subr_copy.c revision 1.13 1 /* $NetBSD: subr_copy.c,v 1.13 2020/03/14 18:08:39 ad Exp $ */
2
3 /*-
4 * Copyright (c) 1997, 1998, 1999, 2002, 2007, 2008, 2019
5 * The NetBSD Foundation, Inc.
6 * All rights reserved.
7 *
8 * This code is derived from software contributed to The NetBSD Foundation
9 * by Jason R. Thorpe of the Numerical Aerospace Simulation Facility,
10 * NASA Ames Research Center.
11 *
12 * Redistribution and use in source and binary forms, with or without
13 * modification, are permitted provided that the following conditions
14 * are met:
15 * 1. Redistributions of source code must retain the above copyright
16 * notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 * notice, this list of conditions and the following disclaimer in the
19 * documentation and/or other materials provided with the distribution.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
22 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
23 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
24 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
25 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
26 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
27 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
28 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
29 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
30 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
31 * POSSIBILITY OF SUCH DAMAGE.
32 */
33
34 /*
35 * Copyright (c) 1982, 1986, 1991, 1993
36 * The Regents of the University of California. All rights reserved.
37 * (c) UNIX System Laboratories, Inc.
38 * All or some portions of this file are derived from material licensed
39 * to the University of California by American Telephone and Telegraph
40 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
41 * the permission of UNIX System Laboratories, Inc.
42 *
43 * Copyright (c) 1992, 1993
44 * The Regents of the University of California. All rights reserved.
45 *
46 * This software was developed by the Computer Systems Engineering group
47 * at Lawrence Berkeley Laboratory under DARPA contract BG 91-66 and
48 * contributed to Berkeley.
49 *
50 * All advertising materials mentioning features or use of this software
51 * must display the following acknowledgement:
52 * This product includes software developed by the University of
53 * California, Lawrence Berkeley Laboratory.
54 *
55 * Redistribution and use in source and binary forms, with or without
56 * modification, are permitted provided that the following conditions
57 * are met:
58 * 1. Redistributions of source code must retain the above copyright
59 * notice, this list of conditions and the following disclaimer.
60 * 2. Redistributions in binary form must reproduce the above copyright
61 * notice, this list of conditions and the following disclaimer in the
62 * documentation and/or other materials provided with the distribution.
63 * 3. Neither the name of the University nor the names of its contributors
64 * may be used to endorse or promote products derived from this software
65 * without specific prior written permission.
66 *
67 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
68 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
69 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
70 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
71 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
72 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
73 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
74 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
75 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
76 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
77 * SUCH DAMAGE.
78 *
79 * @(#)kern_subr.c 8.4 (Berkeley) 2/14/95
80 */
81
82 #include <sys/cdefs.h>
83 __KERNEL_RCSID(0, "$NetBSD: subr_copy.c,v 1.13 2020/03/14 18:08:39 ad Exp $");
84
85 #define __UFETCHSTORE_PRIVATE
86 #define __UCAS_PRIVATE
87
88 #include <sys/param.h>
89 #include <sys/fcntl.h>
90 #include <sys/proc.h>
91 #include <sys/systm.h>
92
93 #include <uvm/uvm_extern.h>
94
95 void
96 uio_setup_sysspace(struct uio *uio)
97 {
98
99 uio->uio_vmspace = vmspace_kernel();
100 }
101
102 int
103 uiomove(void *buf, size_t n, struct uio *uio)
104 {
105 struct vmspace *vm = uio->uio_vmspace;
106 struct iovec *iov;
107 size_t cnt;
108 int error = 0;
109 char *cp = buf;
110
111 ASSERT_SLEEPABLE();
112
113 KASSERT(uio->uio_rw == UIO_READ || uio->uio_rw == UIO_WRITE);
114 while (n > 0 && uio->uio_resid) {
115 iov = uio->uio_iov;
116 cnt = iov->iov_len;
117 if (cnt == 0) {
118 KASSERT(uio->uio_iovcnt > 0);
119 uio->uio_iov++;
120 uio->uio_iovcnt--;
121 continue;
122 }
123 if (cnt > n)
124 cnt = n;
125 if (!VMSPACE_IS_KERNEL_P(vm)) {
126 preempt_point();
127 }
128
129 if (uio->uio_rw == UIO_READ) {
130 error = copyout_vmspace(vm, cp, iov->iov_base,
131 cnt);
132 } else {
133 error = copyin_vmspace(vm, iov->iov_base, cp,
134 cnt);
135 }
136 if (error) {
137 break;
138 }
139 iov->iov_base = (char *)iov->iov_base + cnt;
140 iov->iov_len -= cnt;
141 uio->uio_resid -= cnt;
142 uio->uio_offset += cnt;
143 cp += cnt;
144 KDASSERT(cnt <= n);
145 n -= cnt;
146 }
147
148 return (error);
149 }
150
151 /*
152 * Wrapper for uiomove() that validates the arguments against a known-good
153 * kernel buffer.
154 */
155 int
156 uiomove_frombuf(void *buf, size_t buflen, struct uio *uio)
157 {
158 size_t offset;
159
160 if (uio->uio_offset < 0 || /* uio->uio_resid < 0 || */
161 (offset = uio->uio_offset) != uio->uio_offset)
162 return (EINVAL);
163 if (offset >= buflen)
164 return (0);
165 return (uiomove((char *)buf + offset, buflen - offset, uio));
166 }
167
168 /*
169 * Give next character to user as result of read.
170 */
171 int
172 ureadc(int c, struct uio *uio)
173 {
174 struct iovec *iov;
175
176 if (uio->uio_resid <= 0)
177 panic("ureadc: non-positive resid");
178 again:
179 if (uio->uio_iovcnt <= 0)
180 panic("ureadc: non-positive iovcnt");
181 iov = uio->uio_iov;
182 if (iov->iov_len <= 0) {
183 uio->uio_iovcnt--;
184 uio->uio_iov++;
185 goto again;
186 }
187 if (!VMSPACE_IS_KERNEL_P(uio->uio_vmspace)) {
188 int error;
189 if ((error = ustore_char(iov->iov_base, c)) != 0)
190 return (error);
191 } else {
192 *(char *)iov->iov_base = c;
193 }
194 iov->iov_base = (char *)iov->iov_base + 1;
195 iov->iov_len--;
196 uio->uio_resid--;
197 uio->uio_offset++;
198 return (0);
199 }
200
201 /*
202 * Like copyin(), but operates on an arbitrary vmspace.
203 */
204 int
205 copyin_vmspace(struct vmspace *vm, const void *uaddr, void *kaddr, size_t len)
206 {
207 struct iovec iov;
208 struct uio uio;
209 int error;
210
211 if (len == 0)
212 return (0);
213
214 if (VMSPACE_IS_KERNEL_P(vm)) {
215 return kcopy(uaddr, kaddr, len);
216 }
217 if (__predict_true(vm == curproc->p_vmspace)) {
218 return copyin(uaddr, kaddr, len);
219 }
220
221 iov.iov_base = kaddr;
222 iov.iov_len = len;
223 uio.uio_iov = &iov;
224 uio.uio_iovcnt = 1;
225 uio.uio_offset = (off_t)(uintptr_t)uaddr;
226 uio.uio_resid = len;
227 uio.uio_rw = UIO_READ;
228 UIO_SETUP_SYSSPACE(&uio);
229 error = uvm_io(&vm->vm_map, &uio, 0);
230
231 return (error);
232 }
233
234 /*
235 * Like copyout(), but operates on an arbitrary vmspace.
236 */
237 int
238 copyout_vmspace(struct vmspace *vm, const void *kaddr, void *uaddr, size_t len)
239 {
240 struct iovec iov;
241 struct uio uio;
242 int error;
243
244 if (len == 0)
245 return (0);
246
247 if (VMSPACE_IS_KERNEL_P(vm)) {
248 return kcopy(kaddr, uaddr, len);
249 }
250 if (__predict_true(vm == curproc->p_vmspace)) {
251 return copyout(kaddr, uaddr, len);
252 }
253
254 iov.iov_base = __UNCONST(kaddr); /* XXXUNCONST cast away const */
255 iov.iov_len = len;
256 uio.uio_iov = &iov;
257 uio.uio_iovcnt = 1;
258 uio.uio_offset = (off_t)(uintptr_t)uaddr;
259 uio.uio_resid = len;
260 uio.uio_rw = UIO_WRITE;
261 UIO_SETUP_SYSSPACE(&uio);
262 error = uvm_io(&vm->vm_map, &uio, 0);
263
264 return (error);
265 }
266
267 /*
268 * Like copyin(), but operates on an arbitrary process.
269 */
270 int
271 copyin_proc(struct proc *p, const void *uaddr, void *kaddr, size_t len)
272 {
273 struct vmspace *vm;
274 int error;
275
276 error = proc_vmspace_getref(p, &vm);
277 if (error) {
278 return error;
279 }
280 error = copyin_vmspace(vm, uaddr, kaddr, len);
281 uvmspace_free(vm);
282
283 return error;
284 }
285
286 /*
287 * Like copyout(), but operates on an arbitrary process.
288 */
289 int
290 copyout_proc(struct proc *p, const void *kaddr, void *uaddr, size_t len)
291 {
292 struct vmspace *vm;
293 int error;
294
295 error = proc_vmspace_getref(p, &vm);
296 if (error) {
297 return error;
298 }
299 error = copyout_vmspace(vm, kaddr, uaddr, len);
300 uvmspace_free(vm);
301
302 return error;
303 }
304
305 /*
306 * Like copyin(), but operates on an arbitrary pid.
307 */
308 int
309 copyin_pid(pid_t pid, const void *uaddr, void *kaddr, size_t len)
310 {
311 struct proc *p;
312 struct vmspace *vm;
313 int error;
314
315 mutex_enter(proc_lock);
316 p = proc_find(pid);
317 if (p == NULL) {
318 mutex_exit(proc_lock);
319 return ESRCH;
320 }
321 mutex_enter(p->p_lock);
322 error = proc_vmspace_getref(p, &vm);
323 mutex_exit(p->p_lock);
324 mutex_exit(proc_lock);
325
326 if (error == 0) {
327 error = copyin_vmspace(vm, uaddr, kaddr, len);
328 uvmspace_free(vm);
329 }
330 return error;
331 }
332
333 /*
334 * Like copyin(), except it operates on kernel addresses when the FKIOCTL
335 * flag is passed in `ioctlflags' from the ioctl call.
336 */
337 int
338 ioctl_copyin(int ioctlflags, const void *src, void *dst, size_t len)
339 {
340 if (ioctlflags & FKIOCTL)
341 return kcopy(src, dst, len);
342 return copyin(src, dst, len);
343 }
344
345 /*
346 * Like copyout(), except it operates on kernel addresses when the FKIOCTL
347 * flag is passed in `ioctlflags' from the ioctl call.
348 */
349 int
350 ioctl_copyout(int ioctlflags, const void *src, void *dst, size_t len)
351 {
352 if (ioctlflags & FKIOCTL)
353 return kcopy(src, dst, len);
354 return copyout(src, dst, len);
355 }
356
357 /*
358 * User-space CAS / fetch / store
359 */
360
361 #ifdef __NO_STRICT_ALIGNMENT
362 #define CHECK_ALIGNMENT(x) __nothing
363 #else /* ! __NO_STRICT_ALIGNMENT */
364 static bool
365 ufetchstore_aligned(uintptr_t uaddr, size_t size)
366 {
367 return (uaddr & (size - 1)) == 0;
368 }
369
370 #define CHECK_ALIGNMENT() \
371 do { \
372 if (!ufetchstore_aligned((uintptr_t)uaddr, sizeof(*uaddr))) \
373 return EFAULT; \
374 } while (/*CONSTCOND*/0)
375 #endif /* __NO_STRICT_ALIGNMENT */
376
377 /*
378 * __HAVE_UCAS_FULL platforms provide _ucas_32() and _ucas_64() themselves.
379 * _RUMPKERNEL also provides it's own _ucas_32() and _ucas_64().
380 *
381 * In all other cases, we provide generic implementations that work on
382 * all platforms.
383 */
384
385 #if !defined(__HAVE_UCAS_FULL) && !defined(_RUMPKERNEL)
386 #if !defined(__HAVE_UCAS_MP) && defined(MULTIPROCESSOR)
387 #include <sys/atomic.h>
388 #include <sys/cpu.h>
389 #include <sys/once.h>
390 #include <sys/mutex.h>
391 #include <sys/ipi.h>
392
393 static int ucas_critical_splcookie;
394 static volatile u_int ucas_critical_pausing_cpus;
395 static u_int ucas_critical_ipi;
396 static ONCE_DECL(ucas_critical_init_once)
397
398 static void
399 ucas_critical_cpu_gate(void *arg __unused)
400 {
401 int count = SPINLOCK_BACKOFF_MIN;
402
403 KASSERT(ucas_critical_pausing_cpus > 0);
404 atomic_dec_uint(&ucas_critical_pausing_cpus);
405 while (ucas_critical_pausing_cpus != (u_int)-1) {
406 SPINLOCK_BACKOFF(count);
407 }
408 }
409
410 static int
411 ucas_critical_init(void)
412 {
413 ucas_critical_ipi = ipi_register(ucas_critical_cpu_gate, NULL);
414 return 0;
415 }
416
417 static void
418 ucas_critical_wait(void)
419 {
420 int count = SPINLOCK_BACKOFF_MIN;
421
422 while (ucas_critical_pausing_cpus > 0) {
423 SPINLOCK_BACKOFF(count);
424 }
425 }
426 #endif /* ! __HAVE_UCAS_MP && MULTIPROCESSOR */
427
428 static inline void
429 ucas_critical_enter(lwp_t * const l)
430 {
431
432 #if !defined(__HAVE_UCAS_MP) && defined(MULTIPROCESSOR)
433 if (ncpu > 1) {
434 RUN_ONCE(&ucas_critical_init_once, ucas_critical_init);
435
436 /*
437 * Acquire the mutex first, then go to splhigh() and
438 * broadcast the IPI to lock all of the other CPUs
439 * behind the gate.
440 *
441 * N.B. Going to splhigh() implicitly disables preemption,
442 * so there's no need to do it explicitly.
443 */
444 mutex_enter(&cpu_lock);
445 ucas_critical_splcookie = splhigh();
446 ucas_critical_pausing_cpus = ncpu - 1;
447 membar_enter();
448
449 ipi_trigger_broadcast(ucas_critical_ipi, true);
450 ucas_critical_wait();
451 return;
452 }
453 #endif /* ! __HAVE_UCAS_MP && MULTIPROCESSOR */
454
455 KPREEMPT_DISABLE(l);
456 }
457
458 static inline void
459 ucas_critical_exit(lwp_t * const l)
460 {
461
462 #if !defined(__HAVE_UCAS_MP) && defined(MULTIPROCESSOR)
463 if (ncpu > 1) {
464 membar_exit();
465 ucas_critical_pausing_cpus = (u_int)-1;
466 splx(ucas_critical_splcookie);
467 mutex_exit(&cpu_lock);
468 return;
469 }
470 #endif /* ! __HAVE_UCAS_MP && MULTIPROCESSOR */
471
472 KPREEMPT_ENABLE(l);
473 }
474
475 int
476 _ucas_32(volatile uint32_t *uaddr, uint32_t old, uint32_t new, uint32_t *ret)
477 {
478 lwp_t * const l = curlwp;
479 uint32_t *uva = ((void *)(uintptr_t)uaddr);
480 int error;
481
482 /*
483 * Wire the user address down to avoid taking a page fault during
484 * the critical section.
485 */
486 error = uvm_vslock(l->l_proc->p_vmspace, uva, sizeof(*uaddr),
487 VM_PROT_READ | VM_PROT_WRITE);
488 if (error)
489 return error;
490
491 ucas_critical_enter(l);
492 error = _ufetch_32(uva, ret);
493 if (error == 0 && *ret == old) {
494 error = _ustore_32(uva, new);
495 }
496 ucas_critical_exit(l);
497
498 uvm_vsunlock(l->l_proc->p_vmspace, uva, sizeof(*uaddr));
499
500 return error;
501 }
502
503 #ifdef _LP64
504 int
505 _ucas_64(volatile uint64_t *uaddr, uint64_t old, uint64_t new, uint64_t *ret)
506 {
507 lwp_t * const l = curlwp;
508 uint64_t *uva = ((void *)(uintptr_t)uaddr);
509 int error;
510
511 /*
512 * Wire the user address down to avoid taking a page fault during
513 * the critical section.
514 */
515 error = uvm_vslock(l->l_proc->p_vmspace, uva, sizeof(*uaddr),
516 VM_PROT_READ | VM_PROT_WRITE);
517 if (error)
518 return error;
519
520 ucas_critical_enter(l);
521 error = _ufetch_64(uva, ret);
522 if (error == 0 && *ret == old) {
523 error = _ustore_64(uva, new);
524 }
525 ucas_critical_exit(l);
526
527 uvm_vsunlock(l->l_proc->p_vmspace, uva, sizeof(*uaddr));
528
529 return error;
530 }
531 #endif /* _LP64 */
532 #endif /* ! __HAVE_UCAS_FULL && ! _RUMPKERNEL */
533
534 int
535 ucas_32(volatile uint32_t *uaddr, uint32_t old, uint32_t new, uint32_t *ret)
536 {
537
538 ASSERT_SLEEPABLE();
539 CHECK_ALIGNMENT();
540 #if (defined(__HAVE_UCAS_MP) && defined(MULTIPROCESSOR)) && \
541 !defined(_RUMPKERNEL)
542 if (ncpu > 1) {
543 return _ucas_32_mp(uaddr, old, new, ret);
544 }
545 #endif /* __HAVE_UCAS_MP && MULTIPROCESSOR */
546 return _ucas_32(uaddr, old, new, ret);
547 }
548
549 #ifdef _LP64
550 int
551 ucas_64(volatile uint64_t *uaddr, uint64_t old, uint64_t new, uint64_t *ret)
552 {
553
554 ASSERT_SLEEPABLE();
555 CHECK_ALIGNMENT();
556 #if (defined(__HAVE_UCAS_MP) && defined(MULTIPROCESSOR)) && \
557 !defined(_RUMPKERNEL)
558 if (ncpu > 1) {
559 return _ucas_64_mp(uaddr, old, new, ret);
560 }
561 #endif /* __HAVE_UCAS_MP && MULTIPROCESSOR */
562 return _ucas_64(uaddr, old, new, ret);
563 }
564 #endif /* _LP64 */
565
566 __strong_alias(ucas_int,ucas_32);
567 #ifdef _LP64
568 __strong_alias(ucas_ptr,ucas_64);
569 #else
570 __strong_alias(ucas_ptr,ucas_32);
571 #endif /* _LP64 */
572
573 int
574 ufetch_8(const uint8_t *uaddr, uint8_t *valp)
575 {
576
577 ASSERT_SLEEPABLE();
578 CHECK_ALIGNMENT();
579 return _ufetch_8(uaddr, valp);
580 }
581
582 int
583 ufetch_16(const uint16_t *uaddr, uint16_t *valp)
584 {
585
586 ASSERT_SLEEPABLE();
587 CHECK_ALIGNMENT();
588 return _ufetch_16(uaddr, valp);
589 }
590
591 int
592 ufetch_32(const uint32_t *uaddr, uint32_t *valp)
593 {
594
595 ASSERT_SLEEPABLE();
596 CHECK_ALIGNMENT();
597 return _ufetch_32(uaddr, valp);
598 }
599
600 #ifdef _LP64
601 int
602 ufetch_64(const uint64_t *uaddr, uint64_t *valp)
603 {
604
605 ASSERT_SLEEPABLE();
606 CHECK_ALIGNMENT();
607 return _ufetch_64(uaddr, valp);
608 }
609 #endif /* _LP64 */
610
611 __strong_alias(ufetch_char,ufetch_8);
612 __strong_alias(ufetch_short,ufetch_16);
613 __strong_alias(ufetch_int,ufetch_32);
614 #ifdef _LP64
615 __strong_alias(ufetch_long,ufetch_64);
616 __strong_alias(ufetch_ptr,ufetch_64);
617 #else
618 __strong_alias(ufetch_long,ufetch_32);
619 __strong_alias(ufetch_ptr,ufetch_32);
620 #endif /* _LP64 */
621
622 int
623 ustore_8(uint8_t *uaddr, uint8_t val)
624 {
625
626 ASSERT_SLEEPABLE();
627 CHECK_ALIGNMENT();
628 return _ustore_8(uaddr, val);
629 }
630
631 int
632 ustore_16(uint16_t *uaddr, uint16_t val)
633 {
634
635 ASSERT_SLEEPABLE();
636 CHECK_ALIGNMENT();
637 return _ustore_16(uaddr, val);
638 }
639
640 int
641 ustore_32(uint32_t *uaddr, uint32_t val)
642 {
643
644 ASSERT_SLEEPABLE();
645 CHECK_ALIGNMENT();
646 return _ustore_32(uaddr, val);
647 }
648
649 #ifdef _LP64
650 int
651 ustore_64(uint64_t *uaddr, uint64_t val)
652 {
653
654 ASSERT_SLEEPABLE();
655 CHECK_ALIGNMENT();
656 return _ustore_64(uaddr, val);
657 }
658 #endif /* _LP64 */
659
660 __strong_alias(ustore_char,ustore_8);
661 __strong_alias(ustore_short,ustore_16);
662 __strong_alias(ustore_int,ustore_32);
663 #ifdef _LP64
664 __strong_alias(ustore_long,ustore_64);
665 __strong_alias(ustore_ptr,ustore_64);
666 #else
667 __strong_alias(ustore_long,ustore_32);
668 __strong_alias(ustore_ptr,ustore_32);
669 #endif /* _LP64 */
670