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 <[email protected]>.
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/sysmacros.h>
27 #include <sys/systeminfo.h>
28 #include <sys/vmsystm.h>
29 #include <sys/kmem.h>
30 #include <sys/kmem_cache.h>
31 #include <sys/vmem.h>
32 #include <sys/mutex.h>
33 #include <sys/rwlock.h>
34 #include <sys/taskq.h>
35 #include <sys/tsd.h>
36 #include <sys/zmod.h>
37 #include <sys/debug.h>
38 #include <sys/proc.h>
39 #include <sys/kstat.h>
40 #include <sys/file.h>
41 #include <sys/sunddi.h>
42 #include <linux/ctype.h>
43 #include <sys/disp.h>
44 #include <sys/random.h>
45 #include <sys/strings.h>
46 #include <linux/kmod.h>
47 #include "zfs_gitrev.h"
48 #include <linux/mod_compat.h>
49 #include <sys/cred.h>
50 #include <sys/vnode.h>
51
52 char spl_gitrev[64] = ZFS_META_GITREV;
53
54 /* BEGIN CSTYLED */
55 unsigned long spl_hostid = 0;
56 EXPORT_SYMBOL(spl_hostid);
57 /* BEGIN CSTYLED */
58 module_param(spl_hostid, ulong, 0644);
59 MODULE_PARM_DESC(spl_hostid, "The system hostid.");
60 /* END CSTYLED */
61
62 proc_t p0;
63 EXPORT_SYMBOL(p0);
64
65 /*
66 * Xorshift Pseudo Random Number Generator based on work by Sebastiano Vigna
67 *
68 * "Further scramblings of Marsaglia's xorshift generators"
69 * http://vigna.di.unimi.it/ftp/papers/xorshiftplus.pdf
70 *
71 * random_get_pseudo_bytes() is an API function on Illumos whose sole purpose
72 * is to provide bytes containing random numbers. It is mapped to /dev/urandom
73 * on Illumos, which uses a "FIPS 186-2 algorithm". No user of the SPL's
74 * random_get_pseudo_bytes() needs bytes that are of cryptographic quality, so
75 * we can implement it using a fast PRNG that we seed using Linux' actual
76 * equivalent to random_get_pseudo_bytes(). We do this by providing each CPU
77 * with an independent seed so that all calls to random_get_pseudo_bytes() are
78 * free of atomic instructions.
79 *
80 * A consequence of using a fast PRNG is that using random_get_pseudo_bytes()
81 * to generate words larger than 128 bits will paradoxically be limited to
82 * `2^128 - 1` possibilities. This is because we have a sequence of `2^128 - 1`
83 * 128-bit words and selecting the first will implicitly select the second. If
84 * a caller finds this behavior undesirable, random_get_bytes() should be used
85 * instead.
86 *
87 * XXX: Linux interrupt handlers that trigger within the critical section
88 * formed by `s[1] = xp[1];` and `xp[0] = s[0];` and call this function will
89 * see the same numbers. Nothing in the code currently calls this in an
90 * interrupt handler, so this is considered to be okay. If that becomes a
91 * problem, we could create a set of per-cpu variables for interrupt handlers
92 * and use them when in_interrupt() from linux/preempt_mask.h evaluates to
93 * true.
94 */
95 void __percpu *spl_pseudo_entropy;
96
97 /*
98 * spl_rand_next()/spl_rand_jump() are copied from the following CC-0 licensed
99 * file:
100 *
101 * http://xorshift.di.unimi.it/xorshift128plus.c
102 */
103
104 static inline uint64_t
spl_rand_next(uint64_t * s)105 spl_rand_next(uint64_t *s)
106 {
107 uint64_t s1 = s[0];
108 const uint64_t s0 = s[1];
109 s[0] = s0;
110 s1 ^= s1 << 23; // a
111 s[1] = s1 ^ s0 ^ (s1 >> 18) ^ (s0 >> 5); // b, c
112 return (s[1] + s0);
113 }
114
115 static inline void
spl_rand_jump(uint64_t * s)116 spl_rand_jump(uint64_t *s)
117 {
118 static const uint64_t JUMP[] =
119 { 0x8a5cd789635d2dff, 0x121fd2155c472f96 };
120
121 uint64_t s0 = 0;
122 uint64_t s1 = 0;
123 int i, b;
124 for (i = 0; i < sizeof (JUMP) / sizeof (*JUMP); i++)
125 for (b = 0; b < 64; b++) {
126 if (JUMP[i] & 1ULL << b) {
127 s0 ^= s[0];
128 s1 ^= s[1];
129 }
130 (void) spl_rand_next(s);
131 }
132
133 s[0] = s0;
134 s[1] = s1;
135 }
136
137 int
random_get_pseudo_bytes(uint8_t * ptr,size_t len)138 random_get_pseudo_bytes(uint8_t *ptr, size_t len)
139 {
140 uint64_t *xp, s[2];
141
142 ASSERT(ptr);
143
144 xp = get_cpu_ptr(spl_pseudo_entropy);
145
146 s[0] = xp[0];
147 s[1] = xp[1];
148
149 while (len) {
150 union {
151 uint64_t ui64;
152 uint8_t byte[sizeof (uint64_t)];
153 }entropy;
154 int i = MIN(len, sizeof (uint64_t));
155
156 len -= i;
157 entropy.ui64 = spl_rand_next(s);
158
159 while (i--)
160 *ptr++ = entropy.byte[i];
161 }
162
163 xp[0] = s[0];
164 xp[1] = s[1];
165
166 put_cpu_ptr(spl_pseudo_entropy);
167
168 return (0);
169 }
170
171
172 EXPORT_SYMBOL(random_get_pseudo_bytes);
173
174 #if BITS_PER_LONG == 32
175
176 /*
177 * Support 64/64 => 64 division on a 32-bit platform. While the kernel
178 * provides a div64_u64() function for this we do not use it because the
179 * implementation is flawed. There are cases which return incorrect
180 * results as late as linux-2.6.35. Until this is fixed upstream the
181 * spl must provide its own implementation.
182 *
183 * This implementation is a slightly modified version of the algorithm
184 * proposed by the book 'Hacker's Delight'. The original source can be
185 * found here and is available for use without restriction.
186 *
187 * http://www.hackersdelight.org/HDcode/newCode/divDouble.c
188 */
189
190 /*
191 * Calculate number of leading of zeros for a 64-bit value.
192 */
193 static int
nlz64(uint64_t x)194 nlz64(uint64_t x)
195 {
196 register int n = 0;
197
198 if (x == 0)
199 return (64);
200
201 if (x <= 0x00000000FFFFFFFFULL) { n = n + 32; x = x << 32; }
202 if (x <= 0x0000FFFFFFFFFFFFULL) { n = n + 16; x = x << 16; }
203 if (x <= 0x00FFFFFFFFFFFFFFULL) { n = n + 8; x = x << 8; }
204 if (x <= 0x0FFFFFFFFFFFFFFFULL) { n = n + 4; x = x << 4; }
205 if (x <= 0x3FFFFFFFFFFFFFFFULL) { n = n + 2; x = x << 2; }
206 if (x <= 0x7FFFFFFFFFFFFFFFULL) { n = n + 1; }
207
208 return (n);
209 }
210
211 /*
212 * Newer kernels have a div_u64() function but we define our own
213 * to simplify portability between kernel versions.
214 */
215 static inline uint64_t
__div_u64(uint64_t u,uint32_t v)216 __div_u64(uint64_t u, uint32_t v)
217 {
218 (void) do_div(u, v);
219 return (u);
220 }
221
222 /*
223 * Turn off missing prototypes warning for these functions. They are
224 * replacements for libgcc-provided functions and will never be called
225 * directly.
226 */
227 #pragma GCC diagnostic push
228 #pragma GCC diagnostic ignored "-Wmissing-prototypes"
229
230 /*
231 * Implementation of 64-bit unsigned division for 32-bit machines.
232 *
233 * First the procedure takes care of the case in which the divisor is a
234 * 32-bit quantity. There are two subcases: (1) If the left half of the
235 * dividend is less than the divisor, one execution of do_div() is all that
236 * is required (overflow is not possible). (2) Otherwise it does two
237 * divisions, using the grade school method.
238 */
239 uint64_t
__udivdi3(uint64_t u,uint64_t v)240 __udivdi3(uint64_t u, uint64_t v)
241 {
242 uint64_t u0, u1, v1, q0, q1, k;
243 int n;
244
245 if (v >> 32 == 0) { // If v < 2**32:
246 if (u >> 32 < v) { // If u/v cannot overflow,
247 return (__div_u64(u, v)); // just do one division.
248 } else { // If u/v would overflow:
249 u1 = u >> 32; // Break u into two halves.
250 u0 = u & 0xFFFFFFFF;
251 q1 = __div_u64(u1, v); // First quotient digit.
252 k = u1 - q1 * v; // First remainder, < v.
253 u0 += (k << 32);
254 q0 = __div_u64(u0, v); // Seconds quotient digit.
255 return ((q1 << 32) + q0);
256 }
257 } else { // If v >= 2**32:
258 n = nlz64(v); // 0 <= n <= 31.
259 v1 = (v << n) >> 32; // Normalize divisor, MSB is 1.
260 u1 = u >> 1; // To ensure no overflow.
261 q1 = __div_u64(u1, v1); // Get quotient from
262 q0 = (q1 << n) >> 31; // Undo normalization and
263 // division of u by 2.
264 if (q0 != 0) // Make q0 correct or
265 q0 = q0 - 1; // too small by 1.
266 if ((u - q0 * v) >= v)
267 q0 = q0 + 1; // Now q0 is correct.
268
269 return (q0);
270 }
271 }
272 EXPORT_SYMBOL(__udivdi3);
273
274 /* BEGIN CSTYLED */
275 #ifndef abs64
276 #define abs64(x) ({ uint64_t t = (x) >> 63; ((x) ^ t) - t; })
277 #endif
278 /* END CSTYLED */
279
280 /*
281 * Implementation of 64-bit signed division for 32-bit machines.
282 */
283 int64_t
__divdi3(int64_t u,int64_t v)284 __divdi3(int64_t u, int64_t v)
285 {
286 int64_t q, t;
287 q = __udivdi3(abs64(u), abs64(v));
288 t = (u ^ v) >> 63; // If u, v have different
289 return ((q ^ t) - t); // signs, negate q.
290 }
291 EXPORT_SYMBOL(__divdi3);
292
293 /*
294 * Implementation of 64-bit unsigned modulo for 32-bit machines.
295 */
296 uint64_t
__umoddi3(uint64_t dividend,uint64_t divisor)297 __umoddi3(uint64_t dividend, uint64_t divisor)
298 {
299 return (dividend - (divisor * __udivdi3(dividend, divisor)));
300 }
301 EXPORT_SYMBOL(__umoddi3);
302
303 /* 64-bit signed modulo for 32-bit machines. */
304 int64_t
__moddi3(int64_t n,int64_t d)305 __moddi3(int64_t n, int64_t d)
306 {
307 int64_t q;
308 boolean_t nn = B_FALSE;
309
310 if (n < 0) {
311 nn = B_TRUE;
312 n = -n;
313 }
314 if (d < 0)
315 d = -d;
316
317 q = __umoddi3(n, d);
318
319 return (nn ? -q : q);
320 }
321 EXPORT_SYMBOL(__moddi3);
322
323 /*
324 * Implementation of 64-bit unsigned division/modulo for 32-bit machines.
325 */
326 uint64_t
__udivmoddi4(uint64_t n,uint64_t d,uint64_t * r)327 __udivmoddi4(uint64_t n, uint64_t d, uint64_t *r)
328 {
329 uint64_t q = __udivdi3(n, d);
330 if (r)
331 *r = n - d * q;
332 return (q);
333 }
334 EXPORT_SYMBOL(__udivmoddi4);
335
336 /*
337 * Implementation of 64-bit signed division/modulo for 32-bit machines.
338 */
339 int64_t
__divmoddi4(int64_t n,int64_t d,int64_t * r)340 __divmoddi4(int64_t n, int64_t d, int64_t *r)
341 {
342 int64_t q, rr;
343 boolean_t nn = B_FALSE;
344 boolean_t nd = B_FALSE;
345 if (n < 0) {
346 nn = B_TRUE;
347 n = -n;
348 }
349 if (d < 0) {
350 nd = B_TRUE;
351 d = -d;
352 }
353
354 q = __udivmoddi4(n, d, (uint64_t *)&rr);
355
356 if (nn != nd)
357 q = -q;
358 if (nn)
359 rr = -rr;
360 if (r)
361 *r = rr;
362 return (q);
363 }
364 EXPORT_SYMBOL(__divmoddi4);
365
366 #if defined(__arm) || defined(__arm__)
367 /*
368 * Implementation of 64-bit (un)signed division for 32-bit arm machines.
369 *
370 * Run-time ABI for the ARM Architecture (page 20). A pair of (unsigned)
371 * long longs is returned in {{r0, r1}, {r2,r3}}, the quotient in {r0, r1},
372 * and the remainder in {r2, r3}. The return type is specifically left
373 * set to 'void' to ensure the compiler does not overwrite these registers
374 * during the return. All results are in registers as per ABI
375 */
376 void
__aeabi_uldivmod(uint64_t u,uint64_t v)377 __aeabi_uldivmod(uint64_t u, uint64_t v)
378 {
379 uint64_t res;
380 uint64_t mod;
381
382 res = __udivdi3(u, v);
383 mod = __umoddi3(u, v);
384 {
385 register uint32_t r0 asm("r0") = (res & 0xFFFFFFFF);
386 register uint32_t r1 asm("r1") = (res >> 32);
387 register uint32_t r2 asm("r2") = (mod & 0xFFFFFFFF);
388 register uint32_t r3 asm("r3") = (mod >> 32);
389
390 /* BEGIN CSTYLED */
391 asm volatile(""
392 : "+r"(r0), "+r"(r1), "+r"(r2),"+r"(r3) /* output */
393 : "r"(r0), "r"(r1), "r"(r2), "r"(r3)); /* input */
394 /* END CSTYLED */
395
396 return; /* r0; */
397 }
398 }
399 EXPORT_SYMBOL(__aeabi_uldivmod);
400
401 void
__aeabi_ldivmod(int64_t u,int64_t v)402 __aeabi_ldivmod(int64_t u, int64_t v)
403 {
404 int64_t res;
405 uint64_t mod;
406
407 res = __divdi3(u, v);
408 mod = __umoddi3(u, v);
409 {
410 register uint32_t r0 asm("r0") = (res & 0xFFFFFFFF);
411 register uint32_t r1 asm("r1") = (res >> 32);
412 register uint32_t r2 asm("r2") = (mod & 0xFFFFFFFF);
413 register uint32_t r3 asm("r3") = (mod >> 32);
414
415 /* BEGIN CSTYLED */
416 asm volatile(""
417 : "+r"(r0), "+r"(r1), "+r"(r2),"+r"(r3) /* output */
418 : "r"(r0), "r"(r1), "r"(r2), "r"(r3)); /* input */
419 /* END CSTYLED */
420
421 return; /* r0; */
422 }
423 }
424 EXPORT_SYMBOL(__aeabi_ldivmod);
425 #endif /* __arm || __arm__ */
426
427 #pragma GCC diagnostic pop
428
429 #endif /* BITS_PER_LONG */
430
431 /*
432 * NOTE: The strtoxx behavior is solely based on my reading of the Solaris
433 * ddi_strtol(9F) man page. I have not verified the behavior of these
434 * functions against their Solaris counterparts. It is possible that I
435 * may have misinterpreted the man page or the man page is incorrect.
436 */
437 int ddi_strtoul(const char *, char **, int, unsigned long *);
438 int ddi_strtol(const char *, char **, int, long *);
439 int ddi_strtoull(const char *, char **, int, unsigned long long *);
440 int ddi_strtoll(const char *, char **, int, long long *);
441
442 #define define_ddi_strtoux(type, valtype) \
443 int ddi_strtou##type(const char *str, char **endptr, \
444 int base, valtype *result) \
445 { \
446 valtype last_value, value = 0; \
447 char *ptr = (char *)str; \
448 int flag = 1, digit; \
449 \
450 if (strlen(ptr) == 0) \
451 return (EINVAL); \
452 \
453 /* Auto-detect base based on prefix */ \
454 if (!base) { \
455 if (str[0] == '0') { \
456 if (tolower(str[1]) == 'x' && isxdigit(str[2])) { \
457 base = 16; /* hex */ \
458 ptr += 2; \
459 } else if (str[1] >= '0' && str[1] < 8) { \
460 base = 8; /* octal */ \
461 ptr += 1; \
462 } else { \
463 return (EINVAL); \
464 } \
465 } else { \
466 base = 10; /* decimal */ \
467 } \
468 } \
469 \
470 while (1) { \
471 if (isdigit(*ptr)) \
472 digit = *ptr - '0'; \
473 else if (isalpha(*ptr)) \
474 digit = tolower(*ptr) - 'a' + 10; \
475 else \
476 break; \
477 \
478 if (digit >= base) \
479 break; \
480 \
481 last_value = value; \
482 value = value * base + digit; \
483 if (last_value > value) /* Overflow */ \
484 return (ERANGE); \
485 \
486 flag = 1; \
487 ptr++; \
488 } \
489 \
490 if (flag) \
491 *result = value; \
492 \
493 if (endptr) \
494 *endptr = (char *)(flag ? ptr : str); \
495 \
496 return (0); \
497 } \
498
499 #define define_ddi_strtox(type, valtype) \
500 int ddi_strto##type(const char *str, char **endptr, \
501 int base, valtype *result) \
502 { \
503 int rc; \
504 \
505 if (*str == '-') { \
506 rc = ddi_strtou##type(str + 1, endptr, base, result); \
507 if (!rc) { \
508 if (*endptr == str + 1) \
509 *endptr = (char *)str; \
510 else \
511 *result = -*result; \
512 } \
513 } else { \
514 rc = ddi_strtou##type(str, endptr, base, result); \
515 } \
516 \
517 return (rc); \
518 }
519
520 define_ddi_strtoux(l, unsigned long)
521 define_ddi_strtox(l, long)
522 define_ddi_strtoux(ll, unsigned long long)
523 define_ddi_strtox(ll, long long)
524
525 EXPORT_SYMBOL(ddi_strtoul);
526 EXPORT_SYMBOL(ddi_strtol);
527 EXPORT_SYMBOL(ddi_strtoll);
528 EXPORT_SYMBOL(ddi_strtoull);
529
530 int
ddi_copyin(const void * from,void * to,size_t len,int flags)531 ddi_copyin(const void *from, void *to, size_t len, int flags)
532 {
533 /* Fake ioctl() issued by kernel, 'from' is a kernel address */
534 if (flags & FKIOCTL) {
535 memcpy(to, from, len);
536 return (0);
537 }
538
539 return (copyin(from, to, len));
540 }
541 EXPORT_SYMBOL(ddi_copyin);
542
543 int
ddi_copyout(const void * from,void * to,size_t len,int flags)544 ddi_copyout(const void *from, void *to, size_t len, int flags)
545 {
546 /* Fake ioctl() issued by kernel, 'from' is a kernel address */
547 if (flags & FKIOCTL) {
548 memcpy(to, from, len);
549 return (0);
550 }
551
552 return (copyout(from, to, len));
553 }
554 EXPORT_SYMBOL(ddi_copyout);
555
556 static ssize_t
spl_kernel_read(struct file * file,void * buf,size_t count,loff_t * pos)557 spl_kernel_read(struct file *file, void *buf, size_t count, loff_t *pos)
558 {
559 #if defined(HAVE_KERNEL_READ_PPOS)
560 return (kernel_read(file, buf, count, pos));
561 #else
562 mm_segment_t saved_fs;
563 ssize_t ret;
564
565 saved_fs = get_fs();
566 set_fs(KERNEL_DS);
567
568 ret = vfs_read(file, (void __user *)buf, count, pos);
569
570 set_fs(saved_fs);
571
572 return (ret);
573 #endif
574 }
575
576 static int
spl_getattr(struct file * filp,struct kstat * stat)577 spl_getattr(struct file *filp, struct kstat *stat)
578 {
579 int rc;
580
581 ASSERT(filp);
582 ASSERT(stat);
583
584 #if defined(HAVE_4ARGS_VFS_GETATTR)
585 rc = vfs_getattr(&filp->f_path, stat, STATX_BASIC_STATS,
586 AT_STATX_SYNC_AS_STAT);
587 #elif defined(HAVE_2ARGS_VFS_GETATTR)
588 rc = vfs_getattr(&filp->f_path, stat);
589 #elif defined(HAVE_3ARGS_VFS_GETATTR)
590 rc = vfs_getattr(filp->f_path.mnt, filp->f_dentry, stat);
591 #else
592 #error "No available vfs_getattr()"
593 #endif
594 if (rc)
595 return (-rc);
596
597 return (0);
598 }
599
600 /*
601 * Read the unique system identifier from the /etc/hostid file.
602 *
603 * The behavior of /usr/bin/hostid on Linux systems with the
604 * regular eglibc and coreutils is:
605 *
606 * 1. Generate the value if the /etc/hostid file does not exist
607 * or if the /etc/hostid file is less than four bytes in size.
608 *
609 * 2. If the /etc/hostid file is at least 4 bytes, then return
610 * the first four bytes [0..3] in native endian order.
611 *
612 * 3. Always ignore bytes [4..] if they exist in the file.
613 *
614 * Only the first four bytes are significant, even on systems that
615 * have a 64-bit word size.
616 *
617 * See:
618 *
619 * eglibc: sysdeps/unix/sysv/linux/gethostid.c
620 * coreutils: src/hostid.c
621 *
622 * Notes:
623 *
624 * The /etc/hostid file on Solaris is a text file that often reads:
625 *
626 * # DO NOT EDIT
627 * "0123456789"
628 *
629 * Directly copying this file to Linux results in a constant
630 * hostid of 4f442023 because the default comment constitutes
631 * the first four bytes of the file.
632 *
633 */
634
635 char *spl_hostid_path = HW_HOSTID_PATH;
636 module_param(spl_hostid_path, charp, 0444);
637 MODULE_PARM_DESC(spl_hostid_path, "The system hostid file (/etc/hostid)");
638
639 static int
hostid_read(uint32_t * hostid)640 hostid_read(uint32_t *hostid)
641 {
642 uint64_t size;
643 uint32_t value = 0;
644 int error;
645 loff_t off;
646 struct file *filp;
647 struct kstat stat;
648
649 filp = filp_open(spl_hostid_path, 0, 0);
650
651 if (IS_ERR(filp))
652 return (ENOENT);
653
654 error = spl_getattr(filp, &stat);
655 if (error) {
656 filp_close(filp, 0);
657 return (error);
658 }
659 size = stat.size;
660 // cppcheck-suppress sizeofwithnumericparameter
661 if (size < sizeof (HW_HOSTID_MASK)) {
662 filp_close(filp, 0);
663 return (EINVAL);
664 }
665
666 off = 0;
667 /*
668 * Read directly into the variable like eglibc does.
669 * Short reads are okay; native behavior is preserved.
670 */
671 error = spl_kernel_read(filp, &value, sizeof (value), &off);
672 if (error < 0) {
673 filp_close(filp, 0);
674 return (EIO);
675 }
676
677 /* Mask down to 32 bits like coreutils does. */
678 *hostid = (value & HW_HOSTID_MASK);
679 filp_close(filp, 0);
680
681 return (0);
682 }
683
684 /*
685 * Return the system hostid. Preferentially use the spl_hostid module option
686 * when set, otherwise use the value in the /etc/hostid file.
687 */
688 uint32_t
zone_get_hostid(void * zone)689 zone_get_hostid(void *zone)
690 {
691 uint32_t hostid;
692
693 ASSERT3P(zone, ==, NULL);
694
695 if (spl_hostid != 0)
696 return ((uint32_t)(spl_hostid & HW_HOSTID_MASK));
697
698 if (hostid_read(&hostid) == 0)
699 return (hostid);
700
701 return (0);
702 }
703 EXPORT_SYMBOL(zone_get_hostid);
704
705 static int
spl_kvmem_init(void)706 spl_kvmem_init(void)
707 {
708 int rc = 0;
709
710 rc = spl_kmem_init();
711 if (rc)
712 return (rc);
713
714 rc = spl_vmem_init();
715 if (rc) {
716 spl_kmem_fini();
717 return (rc);
718 }
719
720 return (rc);
721 }
722
723 /*
724 * We initialize the random number generator with 128 bits of entropy from the
725 * system random number generator. In the improbable case that we have a zero
726 * seed, we fallback to the system jiffies, unless it is also zero, in which
727 * situation we use a preprogrammed seed. We step forward by 2^64 iterations to
728 * initialize each of the per-cpu seeds so that the sequences generated on each
729 * CPU are guaranteed to never overlap in practice.
730 */
731 static void __init
spl_random_init(void)732 spl_random_init(void)
733 {
734 uint64_t s[2];
735 int i = 0;
736
737 spl_pseudo_entropy = __alloc_percpu(2 * sizeof (uint64_t),
738 sizeof (uint64_t));
739
740 get_random_bytes(s, sizeof (s));
741
742 if (s[0] == 0 && s[1] == 0) {
743 if (jiffies != 0) {
744 s[0] = jiffies;
745 s[1] = ~0 - jiffies;
746 } else {
747 (void) memcpy(s, "improbable seed", sizeof (s));
748 }
749 printk("SPL: get_random_bytes() returned 0 "
750 "when generating random seed. Setting initial seed to "
751 "0x%016llx%016llx.\n", cpu_to_be64(s[0]),
752 cpu_to_be64(s[1]));
753 }
754
755 for_each_possible_cpu(i) {
756 uint64_t *wordp = per_cpu_ptr(spl_pseudo_entropy, i);
757
758 spl_rand_jump(s);
759
760 wordp[0] = s[0];
761 wordp[1] = s[1];
762 }
763 }
764
765 static void
spl_random_fini(void)766 spl_random_fini(void)
767 {
768 free_percpu(spl_pseudo_entropy);
769 }
770
771 static void
spl_kvmem_fini(void)772 spl_kvmem_fini(void)
773 {
774 spl_vmem_fini();
775 spl_kmem_fini();
776 }
777
778 static int __init
spl_init(void)779 spl_init(void)
780 {
781 int rc = 0;
782
783 bzero(&p0, sizeof (proc_t));
784 spl_random_init();
785
786 if ((rc = spl_kvmem_init()))
787 goto out1;
788
789 if ((rc = spl_tsd_init()))
790 goto out2;
791
792 if ((rc = spl_taskq_init()))
793 goto out3;
794
795 if ((rc = spl_kmem_cache_init()))
796 goto out4;
797
798 if ((rc = spl_proc_init()))
799 goto out5;
800
801 if ((rc = spl_kstat_init()))
802 goto out6;
803
804 if ((rc = spl_zlib_init()))
805 goto out7;
806
807 return (rc);
808
809 out7:
810 spl_kstat_fini();
811 out6:
812 spl_proc_fini();
813 out5:
814 spl_kmem_cache_fini();
815 out4:
816 spl_taskq_fini();
817 out3:
818 spl_tsd_fini();
819 out2:
820 spl_kvmem_fini();
821 out1:
822 return (rc);
823 }
824
825 static void __exit
spl_fini(void)826 spl_fini(void)
827 {
828 spl_zlib_fini();
829 spl_kstat_fini();
830 spl_proc_fini();
831 spl_kmem_cache_fini();
832 spl_taskq_fini();
833 spl_tsd_fini();
834 spl_kvmem_fini();
835 spl_random_fini();
836 }
837
838 module_init(spl_init);
839 module_exit(spl_fini);
840
841 ZFS_MODULE_DESCRIPTION("Solaris Porting Layer");
842 ZFS_MODULE_AUTHOR(ZFS_META_AUTHOR);
843 ZFS_MODULE_LICENSE("GPL");
844 ZFS_MODULE_VERSION(ZFS_META_VERSION "-" ZFS_META_RELEASE);
845