, char **nptr
, int base
, unsigned long *result
)
751 *result
= strtoul(hw_serial
, &end
, base
);
758 * =========================================================================
759 * kernel emulation setup & teardown
760 * =========================================================================
763 umem_out_of_memory(void)
765 char errmsg
[] = "out of memory -- generating core dump\n";
767 write(fileno(stderr
), errmsg
, sizeof (errmsg
));
773 kernel_init(int mode
)
775 umem_nofail_callback(umem_out_of_memory
);
777 physmem
= sysconf(_SC_PHYS_PAGES
);
779 dprintf("physmem = %llu pages (%.2f GB)\n", physmem
,
780 (double)physmem
* sysconf(_SC_PAGE_SIZE
) / (1ULL << 30));
782 (void) snprintf(hw_serial
, sizeof (hw_serial
), "%ld", gethostid());
784 VERIFY((random_fd
= open("/dev/random", O_RDONLY
)) != -1);
785 VERIFY((urandom_fd
= open("/dev/urandom", O_RDONLY
)) != -1);
805 z_uncompress(void *dst
, size_t *dstlen
, const void *src
, size_t srclen
)
808 uLongf len
= *dstlen
;
810 if ((ret
= uncompress(dst
, &len
, src
, srclen
)) == Z_OK
)
811 *dstlen
= (size_t)len
;
817 z_compress_level(void *dst
, size_t *dstlen
, const void *src
, size_t srclen
,
821 uLongf len
= *dstlen
;
823 if ((ret
= compress2(dst
, &len
, src
, srclen
, level
)) == Z_OK
)
824 *dstlen
= (size_t)len
;
842 crgetngroups(cred_t
*cr
)
848 crgetgroups(cred_t
*cr
)
854 zfs_secpolicy_snapshot_perms(const char *name
, cred_t
*cr
)
860 zfs_secpolicy_rename_perms(const char *from
, const char *to
, cred_t
*cr
)
866 zfs_secpolicy_destroy_perms(const char *name
, cred_t
*cr
)
872 ksid_lookupdomain(const char *dom
)
876 kd
= umem_zalloc(sizeof (ksiddomain_t
), UMEM_NOFAIL
);
877 kd
->kd_name
= spa_strdup(dom
);
882 ksiddomain_rele(ksiddomain_t
*ksid
)
884 spa_strfree(ksid
->kd_name
);
885 umem_free(ksid
, sizeof (ksiddomain_t
));