xref: /freebsd/sys/contrib/openzfs/module/os/linux/spl/spl-generic.c (revision b64c5a0ace59af62eff52bfe110a521dc73c937b)
1 /*
2  *  Copyright (C) 2007-2010 Lawrence Livermore National Security, LLC.
3  *  Copyright (C) 2007 The Regents of the University of California.
4  *  Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
5  *  Written by Brian Behlendorf <behlendorf1@llnl.gov>.
6  *  UCRL-CODE-235197
7  *
8  *  This file is part of the SPL, Solaris Porting Layer.
9  *
10  *  The SPL is free software; you can redistribute it and/or modify it
11  *  under the terms of the GNU General Public License as published by the
12  *  Free Software Foundation; either version 2 of the License, or (at your
13  *  option) any later version.
14  *
15  *  The SPL is distributed in the hope that it will be useful, but WITHOUT
16  *  ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
17  *  FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
18  *  for more details.
19  *
20  *  You should have received a copy of the GNU General Public License along
21  *  with the SPL.  If not, see <http://www.gnu.org/licenses/>.
22  *
23  *  Solaris Porting Layer (SPL) Generic Implementation.
24  */
25 
26 #include <sys/isa_defs.h>
27 #include <sys/sysmacros.h>
28 #include <sys/systeminfo.h>
29 #include <sys/vmsystm.h>
30 #include <sys/kmem.h>
31 #include <sys/kmem_cache.h>
32 #include <sys/vmem.h>
33 #include <sys/mutex.h>
34 #include <sys/rwlock.h>
35 #include <sys/taskq.h>
36 #include <sys/tsd.h>
37 #include <sys/zmod.h>
38 #include <sys/debug.h>
39 #include <sys/proc.h>
40 #include <sys/kstat.h>
41 #include <sys/file.h>
42 #include <sys/sunddi.h>
43 #include <linux/ctype.h>
44 #include <sys/disp.h>
45 #include <sys/random.h>
46 #include <sys/string.h>
47 #include <linux/kmod.h>
48 #include <linux/mod_compat.h>
49 #include <sys/cred.h>
50 #include <sys/vnode.h>
51 #include <sys/misc.h>
52 #include <linux/mod_compat.h>
53 
54 unsigned long spl_hostid = 0;
55 EXPORT_SYMBOL(spl_hostid);
56 
57 module_param(spl_hostid, ulong, 0644);
58 MODULE_PARM_DESC(spl_hostid, "The system hostid.");
59 
60 proc_t p0;
61 EXPORT_SYMBOL(p0);
62 
63 /*
64  * xoshiro256++ 1.0 PRNG by David Blackman and Sebastiano Vigna
65  *
66  * "Scrambled Linear Pseudorandom Number Generators∗"
67  * https://vigna.di.unimi.it/ftp/papers/ScrambledLinear.pdf
68  *
69  * random_get_pseudo_bytes() is an API function on Illumos whose sole purpose
70  * is to provide bytes containing random numbers. It is mapped to /dev/urandom
71  * on Illumos, which uses a "FIPS 186-2 algorithm". No user of the SPL's
72  * random_get_pseudo_bytes() needs bytes that are of cryptographic quality, so
73  * we can implement it using a fast PRNG that we seed using Linux' actual
74  * equivalent to random_get_pseudo_bytes(). We do this by providing each CPU
75  * with an independent seed so that all calls to random_get_pseudo_bytes() are
76  * free of atomic instructions.
77  *
78  * A consequence of using a fast PRNG is that using random_get_pseudo_bytes()
79  * to generate words larger than 256 bits will paradoxically be limited to
80  * `2^256 - 1` possibilities. This is because we have a sequence of `2^256 - 1`
81  * 256-bit words and selecting the first will implicitly select the second. If
82  * a caller finds this behavior undesirable, random_get_bytes() should be used
83  * instead.
84  *
85  * XXX: Linux interrupt handlers that trigger within the critical section
86  * formed by `s[3] = xp[3];` and `xp[0] = s[0];` and call this function will
87  * see the same numbers. Nothing in the code currently calls this in an
88  * interrupt handler, so this is considered to be okay. If that becomes a
89  * problem, we could create a set of per-cpu variables for interrupt handlers
90  * and use them when in_interrupt() from linux/preempt_mask.h evaluates to
91  * true.
92  */
93 static void __percpu *spl_pseudo_entropy;
94 
95 /*
96  * rotl()/spl_rand_next()/spl_rand_jump() are copied from the following CC-0
97  * licensed file:
98  *
99  * https://prng.di.unimi.it/xoshiro256plusplus.c
100  */
101 
102 static inline uint64_t rotl(const uint64_t x, int k)
103 {
104 	return ((x << k) | (x >> (64 - k)));
105 }
106 
107 static inline uint64_t
108 spl_rand_next(uint64_t *s)
109 {
110 	const uint64_t result = rotl(s[0] + s[3], 23) + s[0];
111 
112 	const uint64_t t = s[1] << 17;
113 
114 	s[2] ^= s[0];
115 	s[3] ^= s[1];
116 	s[1] ^= s[2];
117 	s[0] ^= s[3];
118 
119 	s[2] ^= t;
120 
121 	s[3] = rotl(s[3], 45);
122 
123 	return (result);
124 }
125 
126 static inline void
127 spl_rand_jump(uint64_t *s)
128 {
129 	static const uint64_t JUMP[] = { 0x180ec6d33cfd0aba,
130 	    0xd5a61266f0c9392c, 0xa9582618e03fc9aa, 0x39abdc4529b1661c };
131 
132 	uint64_t s0 = 0;
133 	uint64_t s1 = 0;
134 	uint64_t s2 = 0;
135 	uint64_t s3 = 0;
136 	int i, b;
137 	for (i = 0; i < sizeof (JUMP) / sizeof (*JUMP); i++)
138 		for (b = 0; b < 64; b++) {
139 			if (JUMP[i] & 1ULL << b) {
140 				s0 ^= s[0];
141 				s1 ^= s[1];
142 				s2 ^= s[2];
143 				s3 ^= s[3];
144 			}
145 			(void) spl_rand_next(s);
146 		}
147 
148 	s[0] = s0;
149 	s[1] = s1;
150 	s[2] = s2;
151 	s[3] = s3;
152 }
153 
154 int
155 random_get_pseudo_bytes(uint8_t *ptr, size_t len)
156 {
157 	uint64_t *xp, s[4];
158 
159 	ASSERT(ptr);
160 
161 	xp = get_cpu_ptr(spl_pseudo_entropy);
162 
163 	s[0] = xp[0];
164 	s[1] = xp[1];
165 	s[2] = xp[2];
166 	s[3] = xp[3];
167 
168 	while (len) {
169 		union {
170 			uint64_t ui64;
171 			uint8_t byte[sizeof (uint64_t)];
172 		}entropy;
173 		int i = MIN(len, sizeof (uint64_t));
174 
175 		len -= i;
176 		entropy.ui64 = spl_rand_next(s);
177 
178 		/*
179 		 * xoshiro256++ has low entropy lower bytes, so we copy the
180 		 * higher order bytes first.
181 		 */
182 		while (i--)
183 #ifdef _ZFS_BIG_ENDIAN
184 			*ptr++ = entropy.byte[i];
185 #else
186 			*ptr++ = entropy.byte[7 - i];
187 #endif
188 	}
189 
190 	xp[0] = s[0];
191 	xp[1] = s[1];
192 	xp[2] = s[2];
193 	xp[3] = s[3];
194 
195 	put_cpu_ptr(spl_pseudo_entropy);
196 
197 	return (0);
198 }
199 
200 
201 EXPORT_SYMBOL(random_get_pseudo_bytes);
202 
203 #if BITS_PER_LONG == 32
204 
205 /*
206  * Support 64/64 => 64 division on a 32-bit platform.  While the kernel
207  * provides a div64_u64() function for this we do not use it because the
208  * implementation is flawed.  There are cases which return incorrect
209  * results as late as linux-2.6.35.  Until this is fixed upstream the
210  * spl must provide its own implementation.
211  *
212  * This implementation is a slightly modified version of the algorithm
213  * proposed by the book 'Hacker's Delight'.  The original source can be
214  * found here and is available for use without restriction.
215  *
216  * http://www.hackersdelight.org/HDcode/newCode/divDouble.c
217  */
218 
219 /*
220  * Calculate number of leading of zeros for a 64-bit value.
221  */
222 static int
223 nlz64(uint64_t x)
224 {
225 	register int n = 0;
226 
227 	if (x == 0)
228 		return (64);
229 
230 	if (x <= 0x00000000FFFFFFFFULL) { n = n + 32; x = x << 32; }
231 	if (x <= 0x0000FFFFFFFFFFFFULL) { n = n + 16; x = x << 16; }
232 	if (x <= 0x00FFFFFFFFFFFFFFULL) { n = n +  8; x = x <<  8; }
233 	if (x <= 0x0FFFFFFFFFFFFFFFULL) { n = n +  4; x = x <<  4; }
234 	if (x <= 0x3FFFFFFFFFFFFFFFULL) { n = n +  2; x = x <<  2; }
235 	if (x <= 0x7FFFFFFFFFFFFFFFULL) { n = n +  1; }
236 
237 	return (n);
238 }
239 
240 /*
241  * Newer kernels have a div_u64() function but we define our own
242  * to simplify portability between kernel versions.
243  */
244 static inline uint64_t
245 __div_u64(uint64_t u, uint32_t v)
246 {
247 	(void) do_div(u, v);
248 	return (u);
249 }
250 
251 /*
252  * Turn off missing prototypes warning for these functions. They are
253  * replacements for libgcc-provided functions and will never be called
254  * directly.
255  */
256 #if defined(__GNUC__) && !defined(__clang__)
257 #pragma GCC diagnostic push
258 #pragma GCC diagnostic ignored "-Wmissing-prototypes"
259 #endif
260 
261 /*
262  * Implementation of 64-bit unsigned division for 32-bit machines.
263  *
264  * First the procedure takes care of the case in which the divisor is a
265  * 32-bit quantity. There are two subcases: (1) If the left half of the
266  * dividend is less than the divisor, one execution of do_div() is all that
267  * is required (overflow is not possible). (2) Otherwise it does two
268  * divisions, using the grade school method.
269  */
270 uint64_t
271 __udivdi3(uint64_t u, uint64_t v)
272 {
273 	uint64_t u0, u1, v1, q0, q1, k;
274 	int n;
275 
276 	if (v >> 32 == 0) {			// If v < 2**32:
277 		if (u >> 32 < v) {		// If u/v cannot overflow,
278 			return (__div_u64(u, v)); // just do one division.
279 		} else {			// If u/v would overflow:
280 			u1 = u >> 32;		// Break u into two halves.
281 			u0 = u & 0xFFFFFFFF;
282 			q1 = __div_u64(u1, v);	// First quotient digit.
283 			k  = u1 - q1 * v;	// First remainder, < v.
284 			u0 += (k << 32);
285 			q0 = __div_u64(u0, v);	// Seconds quotient digit.
286 			return ((q1 << 32) + q0);
287 		}
288 	} else {				// If v >= 2**32:
289 		n = nlz64(v);			// 0 <= n <= 31.
290 		v1 = (v << n) >> 32;		// Normalize divisor, MSB is 1.
291 		u1 = u >> 1;			// To ensure no overflow.
292 		q1 = __div_u64(u1, v1);		// Get quotient from
293 		q0 = (q1 << n) >> 31;		// Undo normalization and
294 						// division of u by 2.
295 		if (q0 != 0)			// Make q0 correct or
296 			q0 = q0 - 1;		// too small by 1.
297 		if ((u - q0 * v) >= v)
298 			q0 = q0 + 1;		// Now q0 is correct.
299 
300 		return (q0);
301 	}
302 }
303 EXPORT_SYMBOL(__udivdi3);
304 
305 #ifndef abs64
306 /* CSTYLED */
307 #define	abs64(x)	({ uint64_t t = (x) >> 63; ((x) ^ t) - t; })
308 #endif
309 
310 /*
311  * Implementation of 64-bit signed division for 32-bit machines.
312  */
313 int64_t
314 __divdi3(int64_t u, int64_t v)
315 {
316 	int64_t q, t;
317 	q = __udivdi3(abs64(u), abs64(v));
318 	t = (u ^ v) >> 63;	// If u, v have different
319 	return ((q ^ t) - t);	// signs, negate q.
320 }
321 EXPORT_SYMBOL(__divdi3);
322 
323 /*
324  * Implementation of 64-bit unsigned modulo for 32-bit machines.
325  */
326 uint64_t
327 __umoddi3(uint64_t dividend, uint64_t divisor)
328 {
329 	return (dividend - (divisor * __udivdi3(dividend, divisor)));
330 }
331 EXPORT_SYMBOL(__umoddi3);
332 
333 /* 64-bit signed modulo for 32-bit machines. */
334 int64_t
335 __moddi3(int64_t n, int64_t d)
336 {
337 	int64_t q;
338 	boolean_t nn = B_FALSE;
339 
340 	if (n < 0) {
341 		nn = B_TRUE;
342 		n = -n;
343 	}
344 	if (d < 0)
345 		d = -d;
346 
347 	q = __umoddi3(n, d);
348 
349 	return (nn ? -q : q);
350 }
351 EXPORT_SYMBOL(__moddi3);
352 
353 /*
354  * Implementation of 64-bit unsigned division/modulo for 32-bit machines.
355  */
356 uint64_t
357 __udivmoddi4(uint64_t n, uint64_t d, uint64_t *r)
358 {
359 	uint64_t q = __udivdi3(n, d);
360 	if (r)
361 		*r = n - d * q;
362 	return (q);
363 }
364 EXPORT_SYMBOL(__udivmoddi4);
365 
366 /*
367  * Implementation of 64-bit signed division/modulo for 32-bit machines.
368  */
369 int64_t
370 __divmoddi4(int64_t n, int64_t d, int64_t *r)
371 {
372 	int64_t q, rr;
373 	boolean_t nn = B_FALSE;
374 	boolean_t nd = B_FALSE;
375 	if (n < 0) {
376 		nn = B_TRUE;
377 		n = -n;
378 	}
379 	if (d < 0) {
380 		nd = B_TRUE;
381 		d = -d;
382 	}
383 
384 	q = __udivmoddi4(n, d, (uint64_t *)&rr);
385 
386 	if (nn != nd)
387 		q = -q;
388 	if (nn)
389 		rr = -rr;
390 	if (r)
391 		*r = rr;
392 	return (q);
393 }
394 EXPORT_SYMBOL(__divmoddi4);
395 
396 #if defined(__arm) || defined(__arm__)
397 /*
398  * Implementation of 64-bit (un)signed division for 32-bit arm machines.
399  *
400  * Run-time ABI for the ARM Architecture (page 20).  A pair of (unsigned)
401  * long longs is returned in {{r0, r1}, {r2,r3}}, the quotient in {r0, r1},
402  * and the remainder in {r2, r3}.  The return type is specifically left
403  * set to 'void' to ensure the compiler does not overwrite these registers
404  * during the return.  All results are in registers as per ABI
405  */
406 void
407 __aeabi_uldivmod(uint64_t u, uint64_t v)
408 {
409 	uint64_t res;
410 	uint64_t mod;
411 
412 	res = __udivdi3(u, v);
413 	mod = __umoddi3(u, v);
414 	{
415 		register uint32_t r0 asm("r0") = (res & 0xFFFFFFFF);
416 		register uint32_t r1 asm("r1") = (res >> 32);
417 		register uint32_t r2 asm("r2") = (mod & 0xFFFFFFFF);
418 		register uint32_t r3 asm("r3") = (mod >> 32);
419 
420 		asm volatile(""
421 		    : "+r"(r0), "+r"(r1), "+r"(r2), "+r"(r3)  /* output */
422 		    : "r"(r0), "r"(r1), "r"(r2), "r"(r3));    /* input */
423 
424 		return; /* r0; */
425 	}
426 }
427 EXPORT_SYMBOL(__aeabi_uldivmod);
428 
429 void
430 __aeabi_ldivmod(int64_t u, int64_t v)
431 {
432 	int64_t res;
433 	uint64_t mod;
434 
435 	res =  __divdi3(u, v);
436 	mod = __umoddi3(u, v);
437 	{
438 		register uint32_t r0 asm("r0") = (res & 0xFFFFFFFF);
439 		register uint32_t r1 asm("r1") = (res >> 32);
440 		register uint32_t r2 asm("r2") = (mod & 0xFFFFFFFF);
441 		register uint32_t r3 asm("r3") = (mod >> 32);
442 
443 		asm volatile(""
444 		    : "+r"(r0), "+r"(r1), "+r"(r2), "+r"(r3)  /* output */
445 		    : "r"(r0), "r"(r1), "r"(r2), "r"(r3));    /* input */
446 
447 		return; /* r0; */
448 	}
449 }
450 EXPORT_SYMBOL(__aeabi_ldivmod);
451 #endif /* __arm || __arm__ */
452 
453 #if defined(__GNUC__) && !defined(__clang__)
454 #pragma GCC diagnostic pop
455 #endif
456 
457 #endif /* BITS_PER_LONG */
458 
459 /*
460  * NOTE: The strtoxx behavior is solely based on my reading of the Solaris
461  * ddi_strtol(9F) man page.  I have not verified the behavior of these
462  * functions against their Solaris counterparts.  It is possible that I
463  * may have misinterpreted the man page or the man page is incorrect.
464  */
465 int ddi_strtol(const char *, char **, int, long *);
466 int ddi_strtoull(const char *, char **, int, unsigned long long *);
467 int ddi_strtoll(const char *, char **, int, long long *);
468 
469 #define	define_ddi_strtox(type, valtype)				\
470 int ddi_strto##type(const char *str, char **endptr,			\
471     int base, valtype *result)						\
472 {									\
473 	valtype last_value, value = 0;					\
474 	char *ptr = (char *)str;					\
475 	int digit, minus = 0;						\
476 									\
477 	while (strchr(" \t\n\r\f", *ptr))				\
478 		++ptr;							\
479 									\
480 	if (strlen(ptr) == 0)						\
481 		return (EINVAL);					\
482 									\
483 	switch (*ptr) {							\
484 	case '-':							\
485 		minus = 1;						\
486 		zfs_fallthrough;					\
487 	case '+':							\
488 		++ptr;							\
489 		break;							\
490 	}								\
491 									\
492 	/* Auto-detect base based on prefix */				\
493 	if (!base) {							\
494 		if (str[0] == '0') {					\
495 			if (tolower(str[1]) == 'x' && isxdigit(str[2])) { \
496 				base = 16; /* hex */			\
497 				ptr += 2;				\
498 			} else if (str[1] >= '0' && str[1] < '8') {	\
499 				base = 8; /* octal */			\
500 				ptr += 1;				\
501 			} else {					\
502 				return (EINVAL);			\
503 			}						\
504 		} else {						\
505 			base = 10; /* decimal */			\
506 		}							\
507 	}								\
508 									\
509 	while (1) {							\
510 		if (isdigit(*ptr))					\
511 			digit = *ptr - '0';				\
512 		else if (isalpha(*ptr))					\
513 			digit = tolower(*ptr) - 'a' + 10;		\
514 		else							\
515 			break;						\
516 									\
517 		if (digit >= base)					\
518 			break;						\
519 									\
520 		last_value = value;					\
521 		value = value * base + digit;				\
522 		if (last_value > value) /* Overflow */			\
523 			return (ERANGE);				\
524 									\
525 		ptr++;							\
526 	}								\
527 									\
528 	*result = minus ? -value : value;				\
529 									\
530 	if (endptr)							\
531 		*endptr = ptr;						\
532 									\
533 	return (0);							\
534 }									\
535 
536 define_ddi_strtox(l, long)
537 define_ddi_strtox(ull, unsigned long long)
538 define_ddi_strtox(ll, long long)
539 
540 EXPORT_SYMBOL(ddi_strtol);
541 EXPORT_SYMBOL(ddi_strtoll);
542 EXPORT_SYMBOL(ddi_strtoull);
543 
544 int
545 ddi_copyin(const void *from, void *to, size_t len, int flags)
546 {
547 	/* Fake ioctl() issued by kernel, 'from' is a kernel address */
548 	if (flags & FKIOCTL) {
549 		memcpy(to, from, len);
550 		return (0);
551 	}
552 
553 	return (copyin(from, to, len));
554 }
555 EXPORT_SYMBOL(ddi_copyin);
556 
557 #define	define_spl_param(type, fmt)					\
558 int									\
559 spl_param_get_##type(char *buf, zfs_kernel_param_t *kp)			\
560 {									\
561 	return (scnprintf(buf, PAGE_SIZE, fmt "\n",			\
562 	    *(type *)kp->arg));						\
563 }									\
564 int									\
565 spl_param_set_##type(const char *buf, zfs_kernel_param_t *kp)		\
566 {									\
567 	return (kstrto##type(buf, 0, (type *)kp->arg));			\
568 }									\
569 const struct kernel_param_ops spl_param_ops_##type = {			\
570 	.set = spl_param_set_##type,					\
571 	.get = spl_param_get_##type,					\
572 };									\
573 EXPORT_SYMBOL(spl_param_get_##type);					\
574 EXPORT_SYMBOL(spl_param_set_##type);					\
575 EXPORT_SYMBOL(spl_param_ops_##type);
576 
577 define_spl_param(s64, "%lld")
578 define_spl_param(u64, "%llu")
579 
580 /*
581  * Post a uevent to userspace whenever a new vdev adds to the pool. It is
582  * necessary to sync blkid information with udev, which zed daemon uses
583  * during device hotplug to identify the vdev.
584  */
585 void
586 spl_signal_kobj_evt(struct block_device *bdev)
587 {
588 #if defined(HAVE_BDEV_KOBJ) || defined(HAVE_PART_TO_DEV)
589 #ifdef HAVE_BDEV_KOBJ
590 	struct kobject *disk_kobj = bdev_kobj(bdev);
591 #else
592 	struct kobject *disk_kobj = &part_to_dev(bdev->bd_part)->kobj;
593 #endif
594 	if (disk_kobj) {
595 		int ret = kobject_uevent(disk_kobj, KOBJ_CHANGE);
596 		if (ret) {
597 			pr_warn("ZFS: Sending event '%d' to kobject: '%s'"
598 			    " (%p): failed(ret:%d)\n", KOBJ_CHANGE,
599 			    kobject_name(disk_kobj), disk_kobj, ret);
600 		}
601 	}
602 #else
603 /*
604  * This is encountered if neither bdev_kobj() nor part_to_dev() is available
605  * in the kernel - likely due to an API change that needs to be chased down.
606  */
607 #error "Unsupported kernel: unable to get struct kobj from bdev"
608 #endif
609 }
610 EXPORT_SYMBOL(spl_signal_kobj_evt);
611 
612 int
613 ddi_copyout(const void *from, void *to, size_t len, int flags)
614 {
615 	/* Fake ioctl() issued by kernel, 'from' is a kernel address */
616 	if (flags & FKIOCTL) {
617 		memcpy(to, from, len);
618 		return (0);
619 	}
620 
621 	return (copyout(from, to, len));
622 }
623 EXPORT_SYMBOL(ddi_copyout);
624 
625 static int
626 spl_getattr(struct file *filp, struct kstat *stat)
627 {
628 	int rc;
629 
630 	ASSERT(filp);
631 	ASSERT(stat);
632 
633 	rc = vfs_getattr(&filp->f_path, stat, STATX_BASIC_STATS,
634 	    AT_STATX_SYNC_AS_STAT);
635 	if (rc)
636 		return (-rc);
637 
638 	return (0);
639 }
640 
641 /*
642  * Read the unique system identifier from the /etc/hostid file.
643  *
644  * The behavior of /usr/bin/hostid on Linux systems with the
645  * regular eglibc and coreutils is:
646  *
647  *   1. Generate the value if the /etc/hostid file does not exist
648  *      or if the /etc/hostid file is less than four bytes in size.
649  *
650  *   2. If the /etc/hostid file is at least 4 bytes, then return
651  *      the first four bytes [0..3] in native endian order.
652  *
653  *   3. Always ignore bytes [4..] if they exist in the file.
654  *
655  * Only the first four bytes are significant, even on systems that
656  * have a 64-bit word size.
657  *
658  * See:
659  *
660  *   eglibc: sysdeps/unix/sysv/linux/gethostid.c
661  *   coreutils: src/hostid.c
662  *
663  * Notes:
664  *
665  * The /etc/hostid file on Solaris is a text file that often reads:
666  *
667  *   # DO NOT EDIT
668  *   "0123456789"
669  *
670  * Directly copying this file to Linux results in a constant
671  * hostid of 4f442023 because the default comment constitutes
672  * the first four bytes of the file.
673  *
674  */
675 
676 static char *spl_hostid_path = HW_HOSTID_PATH;
677 module_param(spl_hostid_path, charp, 0444);
678 MODULE_PARM_DESC(spl_hostid_path, "The system hostid file (/etc/hostid)");
679 
680 static int
681 hostid_read(uint32_t *hostid)
682 {
683 	uint64_t size;
684 	uint32_t value = 0;
685 	int error;
686 	loff_t off;
687 	struct file *filp;
688 	struct kstat stat;
689 
690 	filp = filp_open(spl_hostid_path, 0, 0);
691 
692 	if (IS_ERR(filp))
693 		return (ENOENT);
694 
695 	error = spl_getattr(filp, &stat);
696 	if (error) {
697 		filp_close(filp, 0);
698 		return (error);
699 	}
700 	size = stat.size;
701 	// cppcheck-suppress sizeofwithnumericparameter
702 	if (size < sizeof (HW_HOSTID_MASK)) {
703 		filp_close(filp, 0);
704 		return (EINVAL);
705 	}
706 
707 	off = 0;
708 	/*
709 	 * Read directly into the variable like eglibc does.
710 	 * Short reads are okay; native behavior is preserved.
711 	 */
712 	error = kernel_read(filp, &value, sizeof (value), &off);
713 	if (error < 0) {
714 		filp_close(filp, 0);
715 		return (EIO);
716 	}
717 
718 	/* Mask down to 32 bits like coreutils does. */
719 	*hostid = (value & HW_HOSTID_MASK);
720 	filp_close(filp, 0);
721 
722 	return (0);
723 }
724 
725 /*
726  * Return the system hostid.  Preferentially use the spl_hostid module option
727  * when set, otherwise use the value in the /etc/hostid file.
728  */
729 uint32_t
730 zone_get_hostid(void *zone)
731 {
732 	uint32_t hostid;
733 
734 	ASSERT3P(zone, ==, NULL);
735 
736 	if (spl_hostid != 0)
737 		return ((uint32_t)(spl_hostid & HW_HOSTID_MASK));
738 
739 	if (hostid_read(&hostid) == 0)
740 		return (hostid);
741 
742 	return (0);
743 }
744 EXPORT_SYMBOL(zone_get_hostid);
745 
746 static int
747 spl_kvmem_init(void)
748 {
749 	int rc = 0;
750 
751 	rc = spl_kmem_init();
752 	if (rc)
753 		return (rc);
754 
755 	rc = spl_vmem_init();
756 	if (rc) {
757 		spl_kmem_fini();
758 		return (rc);
759 	}
760 
761 	return (rc);
762 }
763 
764 /*
765  * We initialize the random number generator with 128 bits of entropy from the
766  * system random number generator. In the improbable case that we have a zero
767  * seed, we fallback to the system jiffies, unless it is also zero, in which
768  * situation we use a preprogrammed seed. We step forward by 2^64 iterations to
769  * initialize each of the per-cpu seeds so that the sequences generated on each
770  * CPU are guaranteed to never overlap in practice.
771  */
772 static int __init
773 spl_random_init(void)
774 {
775 	uint64_t s[4];
776 	int i = 0;
777 
778 	spl_pseudo_entropy = __alloc_percpu(4 * sizeof (uint64_t),
779 	    sizeof (uint64_t));
780 
781 	if (!spl_pseudo_entropy)
782 		return (-ENOMEM);
783 
784 	get_random_bytes(s, sizeof (s));
785 
786 	if (s[0] == 0 && s[1] == 0 && s[2] == 0 && s[3] == 0) {
787 		if (jiffies != 0) {
788 			s[0] = jiffies;
789 			s[1] = ~0 - jiffies;
790 			s[2] = ~jiffies;
791 			s[3] = jiffies - ~0;
792 		} else {
793 			(void) memcpy(s, "improbable seed", 16);
794 		}
795 		printk("SPL: get_random_bytes() returned 0 "
796 		    "when generating random seed. Setting initial seed to "
797 		    "0x%016llx%016llx%016llx%016llx.\n", cpu_to_be64(s[0]),
798 		    cpu_to_be64(s[1]), cpu_to_be64(s[2]), cpu_to_be64(s[3]));
799 	}
800 
801 	for_each_possible_cpu(i) {
802 		uint64_t *wordp = per_cpu_ptr(spl_pseudo_entropy, i);
803 
804 		spl_rand_jump(s);
805 
806 		wordp[0] = s[0];
807 		wordp[1] = s[1];
808 		wordp[2] = s[2];
809 		wordp[3] = s[3];
810 	}
811 
812 	return (0);
813 }
814 
815 static void
816 spl_random_fini(void)
817 {
818 	free_percpu(spl_pseudo_entropy);
819 }
820 
821 static void
822 spl_kvmem_fini(void)
823 {
824 	spl_vmem_fini();
825 	spl_kmem_fini();
826 }
827 
828 static int __init
829 spl_init(void)
830 {
831 	int rc = 0;
832 
833 	if ((rc = spl_random_init()))
834 		goto out0;
835 
836 	if ((rc = spl_kvmem_init()))
837 		goto out1;
838 
839 	if ((rc = spl_tsd_init()))
840 		goto out2;
841 
842 	if ((rc = spl_proc_init()))
843 		goto out3;
844 
845 	if ((rc = spl_kstat_init()))
846 		goto out4;
847 
848 	if ((rc = spl_taskq_init()))
849 		goto out5;
850 
851 	if ((rc = spl_kmem_cache_init()))
852 		goto out6;
853 
854 	if ((rc = spl_zlib_init()))
855 		goto out7;
856 
857 	if ((rc = spl_zone_init()))
858 		goto out8;
859 
860 	return (rc);
861 
862 out8:
863 	spl_zlib_fini();
864 out7:
865 	spl_kmem_cache_fini();
866 out6:
867 	spl_taskq_fini();
868 out5:
869 	spl_kstat_fini();
870 out4:
871 	spl_proc_fini();
872 out3:
873 	spl_tsd_fini();
874 out2:
875 	spl_kvmem_fini();
876 out1:
877 	spl_random_fini();
878 out0:
879 	return (rc);
880 }
881 
882 static void __exit
883 spl_fini(void)
884 {
885 	spl_zone_fini();
886 	spl_zlib_fini();
887 	spl_kmem_cache_fini();
888 	spl_taskq_fini();
889 	spl_kstat_fini();
890 	spl_proc_fini();
891 	spl_tsd_fini();
892 	spl_kvmem_fini();
893 	spl_random_fini();
894 }
895 
896 module_init(spl_init);
897 module_exit(spl_fini);
898 
899 MODULE_DESCRIPTION("Solaris Porting Layer");
900 MODULE_AUTHOR(ZFS_META_AUTHOR);
901 MODULE_LICENSE("GPL");
902 MODULE_VERSION(ZFS_META_VERSION "-" ZFS_META_RELEASE);
903