xref: /freebsd/sys/contrib/openzfs/lib/libzpool/kernel.c (revision 13ec1e3155c7e9bf037b12af186351b7fa9b9450)
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, 2018 by Delphix. All rights reserved.
24  * Copyright (c) 2016 Actifio, Inc. All rights reserved.
25  */
26 
27 #include <assert.h>
28 #include <fcntl.h>
29 #include <libgen.h>
30 #include <poll.h>
31 #include <stdio.h>
32 #include <stdlib.h>
33 #include <string.h>
34 #include <libzutil.h>
35 #include <sys/crypto/icp.h>
36 #include <sys/processor.h>
37 #include <sys/rrwlock.h>
38 #include <sys/spa.h>
39 #include <sys/stat.h>
40 #include <sys/systeminfo.h>
41 #include <sys/time.h>
42 #include <sys/utsname.h>
43 #include <sys/zfs_context.h>
44 #include <sys/zfs_onexit.h>
45 #include <sys/zfs_vfsops.h>
46 #include <sys/zstd/zstd.h>
47 #include <sys/zvol.h>
48 #include <zfs_fletcher.h>
49 #include <zlib.h>
50 
51 /*
52  * Emulation of kernel services in userland.
53  */
54 
55 uint64_t physmem;
56 char hw_serial[HW_HOSTID_LEN];
57 struct utsname hw_utsname;
58 
59 /* If set, all blocks read will be copied to the specified directory. */
60 char *vn_dumpdir = NULL;
61 
62 /* this only exists to have its address taken */
63 struct proc p0;
64 
65 /*
66  * =========================================================================
67  * threads
68  * =========================================================================
69  *
70  * TS_STACK_MIN is dictated by the minimum allowed pthread stack size.  While
71  * TS_STACK_MAX is somewhat arbitrary, it was selected to be large enough for
72  * the expected stack depth while small enough to avoid exhausting address
73  * space with high thread counts.
74  */
75 #define	TS_STACK_MIN	MAX(PTHREAD_STACK_MIN, 32768)
76 #define	TS_STACK_MAX	(256 * 1024)
77 
78 kthread_t *
79 zk_thread_create(void (*func)(void *), void *arg, size_t stksize, int state)
80 {
81 	pthread_attr_t attr;
82 	pthread_t tid;
83 	char *stkstr;
84 	int detachstate = PTHREAD_CREATE_DETACHED;
85 
86 	VERIFY0(pthread_attr_init(&attr));
87 
88 	if (state & TS_JOINABLE)
89 		detachstate = PTHREAD_CREATE_JOINABLE;
90 
91 	VERIFY0(pthread_attr_setdetachstate(&attr, detachstate));
92 
93 	/*
94 	 * We allow the default stack size in user space to be specified by
95 	 * setting the ZFS_STACK_SIZE environment variable.  This allows us
96 	 * the convenience of observing and debugging stack overruns in
97 	 * user space.  Explicitly specified stack sizes will be honored.
98 	 * The usage of ZFS_STACK_SIZE is discussed further in the
99 	 * ENVIRONMENT VARIABLES sections of the ztest(1) man page.
100 	 */
101 	if (stksize == 0) {
102 		stkstr = getenv("ZFS_STACK_SIZE");
103 
104 		if (stkstr == NULL)
105 			stksize = TS_STACK_MAX;
106 		else
107 			stksize = MAX(atoi(stkstr), TS_STACK_MIN);
108 	}
109 
110 	VERIFY3S(stksize, >, 0);
111 	stksize = P2ROUNDUP(MAX(stksize, TS_STACK_MIN), PAGESIZE);
112 
113 	/*
114 	 * If this ever fails, it may be because the stack size is not a
115 	 * multiple of system page size.
116 	 */
117 	VERIFY0(pthread_attr_setstacksize(&attr, stksize));
118 	VERIFY0(pthread_attr_setguardsize(&attr, PAGESIZE));
119 
120 	VERIFY0(pthread_create(&tid, &attr, (void *(*)(void *))func, arg));
121 	VERIFY0(pthread_attr_destroy(&attr));
122 
123 	return ((void *)(uintptr_t)tid);
124 }
125 
126 /*
127  * =========================================================================
128  * kstats
129  * =========================================================================
130  */
131 kstat_t *
132 kstat_create(const char *module, int instance, const char *name,
133     const char *class, uchar_t type, ulong_t ndata, uchar_t ks_flag)
134 {
135 	(void) module, (void) instance, (void) name, (void) class, (void) type,
136 	    (void) ndata, (void) ks_flag;
137 	return (NULL);
138 }
139 
140 void
141 kstat_install(kstat_t *ksp)
142 {
143 	(void) ksp;
144 }
145 
146 void
147 kstat_delete(kstat_t *ksp)
148 {
149 	(void) ksp;
150 }
151 
152 void
153 kstat_set_raw_ops(kstat_t *ksp,
154     int (*headers)(char *buf, size_t size),
155     int (*data)(char *buf, size_t size, void *data),
156     void *(*addr)(kstat_t *ksp, loff_t index))
157 {
158 	(void) ksp, (void) headers, (void) data, (void) addr;
159 }
160 
161 /*
162  * =========================================================================
163  * mutexes
164  * =========================================================================
165  */
166 
167 void
168 mutex_init(kmutex_t *mp, char *name, int type, void *cookie)
169 {
170 	(void) name, (void) type, (void) cookie;
171 	VERIFY0(pthread_mutex_init(&mp->m_lock, NULL));
172 	memset(&mp->m_owner, 0, sizeof (pthread_t));
173 }
174 
175 void
176 mutex_destroy(kmutex_t *mp)
177 {
178 	VERIFY0(pthread_mutex_destroy(&mp->m_lock));
179 }
180 
181 void
182 mutex_enter(kmutex_t *mp)
183 {
184 	VERIFY0(pthread_mutex_lock(&mp->m_lock));
185 	mp->m_owner = pthread_self();
186 }
187 
188 int
189 mutex_tryenter(kmutex_t *mp)
190 {
191 	int error = pthread_mutex_trylock(&mp->m_lock);
192 	if (error == 0) {
193 		mp->m_owner = pthread_self();
194 		return (1);
195 	} else {
196 		VERIFY3S(error, ==, EBUSY);
197 		return (0);
198 	}
199 }
200 
201 void
202 mutex_exit(kmutex_t *mp)
203 {
204 	memset(&mp->m_owner, 0, sizeof (pthread_t));
205 	VERIFY0(pthread_mutex_unlock(&mp->m_lock));
206 }
207 
208 /*
209  * =========================================================================
210  * rwlocks
211  * =========================================================================
212  */
213 
214 void
215 rw_init(krwlock_t *rwlp, char *name, int type, void *arg)
216 {
217 	(void) name, (void) type, (void) arg;
218 	VERIFY0(pthread_rwlock_init(&rwlp->rw_lock, NULL));
219 	rwlp->rw_readers = 0;
220 	rwlp->rw_owner = 0;
221 }
222 
223 void
224 rw_destroy(krwlock_t *rwlp)
225 {
226 	VERIFY0(pthread_rwlock_destroy(&rwlp->rw_lock));
227 }
228 
229 void
230 rw_enter(krwlock_t *rwlp, krw_t rw)
231 {
232 	if (rw == RW_READER) {
233 		VERIFY0(pthread_rwlock_rdlock(&rwlp->rw_lock));
234 		atomic_inc_uint(&rwlp->rw_readers);
235 	} else {
236 		VERIFY0(pthread_rwlock_wrlock(&rwlp->rw_lock));
237 		rwlp->rw_owner = pthread_self();
238 	}
239 }
240 
241 void
242 rw_exit(krwlock_t *rwlp)
243 {
244 	if (RW_READ_HELD(rwlp))
245 		atomic_dec_uint(&rwlp->rw_readers);
246 	else
247 		rwlp->rw_owner = 0;
248 
249 	VERIFY0(pthread_rwlock_unlock(&rwlp->rw_lock));
250 }
251 
252 int
253 rw_tryenter(krwlock_t *rwlp, krw_t rw)
254 {
255 	int error;
256 
257 	if (rw == RW_READER)
258 		error = pthread_rwlock_tryrdlock(&rwlp->rw_lock);
259 	else
260 		error = pthread_rwlock_trywrlock(&rwlp->rw_lock);
261 
262 	if (error == 0) {
263 		if (rw == RW_READER)
264 			atomic_inc_uint(&rwlp->rw_readers);
265 		else
266 			rwlp->rw_owner = pthread_self();
267 
268 		return (1);
269 	}
270 
271 	VERIFY3S(error, ==, EBUSY);
272 
273 	return (0);
274 }
275 
276 uint32_t
277 zone_get_hostid(void *zonep)
278 {
279 	/*
280 	 * We're emulating the system's hostid in userland.
281 	 */
282 	(void) zonep;
283 	return (strtoul(hw_serial, NULL, 10));
284 }
285 
286 int
287 rw_tryupgrade(krwlock_t *rwlp)
288 {
289 	(void) rwlp;
290 	return (0);
291 }
292 
293 /*
294  * =========================================================================
295  * condition variables
296  * =========================================================================
297  */
298 
299 void
300 cv_init(kcondvar_t *cv, char *name, int type, void *arg)
301 {
302 	(void) name, (void) type, (void) arg;
303 	VERIFY0(pthread_cond_init(cv, NULL));
304 }
305 
306 void
307 cv_destroy(kcondvar_t *cv)
308 {
309 	VERIFY0(pthread_cond_destroy(cv));
310 }
311 
312 void
313 cv_wait(kcondvar_t *cv, kmutex_t *mp)
314 {
315 	memset(&mp->m_owner, 0, sizeof (pthread_t));
316 	VERIFY0(pthread_cond_wait(cv, &mp->m_lock));
317 	mp->m_owner = pthread_self();
318 }
319 
320 int
321 cv_wait_sig(kcondvar_t *cv, kmutex_t *mp)
322 {
323 	cv_wait(cv, mp);
324 	return (1);
325 }
326 
327 int
328 cv_timedwait(kcondvar_t *cv, kmutex_t *mp, clock_t abstime)
329 {
330 	int error;
331 	struct timeval tv;
332 	struct timespec ts;
333 	clock_t delta;
334 
335 	delta = abstime - ddi_get_lbolt();
336 	if (delta <= 0)
337 		return (-1);
338 
339 	VERIFY(gettimeofday(&tv, NULL) == 0);
340 
341 	ts.tv_sec = tv.tv_sec + delta / hz;
342 	ts.tv_nsec = tv.tv_usec * NSEC_PER_USEC + (delta % hz) * (NANOSEC / hz);
343 	if (ts.tv_nsec >= NANOSEC) {
344 		ts.tv_sec++;
345 		ts.tv_nsec -= NANOSEC;
346 	}
347 
348 	memset(&mp->m_owner, 0, sizeof (pthread_t));
349 	error = pthread_cond_timedwait(cv, &mp->m_lock, &ts);
350 	mp->m_owner = pthread_self();
351 
352 	if (error == ETIMEDOUT)
353 		return (-1);
354 
355 	VERIFY0(error);
356 
357 	return (1);
358 }
359 
360 int
361 cv_timedwait_hires(kcondvar_t *cv, kmutex_t *mp, hrtime_t tim, hrtime_t res,
362     int flag)
363 {
364 	(void) res;
365 	int error;
366 	struct timeval tv;
367 	struct timespec ts;
368 	hrtime_t delta;
369 
370 	ASSERT(flag == 0 || flag == CALLOUT_FLAG_ABSOLUTE);
371 
372 	delta = tim;
373 	if (flag & CALLOUT_FLAG_ABSOLUTE)
374 		delta -= gethrtime();
375 
376 	if (delta <= 0)
377 		return (-1);
378 
379 	VERIFY0(gettimeofday(&tv, NULL));
380 
381 	ts.tv_sec = tv.tv_sec + delta / NANOSEC;
382 	ts.tv_nsec = tv.tv_usec * NSEC_PER_USEC + (delta % NANOSEC);
383 	if (ts.tv_nsec >= NANOSEC) {
384 		ts.tv_sec++;
385 		ts.tv_nsec -= NANOSEC;
386 	}
387 
388 	memset(&mp->m_owner, 0, sizeof (pthread_t));
389 	error = pthread_cond_timedwait(cv, &mp->m_lock, &ts);
390 	mp->m_owner = pthread_self();
391 
392 	if (error == ETIMEDOUT)
393 		return (-1);
394 
395 	VERIFY0(error);
396 
397 	return (1);
398 }
399 
400 void
401 cv_signal(kcondvar_t *cv)
402 {
403 	VERIFY0(pthread_cond_signal(cv));
404 }
405 
406 void
407 cv_broadcast(kcondvar_t *cv)
408 {
409 	VERIFY0(pthread_cond_broadcast(cv));
410 }
411 
412 /*
413  * =========================================================================
414  * procfs list
415  * =========================================================================
416  */
417 
418 void
419 seq_printf(struct seq_file *m, const char *fmt, ...)
420 {
421 	(void) m, (void) fmt;
422 }
423 
424 void
425 procfs_list_install(const char *module,
426     const char *submodule,
427     const char *name,
428     mode_t mode,
429     procfs_list_t *procfs_list,
430     int (*show)(struct seq_file *f, void *p),
431     int (*show_header)(struct seq_file *f),
432     int (*clear)(procfs_list_t *procfs_list),
433     size_t procfs_list_node_off)
434 {
435 	(void) module, (void) submodule, (void) name, (void) mode, (void) show,
436 	    (void) show_header, (void) clear;
437 	mutex_init(&procfs_list->pl_lock, NULL, MUTEX_DEFAULT, NULL);
438 	list_create(&procfs_list->pl_list,
439 	    procfs_list_node_off + sizeof (procfs_list_node_t),
440 	    procfs_list_node_off + offsetof(procfs_list_node_t, pln_link));
441 	procfs_list->pl_next_id = 1;
442 	procfs_list->pl_node_offset = procfs_list_node_off;
443 }
444 
445 void
446 procfs_list_uninstall(procfs_list_t *procfs_list)
447 {
448 	(void) procfs_list;
449 }
450 
451 void
452 procfs_list_destroy(procfs_list_t *procfs_list)
453 {
454 	ASSERT(list_is_empty(&procfs_list->pl_list));
455 	list_destroy(&procfs_list->pl_list);
456 	mutex_destroy(&procfs_list->pl_lock);
457 }
458 
459 #define	NODE_ID(procfs_list, obj) \
460 		(((procfs_list_node_t *)(((char *)obj) + \
461 		(procfs_list)->pl_node_offset))->pln_id)
462 
463 void
464 procfs_list_add(procfs_list_t *procfs_list, void *p)
465 {
466 	ASSERT(MUTEX_HELD(&procfs_list->pl_lock));
467 	NODE_ID(procfs_list, p) = procfs_list->pl_next_id++;
468 	list_insert_tail(&procfs_list->pl_list, p);
469 }
470 
471 /*
472  * =========================================================================
473  * vnode operations
474  * =========================================================================
475  */
476 
477 /*
478  * =========================================================================
479  * Figure out which debugging statements to print
480  * =========================================================================
481  */
482 
483 static char *dprintf_string;
484 static int dprintf_print_all;
485 
486 int
487 dprintf_find_string(const char *string)
488 {
489 	char *tmp_str = dprintf_string;
490 	int len = strlen(string);
491 
492 	/*
493 	 * Find out if this is a string we want to print.
494 	 * String format: file1.c,function_name1,file2.c,file3.c
495 	 */
496 
497 	while (tmp_str != NULL) {
498 		if (strncmp(tmp_str, string, len) == 0 &&
499 		    (tmp_str[len] == ',' || tmp_str[len] == '\0'))
500 			return (1);
501 		tmp_str = strchr(tmp_str, ',');
502 		if (tmp_str != NULL)
503 			tmp_str++; /* Get rid of , */
504 	}
505 	return (0);
506 }
507 
508 void
509 dprintf_setup(int *argc, char **argv)
510 {
511 	int i, j;
512 
513 	/*
514 	 * Debugging can be specified two ways: by setting the
515 	 * environment variable ZFS_DEBUG, or by including a
516 	 * "debug=..."  argument on the command line.  The command
517 	 * line setting overrides the environment variable.
518 	 */
519 
520 	for (i = 1; i < *argc; i++) {
521 		int len = strlen("debug=");
522 		/* First look for a command line argument */
523 		if (strncmp("debug=", argv[i], len) == 0) {
524 			dprintf_string = argv[i] + len;
525 			/* Remove from args */
526 			for (j = i; j < *argc; j++)
527 				argv[j] = argv[j+1];
528 			argv[j] = NULL;
529 			(*argc)--;
530 		}
531 	}
532 
533 	if (dprintf_string == NULL) {
534 		/* Look for ZFS_DEBUG environment variable */
535 		dprintf_string = getenv("ZFS_DEBUG");
536 	}
537 
538 	/*
539 	 * Are we just turning on all debugging?
540 	 */
541 	if (dprintf_find_string("on"))
542 		dprintf_print_all = 1;
543 
544 	if (dprintf_string != NULL)
545 		zfs_flags |= ZFS_DEBUG_DPRINTF;
546 }
547 
548 /*
549  * =========================================================================
550  * debug printfs
551  * =========================================================================
552  */
553 void
554 __dprintf(boolean_t dprint, const char *file, const char *func,
555     int line, const char *fmt, ...)
556 {
557 	/* Get rid of annoying "../common/" prefix to filename. */
558 	const char *newfile = zfs_basename(file);
559 
560 	va_list adx;
561 	if (dprint) {
562 		/* dprintf messages are printed immediately */
563 
564 		if (!dprintf_print_all &&
565 		    !dprintf_find_string(newfile) &&
566 		    !dprintf_find_string(func))
567 			return;
568 
569 		/* Print out just the function name if requested */
570 		flockfile(stdout);
571 		if (dprintf_find_string("pid"))
572 			(void) printf("%d ", getpid());
573 		if (dprintf_find_string("tid"))
574 			(void) printf("%ju ",
575 			    (uintmax_t)(uintptr_t)pthread_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("dprintf: %s: ", func);
583 		va_start(adx, fmt);
584 		(void) vprintf(fmt, adx);
585 		va_end(adx);
586 		funlockfile(stdout);
587 	} else {
588 		/* zfs_dbgmsg is logged for dumping later */
589 		size_t size;
590 		char *buf;
591 		int i;
592 
593 		size = 1024;
594 		buf = umem_alloc(size, UMEM_NOFAIL);
595 		i = snprintf(buf, size, "%s:%d:%s(): ", newfile, line, func);
596 
597 		if (i < size) {
598 			va_start(adx, fmt);
599 			(void) vsnprintf(buf + i, size - i, fmt, adx);
600 			va_end(adx);
601 		}
602 
603 		__zfs_dbgmsg(buf);
604 
605 		umem_free(buf, size);
606 	}
607 }
608 
609 /*
610  * =========================================================================
611  * cmn_err() and panic()
612  * =========================================================================
613  */
614 static char ce_prefix[CE_IGNORE][10] = { "", "NOTICE: ", "WARNING: ", "" };
615 static char ce_suffix[CE_IGNORE][2] = { "", "\n", "\n", "" };
616 
617 void
618 vpanic(const char *fmt, va_list adx)
619 {
620 	(void) fprintf(stderr, "error: ");
621 	(void) vfprintf(stderr, fmt, adx);
622 	(void) fprintf(stderr, "\n");
623 
624 	abort();	/* think of it as a "user-level crash dump" */
625 }
626 
627 void
628 panic(const char *fmt, ...)
629 {
630 	va_list adx;
631 
632 	va_start(adx, fmt);
633 	vpanic(fmt, adx);
634 	va_end(adx);
635 }
636 
637 void
638 vcmn_err(int ce, const char *fmt, va_list adx)
639 {
640 	if (ce == CE_PANIC)
641 		vpanic(fmt, adx);
642 	if (ce != CE_NOTE) {	/* suppress noise in userland stress testing */
643 		(void) fprintf(stderr, "%s", ce_prefix[ce]);
644 		(void) vfprintf(stderr, fmt, adx);
645 		(void) fprintf(stderr, "%s", ce_suffix[ce]);
646 	}
647 }
648 
649 void
650 cmn_err(int ce, const char *fmt, ...)
651 {
652 	va_list adx;
653 
654 	va_start(adx, fmt);
655 	vcmn_err(ce, fmt, adx);
656 	va_end(adx);
657 }
658 
659 /*
660  * =========================================================================
661  * misc routines
662  * =========================================================================
663  */
664 
665 void
666 delay(clock_t ticks)
667 {
668 	(void) poll(0, 0, ticks * (1000 / hz));
669 }
670 
671 /*
672  * Find highest one bit set.
673  * Returns bit number + 1 of highest bit that is set, otherwise returns 0.
674  * The __builtin_clzll() function is supported by both GCC and Clang.
675  */
676 int
677 highbit64(uint64_t i)
678 {
679 	if (i == 0)
680 	return (0);
681 
682 	return (NBBY * sizeof (uint64_t) - __builtin_clzll(i));
683 }
684 
685 /*
686  * Find lowest one bit set.
687  * Returns bit number + 1 of lowest bit that is set, otherwise returns 0.
688  * The __builtin_ffsll() function is supported by both GCC and Clang.
689  */
690 int
691 lowbit64(uint64_t i)
692 {
693 	if (i == 0)
694 		return (0);
695 
696 	return (__builtin_ffsll(i));
697 }
698 
699 const char *random_path = "/dev/random";
700 const char *urandom_path = "/dev/urandom";
701 static int random_fd = -1, urandom_fd = -1;
702 
703 void
704 random_init(void)
705 {
706 	VERIFY((random_fd = open(random_path, O_RDONLY | O_CLOEXEC)) != -1);
707 	VERIFY((urandom_fd = open(urandom_path, O_RDONLY | O_CLOEXEC)) != -1);
708 }
709 
710 void
711 random_fini(void)
712 {
713 	close(random_fd);
714 	close(urandom_fd);
715 
716 	random_fd = -1;
717 	urandom_fd = -1;
718 }
719 
720 static int
721 random_get_bytes_common(uint8_t *ptr, size_t len, int fd)
722 {
723 	size_t resid = len;
724 	ssize_t bytes;
725 
726 	ASSERT(fd != -1);
727 
728 	while (resid != 0) {
729 		bytes = read(fd, ptr, resid);
730 		ASSERT3S(bytes, >=, 0);
731 		ptr += bytes;
732 		resid -= bytes;
733 	}
734 
735 	return (0);
736 }
737 
738 int
739 random_get_bytes(uint8_t *ptr, size_t len)
740 {
741 	return (random_get_bytes_common(ptr, len, random_fd));
742 }
743 
744 int
745 random_get_pseudo_bytes(uint8_t *ptr, size_t len)
746 {
747 	return (random_get_bytes_common(ptr, len, urandom_fd));
748 }
749 
750 int
751 ddi_strtoul(const char *hw_serial, char **nptr, int base, unsigned long *result)
752 {
753 	(void) nptr;
754 	char *end;
755 
756 	*result = strtoul(hw_serial, &end, base);
757 	if (*result == 0)
758 		return (errno);
759 	return (0);
760 }
761 
762 int
763 ddi_strtoull(const char *str, char **nptr, int base, u_longlong_t *result)
764 {
765 	(void) nptr;
766 	char *end;
767 
768 	*result = strtoull(str, &end, base);
769 	if (*result == 0)
770 		return (errno);
771 	return (0);
772 }
773 
774 utsname_t *
775 utsname(void)
776 {
777 	return (&hw_utsname);
778 }
779 
780 /*
781  * =========================================================================
782  * kernel emulation setup & teardown
783  * =========================================================================
784  */
785 static int
786 umem_out_of_memory(void)
787 {
788 	char errmsg[] = "out of memory -- generating core dump\n";
789 
790 	(void) fprintf(stderr, "%s", errmsg);
791 	abort();
792 	return (0);
793 }
794 
795 void
796 kernel_init(int mode)
797 {
798 	extern uint_t rrw_tsd_key;
799 
800 	umem_nofail_callback(umem_out_of_memory);
801 
802 	physmem = sysconf(_SC_PHYS_PAGES);
803 
804 	dprintf("physmem = %llu pages (%.2f GB)\n", (u_longlong_t)physmem,
805 	    (double)physmem * sysconf(_SC_PAGE_SIZE) / (1ULL << 30));
806 
807 	(void) snprintf(hw_serial, sizeof (hw_serial), "%ld",
808 	    (mode & SPA_MODE_WRITE) ? get_system_hostid() : 0);
809 
810 	random_init();
811 
812 	VERIFY0(uname(&hw_utsname));
813 
814 	system_taskq_init();
815 	icp_init();
816 
817 	zstd_init();
818 
819 	spa_init((spa_mode_t)mode);
820 
821 	fletcher_4_init();
822 
823 	tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
824 }
825 
826 void
827 kernel_fini(void)
828 {
829 	fletcher_4_fini();
830 	spa_fini();
831 
832 	zstd_fini();
833 
834 	icp_fini();
835 	system_taskq_fini();
836 
837 	random_fini();
838 }
839 
840 uid_t
841 crgetuid(cred_t *cr)
842 {
843 	(void) cr;
844 	return (0);
845 }
846 
847 uid_t
848 crgetruid(cred_t *cr)
849 {
850 	(void) cr;
851 	return (0);
852 }
853 
854 gid_t
855 crgetgid(cred_t *cr)
856 {
857 	(void) cr;
858 	return (0);
859 }
860 
861 int
862 crgetngroups(cred_t *cr)
863 {
864 	(void) cr;
865 	return (0);
866 }
867 
868 gid_t *
869 crgetgroups(cred_t *cr)
870 {
871 	(void) cr;
872 	return (NULL);
873 }
874 
875 int
876 zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr)
877 {
878 	(void) name, (void) cr;
879 	return (0);
880 }
881 
882 int
883 zfs_secpolicy_rename_perms(const char *from, const char *to, cred_t *cr)
884 {
885 	(void) from, (void) to, (void) cr;
886 	return (0);
887 }
888 
889 int
890 zfs_secpolicy_destroy_perms(const char *name, cred_t *cr)
891 {
892 	(void) name, (void) cr;
893 	return (0);
894 }
895 
896 int
897 secpolicy_zfs(const cred_t *cr)
898 {
899 	(void) cr;
900 	return (0);
901 }
902 
903 int
904 secpolicy_zfs_proc(const cred_t *cr, proc_t *proc)
905 {
906 	(void) cr, (void) proc;
907 	return (0);
908 }
909 
910 ksiddomain_t *
911 ksid_lookupdomain(const char *dom)
912 {
913 	ksiddomain_t *kd;
914 
915 	kd = umem_zalloc(sizeof (ksiddomain_t), UMEM_NOFAIL);
916 	kd->kd_name = spa_strdup(dom);
917 	return (kd);
918 }
919 
920 void
921 ksiddomain_rele(ksiddomain_t *ksid)
922 {
923 	spa_strfree(ksid->kd_name);
924 	umem_free(ksid, sizeof (ksiddomain_t));
925 }
926 
927 char *
928 kmem_vasprintf(const char *fmt, va_list adx)
929 {
930 	char *buf = NULL;
931 	va_list adx_copy;
932 
933 	va_copy(adx_copy, adx);
934 	VERIFY(vasprintf(&buf, fmt, adx_copy) != -1);
935 	va_end(adx_copy);
936 
937 	return (buf);
938 }
939 
940 char *
941 kmem_asprintf(const char *fmt, ...)
942 {
943 	char *buf = NULL;
944 	va_list adx;
945 
946 	va_start(adx, fmt);
947 	VERIFY(vasprintf(&buf, fmt, adx) != -1);
948 	va_end(adx);
949 
950 	return (buf);
951 }
952 
953 zfs_file_t *
954 zfs_onexit_fd_hold(int fd, minor_t *minorp)
955 {
956 	(void) fd;
957 	*minorp = 0;
958 	return (NULL);
959 }
960 
961 void
962 zfs_onexit_fd_rele(zfs_file_t *fp)
963 {
964 	(void) fp;
965 }
966 
967 int
968 zfs_onexit_add_cb(minor_t minor, void (*func)(void *), void *data,
969     uint64_t *action_handle)
970 {
971 	(void) minor, (void) func, (void) data, (void) action_handle;
972 	return (0);
973 }
974 
975 fstrans_cookie_t
976 spl_fstrans_mark(void)
977 {
978 	return ((fstrans_cookie_t)0);
979 }
980 
981 void
982 spl_fstrans_unmark(fstrans_cookie_t cookie)
983 {
984 	(void) cookie;
985 }
986 
987 int
988 __spl_pf_fstrans_check(void)
989 {
990 	return (0);
991 }
992 
993 int
994 kmem_cache_reap_active(void)
995 {
996 	return (0);
997 }
998 
999 void *zvol_tag = "zvol_tag";
1000 
1001 void
1002 zvol_create_minor(const char *name)
1003 {
1004 	(void) name;
1005 }
1006 
1007 void
1008 zvol_create_minors_recursive(const char *name)
1009 {
1010 	(void) name;
1011 }
1012 
1013 void
1014 zvol_remove_minors(spa_t *spa, const char *name, boolean_t async)
1015 {
1016 	(void) spa, (void) name, (void) async;
1017 }
1018 
1019 void
1020 zvol_rename_minors(spa_t *spa, const char *oldname, const char *newname,
1021     boolean_t async)
1022 {
1023 	(void) spa, (void) oldname, (void) newname, (void) async;
1024 }
1025 
1026 /*
1027  * Open file
1028  *
1029  * path - fully qualified path to file
1030  * flags - file attributes O_READ / O_WRITE / O_EXCL
1031  * fpp - pointer to return file pointer
1032  *
1033  * Returns 0 on success underlying error on failure.
1034  */
1035 int
1036 zfs_file_open(const char *path, int flags, int mode, zfs_file_t **fpp)
1037 {
1038 	int fd = -1;
1039 	int dump_fd = -1;
1040 	int err;
1041 	int old_umask = 0;
1042 	zfs_file_t *fp;
1043 	struct stat64 st;
1044 
1045 	if (!(flags & O_CREAT) && stat64(path, &st) == -1)
1046 		return (errno);
1047 
1048 	if (!(flags & O_CREAT) && S_ISBLK(st.st_mode))
1049 		flags |= O_DIRECT;
1050 
1051 	if (flags & O_CREAT)
1052 		old_umask = umask(0);
1053 
1054 	fd = open64(path, flags, mode);
1055 	if (fd == -1)
1056 		return (errno);
1057 
1058 	if (flags & O_CREAT)
1059 		(void) umask(old_umask);
1060 
1061 	if (vn_dumpdir != NULL) {
1062 		char *dumppath = umem_zalloc(MAXPATHLEN, UMEM_NOFAIL);
1063 		const char *inpath = zfs_basename(path);
1064 
1065 		(void) snprintf(dumppath, MAXPATHLEN,
1066 		    "%s/%s", vn_dumpdir, inpath);
1067 		dump_fd = open64(dumppath, O_CREAT | O_WRONLY, 0666);
1068 		umem_free(dumppath, MAXPATHLEN);
1069 		if (dump_fd == -1) {
1070 			err = errno;
1071 			close(fd);
1072 			return (err);
1073 		}
1074 	} else {
1075 		dump_fd = -1;
1076 	}
1077 
1078 	(void) fcntl(fd, F_SETFD, FD_CLOEXEC);
1079 
1080 	fp = umem_zalloc(sizeof (zfs_file_t), UMEM_NOFAIL);
1081 	fp->f_fd = fd;
1082 	fp->f_dump_fd = dump_fd;
1083 	*fpp = fp;
1084 
1085 	return (0);
1086 }
1087 
1088 void
1089 zfs_file_close(zfs_file_t *fp)
1090 {
1091 	close(fp->f_fd);
1092 	if (fp->f_dump_fd != -1)
1093 		close(fp->f_dump_fd);
1094 
1095 	umem_free(fp, sizeof (zfs_file_t));
1096 }
1097 
1098 /*
1099  * Stateful write - use os internal file pointer to determine where to
1100  * write and update on successful completion.
1101  *
1102  * fp -  pointer to file (pipe, socket, etc) to write to
1103  * buf - buffer to write
1104  * count - # of bytes to write
1105  * resid -  pointer to count of unwritten bytes  (if short write)
1106  *
1107  * Returns 0 on success errno on failure.
1108  */
1109 int
1110 zfs_file_write(zfs_file_t *fp, const void *buf, size_t count, ssize_t *resid)
1111 {
1112 	ssize_t rc;
1113 
1114 	rc = write(fp->f_fd, buf, count);
1115 	if (rc < 0)
1116 		return (errno);
1117 
1118 	if (resid) {
1119 		*resid = count - rc;
1120 	} else if (rc != count) {
1121 		return (EIO);
1122 	}
1123 
1124 	return (0);
1125 }
1126 
1127 /*
1128  * Stateless write - os internal file pointer is not updated.
1129  *
1130  * fp -  pointer to file (pipe, socket, etc) to write to
1131  * buf - buffer to write
1132  * count - # of bytes to write
1133  * off - file offset to write to (only valid for seekable types)
1134  * resid -  pointer to count of unwritten bytes
1135  *
1136  * Returns 0 on success errno on failure.
1137  */
1138 int
1139 zfs_file_pwrite(zfs_file_t *fp, const void *buf,
1140     size_t count, loff_t pos, ssize_t *resid)
1141 {
1142 	ssize_t rc, split, done;
1143 	int sectors;
1144 
1145 	/*
1146 	 * To simulate partial disk writes, we split writes into two
1147 	 * system calls so that the process can be killed in between.
1148 	 * This is used by ztest to simulate realistic failure modes.
1149 	 */
1150 	sectors = count >> SPA_MINBLOCKSHIFT;
1151 	split = (sectors > 0 ? rand() % sectors : 0) << SPA_MINBLOCKSHIFT;
1152 	rc = pwrite64(fp->f_fd, buf, split, pos);
1153 	if (rc != -1) {
1154 		done = rc;
1155 		rc = pwrite64(fp->f_fd, (char *)buf + split,
1156 		    count - split, pos + split);
1157 	}
1158 #ifdef __linux__
1159 	if (rc == -1 && errno == EINVAL) {
1160 		/*
1161 		 * Under Linux, this most likely means an alignment issue
1162 		 * (memory or disk) due to O_DIRECT, so we abort() in order
1163 		 * to catch the offender.
1164 		 */
1165 		abort();
1166 	}
1167 #endif
1168 
1169 	if (rc < 0)
1170 		return (errno);
1171 
1172 	done += rc;
1173 
1174 	if (resid) {
1175 		*resid = count - done;
1176 	} else if (done != count) {
1177 		return (EIO);
1178 	}
1179 
1180 	return (0);
1181 }
1182 
1183 /*
1184  * Stateful read - use os internal file pointer to determine where to
1185  * read and update on successful completion.
1186  *
1187  * fp -  pointer to file (pipe, socket, etc) to read from
1188  * buf - buffer to write
1189  * count - # of bytes to read
1190  * resid -  pointer to count of unread bytes (if short read)
1191  *
1192  * Returns 0 on success errno on failure.
1193  */
1194 int
1195 zfs_file_read(zfs_file_t *fp, void *buf, size_t count, ssize_t *resid)
1196 {
1197 	int rc;
1198 
1199 	rc = read(fp->f_fd, buf, count);
1200 	if (rc < 0)
1201 		return (errno);
1202 
1203 	if (resid) {
1204 		*resid = count - rc;
1205 	} else if (rc != count) {
1206 		return (EIO);
1207 	}
1208 
1209 	return (0);
1210 }
1211 
1212 /*
1213  * Stateless read - os internal file pointer is not updated.
1214  *
1215  * fp -  pointer to file (pipe, socket, etc) to read from
1216  * buf - buffer to write
1217  * count - # of bytes to write
1218  * off - file offset to read from (only valid for seekable types)
1219  * resid -  pointer to count of unwritten bytes (if short write)
1220  *
1221  * Returns 0 on success errno on failure.
1222  */
1223 int
1224 zfs_file_pread(zfs_file_t *fp, void *buf, size_t count, loff_t off,
1225     ssize_t *resid)
1226 {
1227 	ssize_t rc;
1228 
1229 	rc = pread64(fp->f_fd, buf, count, off);
1230 	if (rc < 0) {
1231 #ifdef __linux__
1232 		/*
1233 		 * Under Linux, this most likely means an alignment issue
1234 		 * (memory or disk) due to O_DIRECT, so we abort() in order to
1235 		 * catch the offender.
1236 		 */
1237 		if (errno == EINVAL)
1238 			abort();
1239 #endif
1240 		return (errno);
1241 	}
1242 
1243 	if (fp->f_dump_fd != -1) {
1244 		int status;
1245 
1246 		status = pwrite64(fp->f_dump_fd, buf, rc, off);
1247 		ASSERT(status != -1);
1248 	}
1249 
1250 	if (resid) {
1251 		*resid = count - rc;
1252 	} else if (rc != count) {
1253 		return (EIO);
1254 	}
1255 
1256 	return (0);
1257 }
1258 
1259 /*
1260  * lseek - set / get file pointer
1261  *
1262  * fp -  pointer to file (pipe, socket, etc) to read from
1263  * offp - value to seek to, returns current value plus passed offset
1264  * whence - see man pages for standard lseek whence values
1265  *
1266  * Returns 0 on success errno on failure (ESPIPE for non seekable types)
1267  */
1268 int
1269 zfs_file_seek(zfs_file_t *fp, loff_t *offp, int whence)
1270 {
1271 	loff_t rc;
1272 
1273 	rc = lseek(fp->f_fd, *offp, whence);
1274 	if (rc < 0)
1275 		return (errno);
1276 
1277 	*offp = rc;
1278 
1279 	return (0);
1280 }
1281 
1282 /*
1283  * Get file attributes
1284  *
1285  * filp - file pointer
1286  * zfattr - pointer to file attr structure
1287  *
1288  * Currently only used for fetching size and file mode
1289  *
1290  * Returns 0 on success or error code of underlying getattr call on failure.
1291  */
1292 int
1293 zfs_file_getattr(zfs_file_t *fp, zfs_file_attr_t *zfattr)
1294 {
1295 	struct stat64 st;
1296 
1297 	if (fstat64_blk(fp->f_fd, &st) == -1)
1298 		return (errno);
1299 
1300 	zfattr->zfa_size = st.st_size;
1301 	zfattr->zfa_mode = st.st_mode;
1302 
1303 	return (0);
1304 }
1305 
1306 /*
1307  * Sync file to disk
1308  *
1309  * filp - file pointer
1310  * flags - O_SYNC and or O_DSYNC
1311  *
1312  * Returns 0 on success or error code of underlying sync call on failure.
1313  */
1314 int
1315 zfs_file_fsync(zfs_file_t *fp, int flags)
1316 {
1317 	(void) flags;
1318 
1319 	if (fsync(fp->f_fd) < 0)
1320 		return (errno);
1321 
1322 	return (0);
1323 }
1324 
1325 /*
1326  * fallocate - allocate or free space on disk
1327  *
1328  * fp - file pointer
1329  * mode (non-standard options for hole punching etc)
1330  * offset - offset to start allocating or freeing from
1331  * len - length to free / allocate
1332  *
1333  * OPTIONAL
1334  */
1335 int
1336 zfs_file_fallocate(zfs_file_t *fp, int mode, loff_t offset, loff_t len)
1337 {
1338 #ifdef __linux__
1339 	return (fallocate(fp->f_fd, mode, offset, len));
1340 #else
1341 	(void) fp, (void) mode, (void) offset, (void) len;
1342 	return (EOPNOTSUPP);
1343 #endif
1344 }
1345 
1346 /*
1347  * Request current file pointer offset
1348  *
1349  * fp - pointer to file
1350  *
1351  * Returns current file offset.
1352  */
1353 loff_t
1354 zfs_file_off(zfs_file_t *fp)
1355 {
1356 	return (lseek(fp->f_fd, SEEK_CUR, 0));
1357 }
1358 
1359 /*
1360  * unlink file
1361  *
1362  * path - fully qualified file path
1363  *
1364  * Returns 0 on success.
1365  *
1366  * OPTIONAL
1367  */
1368 int
1369 zfs_file_unlink(const char *path)
1370 {
1371 	return (remove(path));
1372 }
1373 
1374 /*
1375  * Get reference to file pointer
1376  *
1377  * fd - input file descriptor
1378  *
1379  * Returns pointer to file struct or NULL.
1380  * Unsupported in user space.
1381  */
1382 zfs_file_t *
1383 zfs_file_get(int fd)
1384 {
1385 	(void) fd;
1386 	abort();
1387 	return (NULL);
1388 }
1389 /*
1390  * Drop reference to file pointer
1391  *
1392  * fp - pointer to file struct
1393  *
1394  * Unsupported in user space.
1395  */
1396 void
1397 zfs_file_put(zfs_file_t *fp)
1398 {
1399 	abort();
1400 	(void) fp;
1401 }
1402 
1403 void
1404 zfsvfs_update_fromname(const char *oldname, const char *newname)
1405 {
1406 	(void) oldname, (void) newname;
1407 }
1408