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 * Copyright (c) 2012, 2015 by Delphix. All rights reserved.
24 * Copyright (c) 2013, Joyent, Inc. All rights reserved.
25 */
26
27 #include <assert.h>
28 #include <fcntl.h>
29 #include <poll.h>
30 #include <stdio.h>
31 #include <stdlib.h>
32 #include <string.h>
33 #include <zlib.h>
34 #include <libgen.h>
35 #include <sys/spa.h>
36 #include <sys/stat.h>
37 #include <sys/processor.h>
38 #include <sys/zfs_context.h>
39 #include <sys/rrwlock.h>
40 #include <sys/zmod.h>
41 #include <sys/utsname.h>
42 #include <sys/systeminfo.h>
43
44 #ifdef __NetBSD__
45 #include <sys/dkio.h>
46 #include <sys/ioctl.h>
47 #endif
48
49 /*
50 * Emulation of kernel services in userland.
51 */
52
53 #ifndef __FreeBSD__
54 int aok;
55 #endif
56 uint64_t physmem;
57 vnode_t *rootdir = (vnode_t *)0xabcd1234;
58 char hw_serial[HW_HOSTID_LEN];
59 #ifdef illumos
60 kmutex_t cpu_lock;
61 #endif
62
63 /* If set, all blocks read will be copied to the specified directory. */
64 char *vn_dumpdir = NULL;
65
66 struct utsname utsname = {
67 "userland", "libzpool", "1", "1", "na"
68 };
69
70 /* this only exists to have its address taken */
71 struct proc p0;
72
73 /*
74 * =========================================================================
75 * threads
76 * =========================================================================
77 */
78 /*ARGSUSED*/
79 kthread_t *
zk_thread_create(void (* func)(),void * arg)80 zk_thread_create(void (*func)(), void *arg)
81 {
82 thread_t tid;
83
84 VERIFY(thr_create(0, 0, (void *(*)(void *))func, arg, THR_DETACHED,
85 &tid) == 0);
86
87 return ((void *)(uintptr_t)tid);
88 }
89
90 /*
91 * =========================================================================
92 * kstats
93 * =========================================================================
94 */
95 /*ARGSUSED*/
96 kstat_t *
kstat_create(char * module,int instance,char * name,char * class,uchar_t type,ulong_t ndata,uchar_t ks_flag)97 kstat_create(char *module, int instance, char *name, char *class,
98 uchar_t type, ulong_t ndata, uchar_t ks_flag)
99 {
100 return (NULL);
101 }
102
103 /*ARGSUSED*/
104 void
kstat_named_init(kstat_named_t * knp,const char * name,uchar_t type)105 kstat_named_init(kstat_named_t *knp, const char *name, uchar_t type)
106 {}
107
108 /*ARGSUSED*/
109 void
kstat_install(kstat_t * ksp)110 kstat_install(kstat_t *ksp)
111 {}
112
113 /*ARGSUSED*/
114 void
kstat_delete(kstat_t * ksp)115 kstat_delete(kstat_t *ksp)
116 {}
117
118 /*
119 * =========================================================================
120 * mutexes
121 * =========================================================================
122 */
123 void
zmutex_init(kmutex_t * mp)124 zmutex_init(kmutex_t *mp)
125 {
126 mp->m_owner = NULL;
127 mp->initialized = B_TRUE;
128 (void) _mutex_init(&mp->m_lock, USYNC_THREAD, NULL);
129 }
130
131 void
zmutex_destroy(kmutex_t * mp)132 zmutex_destroy(kmutex_t *mp)
133 {
134 ASSERT(mp->initialized == B_TRUE);
135 ASSERT(mp->m_owner == NULL);
136 (void) _mutex_destroy(&(mp)->m_lock);
137 mp->m_owner = (void *)-1UL;
138 mp->initialized = B_FALSE;
139 }
140
141 int
zmutex_owned(kmutex_t * mp)142 zmutex_owned(kmutex_t *mp)
143 {
144 ASSERT(mp->initialized == B_TRUE);
145
146 return (mp->m_owner == curthread);
147 }
148
149 void
mutex_enter(kmutex_t * mp)150 mutex_enter(kmutex_t *mp)
151 {
152 ASSERT(mp->initialized == B_TRUE);
153 ASSERT(mp->m_owner != (void *)-1UL);
154 ASSERT(mp->m_owner != curthread);
155 VERIFY(mutex_lock(&mp->m_lock) == 0);
156 ASSERT(mp->m_owner == NULL);
157 mp->m_owner = curthread;
158 }
159
160 int
mutex_tryenter(kmutex_t * mp)161 mutex_tryenter(kmutex_t *mp)
162 {
163 ASSERT(mp->initialized == B_TRUE);
164 ASSERT(mp->m_owner != (void *)-1UL);
165 if (0 == mutex_trylock(&mp->m_lock)) {
166 ASSERT(mp->m_owner == NULL);
167 mp->m_owner = curthread;
168 return (1);
169 } else {
170 return (0);
171 }
172 }
173
174 void
mutex_exit(kmutex_t * mp)175 mutex_exit(kmutex_t *mp)
176 {
177 ASSERT(mp->initialized == B_TRUE);
178 ASSERT(mutex_owner(mp) == curthread);
179 mp->m_owner = NULL;
180 VERIFY(mutex_unlock(&mp->m_lock) == 0);
181 }
182
183 void *
mutex_owner(kmutex_t * mp)184 mutex_owner(kmutex_t *mp)
185 {
186 ASSERT(mp->initialized == B_TRUE);
187 return (mp->m_owner);
188 }
189
190 /*
191 * =========================================================================
192 * rwlocks
193 * =========================================================================
194 */
195 /*ARGSUSED*/
196 void
rw_init(krwlock_t * rwlp,char * name,int type,void * arg)197 rw_init(krwlock_t *rwlp, char *name, int type, void *arg)
198 {
199 rwlock_init(&rwlp->rw_lock, USYNC_THREAD, NULL);
200 rwlp->rw_owner = NULL;
201 rwlp->initialized = B_TRUE;
202 rwlp->rw_count = 0;
203 }
204
205 void
rw_destroy(krwlock_t * rwlp)206 rw_destroy(krwlock_t *rwlp)
207 {
208 ASSERT(rwlp->rw_count == 0);
209 rwlock_destroy(&rwlp->rw_lock);
210 rwlp->rw_owner = (void *)-1UL;
211 rwlp->initialized = B_FALSE;
212 }
213
214 void
rw_enter(krwlock_t * rwlp,krw_t rw)215 rw_enter(krwlock_t *rwlp, krw_t rw)
216 {
217 //ASSERT(!RW_LOCK_HELD(rwlp));
218 ASSERT(rwlp->initialized == B_TRUE);
219 ASSERT(rwlp->rw_owner != (void *)-1UL);
220 ASSERT(rwlp->rw_owner != curthread);
221
222 if (rw == RW_READER) {
223 VERIFY(rw_rdlock(&rwlp->rw_lock) == 0);
224 ASSERT(rwlp->rw_count >= 0);
225 atomic_add_int(&rwlp->rw_count, 1);
226 } else {
227 VERIFY(rw_wrlock(&rwlp->rw_lock) == 0);
228 ASSERT(rwlp->rw_count == 0);
229 rwlp->rw_count = -1;
230 rwlp->rw_owner = curthread;
231 }
232 }
233
234 void
rw_exit(krwlock_t * rwlp)235 rw_exit(krwlock_t *rwlp)
236 {
237 ASSERT(rwlp->initialized == B_TRUE);
238 ASSERT(rwlp->rw_owner != (void *)-1UL);
239
240 if (rwlp->rw_owner == curthread) {
241 /* Write locked. */
242 ASSERT(rwlp->rw_count == -1);
243 rwlp->rw_count = 0;
244 rwlp->rw_owner = NULL;
245 } else {
246 /* Read locked. */
247 ASSERT(rwlp->rw_count > 0);
248 atomic_add_int(&rwlp->rw_count, -1);
249 }
250 VERIFY(rw_unlock(&rwlp->rw_lock) == 0);
251 }
252
253 int
rw_tryenter(krwlock_t * rwlp,krw_t rw)254 rw_tryenter(krwlock_t *rwlp, krw_t rw)
255 {
256 int rv;
257
258 ASSERT(rwlp->initialized == B_TRUE);
259 ASSERT(rwlp->rw_owner != (void *)-1UL);
260 ASSERT(rwlp->rw_owner != curthread);
261
262 if (rw == RW_READER)
263 rv = rw_tryrdlock(&rwlp->rw_lock);
264 else
265 rv = rw_trywrlock(&rwlp->rw_lock);
266
267 if (rv == 0) {
268 ASSERT(rwlp->rw_owner == NULL);
269 if (rw == RW_READER) {
270 ASSERT(rwlp->rw_count >= 0);
271 atomic_add_int(&rwlp->rw_count, 1);
272 } else {
273 ASSERT(rwlp->rw_count == 0);
274 rwlp->rw_count = -1;
275 rwlp->rw_owner = curthread;
276 }
277 return (1);
278 }
279
280 return (0);
281 }
282
283 /*ARGSUSED*/
284 int
rw_tryupgrade(krwlock_t * rwlp)285 rw_tryupgrade(krwlock_t *rwlp)
286 {
287 ASSERT(rwlp->initialized == B_TRUE);
288 ASSERT(rwlp->rw_owner != (void *)-1UL);
289
290 return (0);
291 }
292
293 int
rw_lock_held(krwlock_t * rwlp)294 rw_lock_held(krwlock_t *rwlp)
295 {
296
297 return (rwlp->rw_count != 0);
298 }
299
300 /*
301 * =========================================================================
302 * condition variables
303 * =========================================================================
304 */
305 /*ARGSUSED*/
306 void
cv_init(kcondvar_t * cv,char * name,int type,void * arg)307 cv_init(kcondvar_t *cv, char *name, int type, void *arg)
308 {
309 VERIFY(cond_init(cv, name, NULL) == 0);
310 }
311
312 void
cv_destroy(kcondvar_t * cv)313 cv_destroy(kcondvar_t *cv)
314 {
315 VERIFY(cond_destroy(cv) == 0);
316 }
317
318 void
cv_wait(kcondvar_t * cv,kmutex_t * mp)319 cv_wait(kcondvar_t *cv, kmutex_t *mp)
320 {
321 ASSERT(mutex_owner(mp) == curthread);
322 mp->m_owner = NULL;
323 int ret = cond_wait(cv, &mp->m_lock);
324 VERIFY(ret == 0 || ret == EINTR);
325 mp->m_owner = curthread;
326 }
327
328 clock_t
cv_timedwait(kcondvar_t * cv,kmutex_t * mp,clock_t abstime)329 cv_timedwait(kcondvar_t *cv, kmutex_t *mp, clock_t abstime)
330 {
331 int error;
332 struct timespec ts;
333 struct timeval tv;
334 clock_t delta;
335
336 abstime += ddi_get_lbolt();
337 top:
338 delta = abstime - ddi_get_lbolt();
339 if (delta <= 0)
340 return (-1);
341
342 if (gettimeofday(&tv, NULL) != 0)
343 assert(!"gettimeofday() failed");
344
345 ts.tv_sec = tv.tv_sec + delta / hz;
346 ts.tv_nsec = tv.tv_usec * 1000 + (delta % hz) * (NANOSEC / hz);
347 ASSERT(ts.tv_nsec >= 0);
348
349 if (ts.tv_nsec >= NANOSEC) {
350 ts.tv_sec++;
351 ts.tv_nsec -= NANOSEC;
352 }
353
354 ASSERT(mutex_owner(mp) == curthread);
355 mp->m_owner = NULL;
356 error = pthread_cond_timedwait(cv, &mp->m_lock, &ts);
357 mp->m_owner = curthread;
358
359 if (error == EINTR)
360 goto top;
361
362 if (error == ETIMEDOUT)
363 return (-1);
364
365 ASSERT(error == 0);
366
367 return (1);
368 }
369
370 /*ARGSUSED*/
371 clock_t
cv_timedwait_hires(kcondvar_t * cv,kmutex_t * mp,hrtime_t tim,hrtime_t res,int flag)372 cv_timedwait_hires(kcondvar_t *cv, kmutex_t *mp, hrtime_t tim, hrtime_t res,
373 int flag)
374 {
375 int error;
376 timestruc_t ts;
377 hrtime_t delta;
378
379 ASSERT(flag == 0 || flag == CALLOUT_FLAG_ABSOLUTE);
380
381 top:
382 delta = tim;
383 if (flag & CALLOUT_FLAG_ABSOLUTE)
384 delta -= gethrtime();
385
386 if (delta <= 0)
387 return (-1);
388
389 ts.tv_sec = delta / NANOSEC;
390 ts.tv_nsec = delta % NANOSEC;
391
392 ASSERT(mutex_owner(mp) == curthread);
393 mp->m_owner = NULL;
394 error = pthread_cond_timedwait(cv, &mp->m_lock, &ts);
395 mp->m_owner = curthread;
396
397 if (error == ETIMEDOUT)
398 return (-1);
399
400 if (error == EINTR)
401 goto top;
402
403 ASSERT(error == 0);
404
405 return (1);
406 }
407
408 void
cv_signal(kcondvar_t * cv)409 cv_signal(kcondvar_t *cv)
410 {
411 VERIFY(cond_signal(cv) == 0);
412 }
413
414 void
cv_broadcast(kcondvar_t * cv)415 cv_broadcast(kcondvar_t *cv)
416 {
417 VERIFY(cond_broadcast(cv) == 0);
418 }
419
420 /*
421 * =========================================================================
422 * vnode operations
423 * =========================================================================
424 */
425 /*
426 * Note: for the xxxat() versions of these functions, we assume that the
427 * starting vp is always rootdir (which is true for spa_directory.c, the only
428 * ZFS consumer of these interfaces). We assert this is true, and then emulate
429 * them by adding '/' in front of the path.
430 */
431
432 /*ARGSUSED*/
433 int
vn_open(char * path,int x1,int flags,int mode,vnode_t ** vpp,int x2,int x3)434 vn_open(char *path, int x1, int flags, int mode, vnode_t **vpp, int x2, int x3)
435 {
436 int fd;
437 int dump_fd;
438 vnode_t *vp;
439 int old_umask;
440 char realpath[MAXPATHLEN];
441 struct stat64 st;
442
443 /*
444 * If we're accessing a real disk from userland, we need to use
445 * the character interface to avoid caching. This is particularly
446 * important if we're trying to look at a real in-kernel storage
447 * pool from userland, e.g. via zdb, because otherwise we won't
448 * see the changes occurring under the segmap cache.
449 * On the other hand, the stupid character device returns zero
450 * for its size. So -- gag -- we open the block device to get
451 * its size, and remember it for subsequent VOP_GETATTR().
452 */
453 if (strncmp(path, "/dev/", 5) == 0) {
454 char *dsk;
455 #ifdef __NetBSD__
456 /*
457 * For NetBSD, we've been passed in a block device name
458 * but need to convert to the character device name.
459 * XXX a bit ugly...
460 */
461 char rawpath[MAXPATHLEN];
462
463 snprintf(rawpath, sizeof(rawpath), "/dev/r%s", path + 5);
464 path = rawpath; /* gets strdup()'d below */
465 #endif /* __NetBSD__ */
466 fd = open64(path, O_RDONLY);
467 if (fd == -1)
468 return (errno);
469 if (fstat64(fd, &st) == -1) {
470 close(fd);
471 return (errno);
472 }
473 #ifdef __NetBSD__
474 if (st.st_size == 0) {
475 off_t dsize;
476
477 if (ioctl(fd, DIOCGMEDIASIZE, &dsize) == 0)
478 st.st_size = dsize;
479 }
480 #endif /* __NetBSD__ */
481 close(fd);
482 (void) sprintf(realpath, "%s", path);
483 dsk = strstr(path, "/dsk/");
484 if (dsk != NULL)
485 (void) sprintf(realpath + (dsk - path) + 1, "r%s",
486 dsk + 1);
487 } else {
488 (void) sprintf(realpath, "%s", path);
489 if (!(flags & FCREAT) && stat64(realpath, &st) == -1)
490 return (errno);
491 }
492
493 if (flags & FCREAT)
494 old_umask = umask(0);
495
496 /*
497 * The construct 'flags - FREAD' conveniently maps combinations of
498 * FREAD and FWRITE to the corresponding O_RDONLY, O_WRONLY, and O_RDWR.
499 */
500 fd = open64(realpath, flags - FREAD, mode);
501
502 if (flags & FCREAT)
503 (void) umask(old_umask);
504
505 if (vn_dumpdir != NULL) {
506 char dumppath[MAXPATHLEN];
507 (void) snprintf(dumppath, sizeof (dumppath),
508 "%s/%s", vn_dumpdir, basename(realpath));
509 dump_fd = open64(dumppath, O_CREAT | O_WRONLY, 0666);
510 if (dump_fd == -1)
511 return (errno);
512 } else {
513 dump_fd = -1;
514 }
515
516 if (fd == -1)
517 return (errno);
518
519 if (fstat64(fd, &st) == -1) {
520 close(fd);
521 return (errno);
522 }
523
524 (void) fcntl(fd, F_SETFD, FD_CLOEXEC);
525
526 *vpp = vp = umem_zalloc(sizeof (vnode_t), UMEM_NOFAIL);
527
528 vp->v_fd = fd;
529 vp->v_size = st.st_size;
530 vp->v_path = spa_strdup(path);
531 vp->v_dump_fd = dump_fd;
532
533 return (0);
534 }
535
536 /*ARGSUSED*/
537 int
vn_openat(char * path,int x1,int flags,int mode,vnode_t ** vpp,int x2,int x3,vnode_t * startvp,int fd)538 vn_openat(char *path, int x1, int flags, int mode, vnode_t **vpp, int x2,
539 int x3, vnode_t *startvp, int fd)
540 {
541 char *realpath = umem_alloc(strlen(path) + 2, UMEM_NOFAIL);
542 int ret;
543
544 ASSERT(startvp == rootdir);
545 (void) sprintf(realpath, "/%s", path);
546
547 /* fd ignored for now, need if want to simulate nbmand support */
548 ret = vn_open(realpath, x1, flags, mode, vpp, x2, x3);
549
550 umem_free(realpath, strlen(path) + 2);
551
552 return (ret);
553 }
554
555 /*ARGSUSED*/
556 int
vn_rdwr(int uio,vnode_t * vp,void * addr,ssize_t len,offset_t offset,int x1,int x2,rlim64_t x3,void * x4,ssize_t * residp)557 vn_rdwr(int uio, vnode_t *vp, void *addr, ssize_t len, offset_t offset,
558 int x1, int x2, rlim64_t x3, void *x4, ssize_t *residp)
559 {
560 ssize_t iolen, split;
561
562 if (uio == UIO_READ) {
563 iolen = pread64(vp->v_fd, addr, len, offset);
564 if (vp->v_dump_fd != -1) {
565 int status =
566 pwrite64(vp->v_dump_fd, addr, iolen, offset);
567 ASSERT(status != -1);
568 }
569 } else {
570 /*
571 * To simulate partial disk writes, we split writes into two
572 * system calls so that the process can be killed in between.
573 */
574 int sectors = len >> SPA_MINBLOCKSHIFT;
575 split = (sectors > 0 ? rand() % sectors : 0) <<
576 SPA_MINBLOCKSHIFT;
577 iolen = pwrite64(vp->v_fd, addr, split, offset);
578 iolen += pwrite64(vp->v_fd, (char *)addr + split,
579 len - split, offset + split);
580 }
581
582 if (iolen == -1)
583 return (errno);
584 if (residp)
585 *residp = len - iolen;
586 else if (iolen != len)
587 return (EIO);
588 return (0);
589 }
590
591 void
vn_close(vnode_t * vp,int openflag,cred_t * cr,kthread_t * td)592 vn_close(vnode_t *vp, int openflag, cred_t *cr, kthread_t *td)
593 {
594 close(vp->v_fd);
595 if (vp->v_dump_fd != -1)
596 close(vp->v_dump_fd);
597 spa_strfree(vp->v_path);
598 umem_free(vp, sizeof (vnode_t));
599 }
600
601 /*
602 * At a minimum we need to update the size since vdev_reopen()
603 * will no longer call vn_openat().
604 */
605 int
fop_getattr(vnode_t * vp,vattr_t * vap)606 fop_getattr(vnode_t *vp, vattr_t *vap)
607 {
608 struct stat64 st;
609
610 if (fstat64(vp->v_fd, &st) == -1) {
611 close(vp->v_fd);
612 return (errno);
613 }
614 #ifdef __NetBSD__
615 if (st.st_size == 0) {
616 off_t dsize;
617
618 if (ioctl(vp->v_fd, DIOCGMEDIASIZE, &dsize) == 0)
619 st.st_size = dsize;
620 }
621 #endif /* __NetBSD__ */
622
623 vap->va_size = st.st_size;
624 return (0);
625 }
626
627 #ifdef ZFS_DEBUG
628
629 /*
630 * =========================================================================
631 * Figure out which debugging statements to print
632 * =========================================================================
633 */
634
635 static char *dprintf_string;
636 static int dprintf_print_all;
637
638 int
dprintf_find_string(const char * string)639 dprintf_find_string(const char *string)
640 {
641 char *tmp_str = dprintf_string;
642 int len = strlen(string);
643
644 /*
645 * Find out if this is a string we want to print.
646 * String format: file1.c,function_name1,file2.c,file3.c
647 */
648
649 while (tmp_str != NULL) {
650 if (strncmp(tmp_str, string, len) == 0 &&
651 (tmp_str[len] == ',' || tmp_str[len] == '\0'))
652 return (1);
653 tmp_str = strchr(tmp_str, ',');
654 if (tmp_str != NULL)
655 tmp_str++; /* Get rid of , */
656 }
657 return (0);
658 }
659
660 void
dprintf_setup(int * argc,char ** argv)661 dprintf_setup(int *argc, char **argv)
662 {
663 int i, j;
664
665 /*
666 * Debugging can be specified two ways: by setting the
667 * environment variable ZFS_DEBUG, or by including a
668 * "debug=..." argument on the command line. The command
669 * line setting overrides the environment variable.
670 */
671
672 for (i = 1; i < *argc; i++) {
673 int len = strlen("debug=");
674 /* First look for a command line argument */
675 if (strncmp("debug=", argv[i], len) == 0) {
676 dprintf_string = argv[i] + len;
677 /* Remove from args */
678 for (j = i; j < *argc; j++)
679 argv[j] = argv[j+1];
680 argv[j] = NULL;
681 (*argc)--;
682 }
683 }
684
685 if (dprintf_string == NULL) {
686 /* Look for ZFS_DEBUG environment variable */
687 dprintf_string = getenv("ZFS_DEBUG");
688 }
689
690 /*
691 * Are we just turning on all debugging?
692 */
693 if (dprintf_find_string("on"))
694 dprintf_print_all = 1;
695
696 if (dprintf_string != NULL)
697 zfs_flags |= ZFS_DEBUG_DPRINTF;
698 }
699
700 int
sysctl_handle_64(SYSCTL_HANDLER_ARGS)701 sysctl_handle_64(SYSCTL_HANDLER_ARGS)
702 {
703 return (0);
704 }
705
706 /*
707 * =========================================================================
708 * debug printfs
709 * =========================================================================
710 */
711 void
__dprintf(const char * file,const char * func,int line,const char * fmt,...)712 __dprintf(const char *file, const char *func, int line, const char *fmt, ...)
713 {
714 const char *newfile;
715 va_list adx;
716
717 /*
718 * Get rid of annoying "../common/" prefix to filename.
719 */
720 newfile = strrchr(file, '/');
721 if (newfile != NULL) {
722 newfile = newfile + 1; /* Get rid of leading / */
723 } else {
724 newfile = file;
725 }
726
727 if (dprintf_print_all ||
728 dprintf_find_string(newfile) ||
729 dprintf_find_string(func)) {
730 /* Print out just the function name if requested */
731 flockfile(stdout);
732 if (dprintf_find_string("pid"))
733 (void) printf("%d ", getpid());
734 if (dprintf_find_string("tid"))
735 (void) printf("%lu ", thr_self());
736 #if 0
737 if (dprintf_find_string("cpu"))
738 (void) printf("%u ", getcpuid());
739 #endif
740 if (dprintf_find_string("time"))
741 (void) printf("%llu ", gethrtime());
742 if (dprintf_find_string("long"))
743 (void) printf("%s, line %d: ", newfile, line);
744 (void) printf("%s: ", func);
745 va_start(adx, fmt);
746 (void) vprintf(fmt, adx);
747 va_end(adx);
748 funlockfile(stdout);
749 }
750 }
751
752 #endif /* ZFS_DEBUG */
753
754 /*
755 * =========================================================================
756 * cmn_err() and panic()
757 * =========================================================================
758 */
759 static char ce_prefix[CE_IGNORE][10] = { "", "NOTICE: ", "WARNING: ", "" };
760 static char ce_suffix[CE_IGNORE][2] = { "", "\n", "\n", "" };
761
762 void
vpanic(const char * fmt,va_list adx)763 vpanic(const char *fmt, va_list adx)
764 {
765 (void) fprintf(stderr, "error: ");
766 (void) vfprintf(stderr, fmt, adx);
767 (void) fprintf(stderr, "\n");
768
769 abort(); /* think of it as a "user-level crash dump" */
770 }
771
772 void
panic(const char * fmt,...)773 panic(const char *fmt, ...)
774 {
775 va_list adx;
776
777 va_start(adx, fmt);
778 vpanic(fmt, adx);
779 va_end(adx);
780 }
781
782 void
vcmn_err(int ce,const char * fmt,va_list adx)783 vcmn_err(int ce, const char *fmt, va_list adx)
784 {
785 if (ce == CE_PANIC)
786 vpanic(fmt, adx);
787 if (ce != CE_NOTE) { /* suppress noise in userland stress testing */
788 (void) fprintf(stderr, "%s", ce_prefix[ce]);
789 (void) vfprintf(stderr, fmt, adx);
790 (void) fprintf(stderr, "%s", ce_suffix[ce]);
791 }
792 }
793
794 /*PRINTFLIKE2*/
795 void
cmn_err(int ce,const char * fmt,...)796 cmn_err(int ce, const char *fmt, ...)
797 {
798 va_list adx;
799
800 va_start(adx, fmt);
801 vcmn_err(ce, fmt, adx);
802 va_end(adx);
803 }
804
805 /*
806 * =========================================================================
807 * kobj interfaces
808 * =========================================================================
809 */
810 struct _buf *
kobj_open_file(char * name)811 kobj_open_file(char *name)
812 {
813 struct _buf *file;
814 vnode_t *vp;
815
816 /* set vp as the _fd field of the file */
817 if (vn_openat(name, UIO_SYSSPACE, FREAD, 0, &vp, 0, 0, rootdir,
818 -1) != 0)
819 return ((void *)-1UL);
820
821 file = umem_zalloc(sizeof (struct _buf), UMEM_NOFAIL);
822 file->_fd = (intptr_t)vp;
823 return (file);
824 }
825
826 int
kobj_read_file(struct _buf * file,char * buf,unsigned size,unsigned off)827 kobj_read_file(struct _buf *file, char *buf, unsigned size, unsigned off)
828 {
829 ssize_t resid;
830
831 vn_rdwr(UIO_READ, (vnode_t *)file->_fd, buf, size, (offset_t)off,
832 UIO_SYSSPACE, 0, 0, 0, &resid);
833
834 return (size - resid);
835 }
836
837 void
kobj_close_file(struct _buf * file)838 kobj_close_file(struct _buf *file)
839 {
840 vn_close((vnode_t *)file->_fd, 0, NULL, NULL);
841 umem_free(file, sizeof (struct _buf));
842 }
843
844 int
kobj_get_filesize(struct _buf * file,uint64_t * size)845 kobj_get_filesize(struct _buf *file, uint64_t *size)
846 {
847 struct stat64 st;
848 vnode_t *vp = (vnode_t *)file->_fd;
849
850 if (fstat64(vp->v_fd, &st) == -1) {
851 vn_close(vp, 0, NULL, NULL);
852 return (errno);
853 }
854 *size = st.st_size;
855 return (0);
856 }
857
858 /*
859 * =========================================================================
860 * misc routines
861 * =========================================================================
862 */
863
864 void
delay(clock_t ticks)865 delay(clock_t ticks)
866 {
867 poll(0, 0, ticks * (1000 / hz));
868 }
869
870 #if 0
871 /*
872 * Find highest one bit set.
873 * Returns bit number + 1 of highest bit that is set, otherwise returns 0.
874 */
875 int
876 highbit64(uint64_t i)
877 {
878 int h = 1;
879
880 if (i == 0)
881 return (0);
882 if (i & 0xffffffff00000000ULL) {
883 h += 32; i >>= 32;
884 }
885 if (i & 0xffff0000) {
886 h += 16; i >>= 16;
887 }
888 if (i & 0xff00) {
889 h += 8; i >>= 8;
890 }
891 if (i & 0xf0) {
892 h += 4; i >>= 4;
893 }
894 if (i & 0xc) {
895 h += 2; i >>= 2;
896 }
897 if (i & 0x2) {
898 h += 1;
899 }
900 return (h);
901 }
902 #endif
903
904 static int random_fd = -1, urandom_fd = -1;
905
906 static int
random_get_bytes_common(uint8_t * ptr,size_t len,int fd)907 random_get_bytes_common(uint8_t *ptr, size_t len, int fd)
908 {
909 size_t resid = len;
910 ssize_t bytes;
911
912 ASSERT(fd != -1);
913
914 while (resid != 0) {
915 bytes = read(fd, ptr, resid);
916 ASSERT3S(bytes, >=, 0);
917 ptr += bytes;
918 resid -= bytes;
919 }
920
921 return (0);
922 }
923
924 int
random_get_bytes(uint8_t * ptr,size_t len)925 random_get_bytes(uint8_t *ptr, size_t len)
926 {
927 return (random_get_bytes_common(ptr, len, random_fd));
928 }
929
930 int
random_get_pseudo_bytes(uint8_t * ptr,size_t len)931 random_get_pseudo_bytes(uint8_t *ptr, size_t len)
932 {
933 return (random_get_bytes_common(ptr, len, urandom_fd));
934 }
935
936 int
ddi_strtoul(const char * hw_serial,char ** nptr,int base,unsigned long * result)937 ddi_strtoul(const char *hw_serial, char **nptr, int base, unsigned long *result)
938 {
939 char *end;
940
941 *result = strtoul(hw_serial, &end, base);
942 if (*result == 0)
943 return (errno);
944 return (0);
945 }
946
947 int
ddi_strtoull(const char * str,char ** nptr,int base,u_longlong_t * result)948 ddi_strtoull(const char *str, char **nptr, int base, u_longlong_t *result)
949 {
950 char *end;
951
952 *result = strtoull(str, &end, base);
953 if (*result == 0)
954 return (errno);
955 return (0);
956 }
957
958 #ifndef __FreeBSD__
959 /* ARGSUSED */
960 cyclic_id_t
cyclic_add(cyc_handler_t * hdlr,cyc_time_t * when)961 cyclic_add(cyc_handler_t *hdlr, cyc_time_t *when)
962 {
963 return (1);
964 }
965
966 /* ARGSUSED */
967 void
cyclic_remove(cyclic_id_t id)968 cyclic_remove(cyclic_id_t id)
969 {
970 }
971
972 /* ARGSUSED */
973 int
cyclic_reprogram(cyclic_id_t id,hrtime_t expiration)974 cyclic_reprogram(cyclic_id_t id, hrtime_t expiration)
975 {
976 return (1);
977 }
978 #endif
979
980 /*
981 * =========================================================================
982 * kernel emulation setup & teardown
983 * =========================================================================
984 */
985 static int
umem_out_of_memory(void)986 umem_out_of_memory(void)
987 {
988 char errmsg[] = "out of memory -- generating core dump\n";
989
990 write(fileno(stderr), errmsg, sizeof (errmsg));
991 abort();
992 return (0);
993 }
994
995 void
kernel_init(int mode)996 kernel_init(int mode)
997 {
998 extern uint_t rrw_tsd_key;
999
1000 umem_nofail_callback(umem_out_of_memory);
1001
1002 physmem = sysconf(_SC_PHYS_PAGES);
1003
1004 dprintf("physmem = %llu pages (%.2f GB)\n", physmem,
1005 (double)physmem * sysconf(_SC_PAGE_SIZE) / (1ULL << 30));
1006
1007 (void) snprintf(hw_serial, sizeof (hw_serial), "%lu",
1008 (mode & FWRITE) ? (unsigned long)gethostid() : 0);
1009
1010 VERIFY((random_fd = open("/dev/random", O_RDONLY)) != -1);
1011 VERIFY((urandom_fd = open("/dev/urandom", O_RDONLY)) != -1);
1012
1013 system_taskq_init();
1014
1015 #ifdef illumos
1016 mutex_init(&cpu_lock, NULL, MUTEX_DEFAULT, NULL);
1017 #endif
1018
1019 spa_init(mode);
1020
1021 tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
1022 }
1023
1024 void
kernel_fini(void)1025 kernel_fini(void)
1026 {
1027 spa_fini();
1028
1029 system_taskq_fini();
1030
1031 close(random_fd);
1032 close(urandom_fd);
1033
1034 random_fd = -1;
1035 urandom_fd = -1;
1036 }
1037
1038 int
z_uncompress(void * dst,size_t * dstlen,const void * src,size_t srclen)1039 z_uncompress(void *dst, size_t *dstlen, const void *src, size_t srclen)
1040 {
1041 int ret;
1042 uLongf len = *dstlen;
1043
1044 if ((ret = uncompress(dst, &len, src, srclen)) == Z_OK)
1045 *dstlen = (size_t)len;
1046
1047 return (ret);
1048 }
1049
1050 int
z_compress_level(void * dst,size_t * dstlen,const void * src,size_t srclen,int level)1051 z_compress_level(void *dst, size_t *dstlen, const void *src, size_t srclen,
1052 int level)
1053 {
1054 int ret;
1055 uLongf len = *dstlen;
1056
1057 if ((ret = compress2(dst, &len, src, srclen, level)) == Z_OK)
1058 *dstlen = (size_t)len;
1059
1060 return (ret);
1061 }
1062
1063 uid_t
crgetuid(cred_t * cr)1064 crgetuid(cred_t *cr)
1065 {
1066 return (0);
1067 }
1068
1069 uid_t
crgetruid(cred_t * cr)1070 crgetruid(cred_t *cr)
1071 {
1072 return (0);
1073 }
1074
1075 gid_t
crgetgid(cred_t * cr)1076 crgetgid(cred_t *cr)
1077 {
1078 return (0);
1079 }
1080
1081 int
crgetngroups(cred_t * cr)1082 crgetngroups(cred_t *cr)
1083 {
1084 return (0);
1085 }
1086
1087 gid_t *
crgetgroups(cred_t * cr)1088 crgetgroups(cred_t *cr)
1089 {
1090 return (NULL);
1091 }
1092
1093 int
zfs_secpolicy_snapshot_perms(const char * name,cred_t * cr)1094 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
1095 {
1096 return (0);
1097 }
1098
1099 int
zfs_secpolicy_rename_perms(const char * from,const char * to,cred_t * cr)1100 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
1101 {
1102 return (0);
1103 }
1104
1105 int
zfs_secpolicy_destroy_perms(const char * name,cred_t * cr)1106 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
1107 {
1108 return (0);
1109 }
1110
1111 ksiddomain_t *
ksid_lookupdomain(const char * dom)1112 ksid_lookupdomain(const char *dom)
1113 {
1114 ksiddomain_t *kd;
1115
1116 kd = umem_zalloc(sizeof (ksiddomain_t), UMEM_NOFAIL);
1117 kd->kd_name = spa_strdup(dom);
1118 return (kd);
1119 }
1120
1121 void
ksiddomain_rele(ksiddomain_t * ksid)1122 ksiddomain_rele(ksiddomain_t *ksid)
1123 {
1124 spa_strfree(ksid->kd_name);
1125 umem_free(ksid, sizeof (ksiddomain_t));
1126 }
1127
1128 /*
1129 * Do not change the length of the returned string; it must be freed
1130 * with strfree().
1131 */
1132 char *
kmem_asprintf(const char * fmt,...)1133 kmem_asprintf(const char *fmt, ...)
1134 {
1135 int size;
1136 va_list adx;
1137 char *buf;
1138
1139 va_start(adx, fmt);
1140 size = vsnprintf(NULL, 0, fmt, adx) + 1;
1141 va_end(adx);
1142
1143 buf = kmem_alloc(size, KM_SLEEP);
1144
1145 va_start(adx, fmt);
1146 size = vsnprintf(buf, size, fmt, adx);
1147 va_end(adx);
1148
1149 return (buf);
1150 }
1151
1152 /* ARGSUSED */
1153 int
zfs_onexit_fd_hold(int fd,minor_t * minorp)1154 zfs_onexit_fd_hold(int fd, minor_t *minorp)
1155 {
1156 *minorp = 0;
1157 return (0);
1158 }
1159
1160 /* ARGSUSED */
1161 void
zfs_onexit_fd_rele(int fd)1162 zfs_onexit_fd_rele(int fd)
1163 {
1164 }
1165
1166 /* ARGSUSED */
1167 int
zfs_onexit_add_cb(minor_t minor,void (* func)(void *),void * data,uint64_t * action_handle)1168 zfs_onexit_add_cb(minor_t minor, void (*func)(void *), void *data,
1169 uint64_t *action_handle)
1170 {
1171 return (0);
1172 }
1173
1174 /* ARGSUSED */
1175 int
zfs_onexit_del_cb(minor_t minor,uint64_t action_handle,boolean_t fire)1176 zfs_onexit_del_cb(minor_t minor, uint64_t action_handle, boolean_t fire)
1177 {
1178 return (0);
1179 }
1180
1181 /* ARGSUSED */
1182 int
zfs_onexit_cb_data(minor_t minor,uint64_t action_handle,void ** data)1183 zfs_onexit_cb_data(minor_t minor, uint64_t action_handle, void **data)
1184 {
1185 return (0);
1186 }
1187
1188 #ifdef __FreeBSD__
1189 /* ARGSUSED */
1190 int
zvol_create_minors(const char * name)1191 zvol_create_minors(const char *name)
1192 {
1193 return (0);
1194 }
1195 #endif
1196
1197 #ifdef illumos
1198 void
bioinit(buf_t * bp)1199 bioinit(buf_t *bp)
1200 {
1201 bzero(bp, sizeof (buf_t));
1202 }
1203
1204 void
biodone(buf_t * bp)1205 biodone(buf_t *bp)
1206 {
1207 if (bp->b_iodone != NULL) {
1208 (*(bp->b_iodone))(bp);
1209 return;
1210 }
1211 ASSERT((bp->b_flags & B_DONE) == 0);
1212 bp->b_flags |= B_DONE;
1213 }
1214
1215 void
bioerror(buf_t * bp,int error)1216 bioerror(buf_t *bp, int error)
1217 {
1218 ASSERT(bp != NULL);
1219 ASSERT(error >= 0);
1220
1221 if (error != 0) {
1222 bp->b_flags |= B_ERROR;
1223 } else {
1224 bp->b_flags &= ~B_ERROR;
1225 }
1226 bp->b_error = error;
1227 }
1228
1229
1230 int
geterror(struct buf * bp)1231 geterror(struct buf *bp)
1232 {
1233 int error = 0;
1234
1235 if (bp->b_flags & B_ERROR) {
1236 error = bp->b_error;
1237 if (!error)
1238 error = EIO;
1239 }
1240 return (error);
1241 }
1242 #endif
1243