xref: /dflybsd-src/sys/kern/kern_memio.c (revision 70e491c09b1cb18f96facdbc91052080f2a455c3)
1 /*-
2  * Copyright (c) 1988 University of Utah.
3  * Copyright (c) 1982, 1986, 1990 The Regents of the University of California.
4  * All rights reserved.
5  *
6  * This code is derived from software contributed to Berkeley by
7  * the Systems Programming Group of the University of Utah Computer
8  * Science Department, and code derived from software contributed to
9  * Berkeley by William Jolitz.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 3. All advertising materials mentioning features or use of this software
20  *    must display the following acknowledgement:
21  *	This product includes software developed by the University of
22  *	California, Berkeley and its contributors.
23  * 4. Neither the name of the University nor the names of its contributors
24  *    may be used to endorse or promote products derived from this software
25  *    without specific prior written permission.
26  *
27  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
28  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
29  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
30  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
31  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
32  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
33  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
34  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
35  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
36  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
37  * SUCH DAMAGE.
38  *
39  *	from: Utah $Hdr: mem.c 1.13 89/10/08$
40  *	from: @(#)mem.c	7.2 (Berkeley) 5/9/91
41  * $FreeBSD: src/sys/i386/i386/mem.c,v 1.79.2.9 2003/01/04 22:58:01 njl Exp $
42  * $DragonFly: src/sys/kern/kern_memio.c,v 1.32 2008/07/23 16:39:28 dillon Exp $
43  */
44 
45 /*
46  * Memory special file
47  */
48 
49 #include <sys/param.h>
50 #include <sys/systm.h>
51 #include <sys/buf.h>
52 #include <sys/conf.h>
53 #include <sys/fcntl.h>
54 #include <sys/filio.h>
55 #include <sys/kernel.h>
56 #include <sys/malloc.h>
57 #include <sys/memrange.h>
58 #include <sys/proc.h>
59 #include <sys/priv.h>
60 #include <sys/random.h>
61 #include <sys/signalvar.h>
62 #include <sys/signal2.h>
63 #include <sys/uio.h>
64 #include <sys/vnode.h>
65 
66 #include <vm/vm.h>
67 #include <vm/pmap.h>
68 #include <vm/vm_extern.h>
69 
70 
71 static	d_open_t	mmopen;
72 static	d_close_t	mmclose;
73 static	d_read_t	mmread;
74 static	d_write_t	mmwrite;
75 static	d_ioctl_t	mmioctl;
76 static	d_mmap_t	memmmap;
77 static	d_poll_t	mmpoll;
78 static	d_kqfilter_t	mmkqfilter;
79 
80 #define CDEV_MAJOR 2
81 static struct dev_ops mem_ops = {
82 	{ "mem", CDEV_MAJOR, D_MEM | D_MPSAFE_READ | D_MPSAFE_WRITE | D_KQFILTER },
83 	.d_open =	mmopen,
84 	.d_close =	mmclose,
85 	.d_read =	mmread,
86 	.d_write =	mmwrite,
87 	.d_ioctl =	mmioctl,
88 	.d_poll =	mmpoll,
89 	.d_kqfilter =	mmkqfilter,
90 	.d_mmap =	memmmap,
91 };
92 
93 static int rand_bolt;
94 static caddr_t	zbuf;
95 static cdev_t	zerodev = NULL;
96 
97 MALLOC_DEFINE(M_MEMDESC, "memdesc", "memory range descriptors");
98 static int mem_ioctl (cdev_t, u_long, caddr_t, int, struct ucred *);
99 static int random_ioctl (cdev_t, u_long, caddr_t, int, struct ucred *);
100 
101 struct mem_range_softc mem_range_softc;
102 
103 
104 static int
105 mmopen(struct dev_open_args *ap)
106 {
107 	cdev_t dev = ap->a_head.a_dev;
108 	int error;
109 
110 	switch (minor(dev)) {
111 	case 0:
112 	case 1:
113 		if (ap->a_oflags & FWRITE) {
114 			if (securelevel > 0 || kernel_mem_readonly)
115 				return (EPERM);
116 		}
117 		error = 0;
118 		break;
119 	case 14:
120 		error = priv_check_cred(ap->a_cred, PRIV_ROOT, 0);
121 		if (error != 0)
122 			break;
123 		if (securelevel > 0 || kernel_mem_readonly) {
124 			error = EPERM;
125 			break;
126 		}
127 		error = cpu_set_iopl();
128 		break;
129 	default:
130 		error = 0;
131 		break;
132 	}
133 	return (error);
134 }
135 
136 static int
137 mmclose(struct dev_close_args *ap)
138 {
139 	cdev_t dev = ap->a_head.a_dev;
140 	int error;
141 
142 	switch (minor(dev)) {
143 	case 14:
144 		error = cpu_clr_iopl();
145 		break;
146 	default:
147 		error = 0;
148 		break;
149 	}
150 	return (error);
151 }
152 
153 
154 static int
155 mmrw(cdev_t dev, struct uio *uio, int flags)
156 {
157 	int o;
158 	u_int c, v;
159 	u_int poolsize;
160 	struct iovec *iov;
161 	int error = 0;
162 	caddr_t buf = NULL;
163 
164 	while (uio->uio_resid > 0 && error == 0) {
165 		iov = uio->uio_iov;
166 		if (iov->iov_len == 0) {
167 			uio->uio_iov++;
168 			uio->uio_iovcnt--;
169 			if (uio->uio_iovcnt < 0)
170 				panic("mmrw");
171 			continue;
172 		}
173 		switch (minor(dev)) {
174 		case 0:
175 			/*
176 			 * minor device 0 is physical memory, /dev/mem
177 			 */
178 			v = uio->uio_offset;
179 			v &= ~PAGE_MASK;
180 			pmap_kenter((vm_offset_t)ptvmmap, v);
181 			o = (int)uio->uio_offset & PAGE_MASK;
182 			c = (u_int)(PAGE_SIZE - ((uintptr_t)iov->iov_base & PAGE_MASK));
183 			c = min(c, (u_int)(PAGE_SIZE - o));
184 			c = min(c, (u_int)iov->iov_len);
185 			error = uiomove((caddr_t)&ptvmmap[o], (int)c, uio);
186 			pmap_kremove((vm_offset_t)ptvmmap);
187 			continue;
188 
189 		case 1: {
190 			/*
191 			 * minor device 1 is kernel memory, /dev/kmem
192 			 */
193 			vm_offset_t saddr, eaddr;
194 			int prot;
195 
196 			c = iov->iov_len;
197 
198 			/*
199 			 * Make sure that all of the pages are currently
200 			 * resident so that we don't create any zero-fill
201 			 * pages.
202 			 */
203 			saddr = trunc_page(uio->uio_offset);
204 			eaddr = round_page(uio->uio_offset + c);
205 			if (saddr > eaddr)
206 				return EFAULT;
207 
208 			/*
209 			 * Make sure the kernel addresses are mapped.
210 			 * platform_direct_mapped() can be used to bypass
211 			 * default mapping via the page table (virtual kernels
212 			 * contain a lot of out-of-band data).
213 			 */
214 			prot = VM_PROT_READ;
215 			if (uio->uio_rw != UIO_READ)
216 				prot |= VM_PROT_WRITE;
217 			error = kvm_access_check(saddr, eaddr, prot);
218 			if (error)
219 				return (error);
220 			error = uiomove((caddr_t)(vm_offset_t)uio->uio_offset,
221 					(int)c, uio);
222 			continue;
223 		}
224 		case 2:
225 			/*
226 			 * minor device 2 is EOF/RATHOLE
227 			 */
228 			if (uio->uio_rw == UIO_READ)
229 				return (0);
230 			c = iov->iov_len;
231 			break;
232 		case 3:
233 			/*
234 			 * minor device 3 (/dev/random) is source of filth
235 			 * on read, seeder on write
236 			 */
237 			if (buf == NULL)
238 				buf = kmalloc(PAGE_SIZE, M_TEMP, M_WAITOK);
239 			c = min(iov->iov_len, PAGE_SIZE);
240 			if (uio->uio_rw == UIO_WRITE) {
241 				error = uiomove(buf, (int)c, uio);
242 				if (error == 0)
243 					error = add_buffer_randomness(buf, c);
244 			} else {
245 				poolsize = read_random(buf, c);
246 				if (poolsize == 0) {
247 					if (buf)
248 						kfree(buf, M_TEMP);
249 					if ((flags & IO_NDELAY) != 0)
250 						return (EWOULDBLOCK);
251 					return (0);
252 				}
253 				c = min(c, poolsize);
254 				error = uiomove(buf, (int)c, uio);
255 			}
256 			continue;
257 		case 4:
258 			/*
259 			 * minor device 4 (/dev/urandom) is source of muck
260 			 * on read, writes are disallowed.
261 			 */
262 			c = min(iov->iov_len, PAGE_SIZE);
263 			if (uio->uio_rw == UIO_WRITE) {
264 				error = EPERM;
265 				break;
266 			}
267 			if (CURSIG(curthread->td_lwp) != 0) {
268 				/*
269 				 * Use tsleep() to get the error code right.
270 				 * It should return immediately.
271 				 */
272 				error = tsleep(&rand_bolt, PCATCH, "urand", 1);
273 				if (error != 0 && error != EWOULDBLOCK)
274 					continue;
275 			}
276 			if (buf == NULL)
277 				buf = kmalloc(PAGE_SIZE, M_TEMP, M_WAITOK);
278 			poolsize = read_random_unlimited(buf, c);
279 			c = min(c, poolsize);
280 			error = uiomove(buf, (int)c, uio);
281 			continue;
282 		case 12:
283 			/*
284 			 * minor device 12 (/dev/zero) is source of nulls
285 			 * on read, write are disallowed.
286 			 */
287 			if (uio->uio_rw == UIO_WRITE) {
288 				c = iov->iov_len;
289 				break;
290 			}
291 			if (zbuf == NULL) {
292 				zbuf = (caddr_t)kmalloc(PAGE_SIZE, M_TEMP,
293 				    M_WAITOK | M_ZERO);
294 			}
295 			c = min(iov->iov_len, PAGE_SIZE);
296 			error = uiomove(zbuf, (int)c, uio);
297 			continue;
298 		default:
299 			return (ENODEV);
300 		}
301 		if (error)
302 			break;
303 		iov->iov_base = (char *)iov->iov_base + c;
304 		iov->iov_len -= c;
305 		uio->uio_offset += c;
306 		uio->uio_resid -= c;
307 	}
308 	if (buf)
309 		kfree(buf, M_TEMP);
310 	return (error);
311 }
312 
313 static int
314 mmread(struct dev_read_args *ap)
315 {
316 	return(mmrw(ap->a_head.a_dev, ap->a_uio, ap->a_ioflag));
317 }
318 
319 static int
320 mmwrite(struct dev_write_args *ap)
321 {
322 	return(mmrw(ap->a_head.a_dev, ap->a_uio, ap->a_ioflag));
323 }
324 
325 
326 
327 
328 
329 /*******************************************************\
330 * allow user processes to MMAP some memory sections	*
331 * instead of going through read/write			*
332 \*******************************************************/
333 
334 static int
335 memmmap(struct dev_mmap_args *ap)
336 {
337 	cdev_t dev = ap->a_head.a_dev;
338 
339 	switch (minor(dev)) {
340 	case 0:
341 		/*
342 		 * minor device 0 is physical memory
343 		 */
344 #if defined(__i386__)
345         	ap->a_result = i386_btop(ap->a_offset);
346 #elif defined(__x86_64__)
347 		ap->a_result = x86_64_btop(ap->a_offset);
348 #endif
349 		return 0;
350 	case 1:
351 		/*
352 		 * minor device 1 is kernel memory
353 		 */
354 #if defined(__i386__)
355         	ap->a_result = i386_btop(vtophys(ap->a_offset));
356 #elif defined(__x86_64__)
357         	ap->a_result = x86_64_btop(vtophys(ap->a_offset));
358 #endif
359 		return 0;
360 
361 	default:
362 		return EINVAL;
363 	}
364 }
365 
366 static int
367 mmioctl(struct dev_ioctl_args *ap)
368 {
369 	cdev_t dev = ap->a_head.a_dev;
370 
371 	switch (minor(dev)) {
372 	case 0:
373 		return mem_ioctl(dev, ap->a_cmd, ap->a_data,
374 				 ap->a_fflag, ap->a_cred);
375 	case 3:
376 	case 4:
377 		return random_ioctl(dev, ap->a_cmd, ap->a_data,
378 				    ap->a_fflag, ap->a_cred);
379 	}
380 	return (ENODEV);
381 }
382 
383 /*
384  * Operations for changing memory attributes.
385  *
386  * This is basically just an ioctl shim for mem_range_attr_get
387  * and mem_range_attr_set.
388  */
389 static int
390 mem_ioctl(cdev_t dev, u_long cmd, caddr_t data, int flags, struct ucred *cred)
391 {
392 	int nd, error = 0;
393 	struct mem_range_op *mo = (struct mem_range_op *)data;
394 	struct mem_range_desc *md;
395 
396 	/* is this for us? */
397 	if ((cmd != MEMRANGE_GET) &&
398 	    (cmd != MEMRANGE_SET))
399 		return (ENOTTY);
400 
401 	/* any chance we can handle this? */
402 	if (mem_range_softc.mr_op == NULL)
403 		return (EOPNOTSUPP);
404 
405 	/* do we have any descriptors? */
406 	if (mem_range_softc.mr_ndesc == 0)
407 		return (ENXIO);
408 
409 	switch (cmd) {
410 	case MEMRANGE_GET:
411 		nd = imin(mo->mo_arg[0], mem_range_softc.mr_ndesc);
412 		if (nd > 0) {
413 			md = (struct mem_range_desc *)
414 				kmalloc(nd * sizeof(struct mem_range_desc),
415 				       M_MEMDESC, M_WAITOK);
416 			error = mem_range_attr_get(md, &nd);
417 			if (!error)
418 				error = copyout(md, mo->mo_desc,
419 					nd * sizeof(struct mem_range_desc));
420 			kfree(md, M_MEMDESC);
421 		} else {
422 			nd = mem_range_softc.mr_ndesc;
423 		}
424 		mo->mo_arg[0] = nd;
425 		break;
426 
427 	case MEMRANGE_SET:
428 		md = (struct mem_range_desc *)kmalloc(sizeof(struct mem_range_desc),
429 						    M_MEMDESC, M_WAITOK);
430 		error = copyin(mo->mo_desc, md, sizeof(struct mem_range_desc));
431 		/* clamp description string */
432 		md->mr_owner[sizeof(md->mr_owner) - 1] = 0;
433 		if (error == 0)
434 			error = mem_range_attr_set(md, &mo->mo_arg[0]);
435 		kfree(md, M_MEMDESC);
436 		break;
437 	}
438 	return (error);
439 }
440 
441 /*
442  * Implementation-neutral, kernel-callable functions for manipulating
443  * memory range attributes.
444  */
445 int
446 mem_range_attr_get(struct mem_range_desc *mrd, int *arg)
447 {
448 	/* can we handle this? */
449 	if (mem_range_softc.mr_op == NULL)
450 		return (EOPNOTSUPP);
451 
452 	if (*arg == 0) {
453 		*arg = mem_range_softc.mr_ndesc;
454 	} else {
455 		bcopy(mem_range_softc.mr_desc, mrd, (*arg) * sizeof(struct mem_range_desc));
456 	}
457 	return (0);
458 }
459 
460 int
461 mem_range_attr_set(struct mem_range_desc *mrd, int *arg)
462 {
463 	/* can we handle this? */
464 	if (mem_range_softc.mr_op == NULL)
465 		return (EOPNOTSUPP);
466 
467 	return (mem_range_softc.mr_op->set(&mem_range_softc, mrd, arg));
468 }
469 
470 #ifdef SMP
471 void
472 mem_range_AP_init(void)
473 {
474 	if (mem_range_softc.mr_op && mem_range_softc.mr_op->initAP)
475 		return (mem_range_softc.mr_op->initAP(&mem_range_softc));
476 }
477 #endif
478 
479 static int
480 random_ioctl(cdev_t dev, u_long cmd, caddr_t data, int flags, struct ucred *cred)
481 {
482 	int error;
483 	int intr;
484 
485 	/*
486 	 * Even inspecting the state is privileged, since it gives a hint
487 	 * about how easily the randomness might be guessed.
488 	 */
489 	error = 0;
490 
491 	switch (cmd) {
492 	/* Really handled in upper layer */
493 	case FIOASYNC:
494 		break;
495 	case MEM_SETIRQ:
496 		intr = *(int16_t *)data;
497 		if ((error = priv_check_cred(cred, PRIV_ROOT, 0)) != 0)
498 			break;
499 		if (intr < 0 || intr >= MAX_INTS)
500 			return (EINVAL);
501 		register_randintr(intr);
502 		break;
503 	case MEM_CLEARIRQ:
504 		intr = *(int16_t *)data;
505 		if ((error = priv_check_cred(cred, PRIV_ROOT, 0)) != 0)
506 			break;
507 		if (intr < 0 || intr >= MAX_INTS)
508 			return (EINVAL);
509 		unregister_randintr(intr);
510 		break;
511 	case MEM_RETURNIRQ:
512 		error = ENOTSUP;
513 		break;
514 	case MEM_FINDIRQ:
515 		intr = *(int16_t *)data;
516 		if ((error = priv_check_cred(cred, PRIV_ROOT, 0)) != 0)
517 			break;
518 		if (intr < 0 || intr >= MAX_INTS)
519 			return (EINVAL);
520 		intr = next_registered_randintr(intr);
521 		if (intr == MAX_INTS)
522 			return (ENOENT);
523 		*(u_int16_t *)data = intr;
524 		break;
525 	default:
526 		error = ENOTSUP;
527 		break;
528 	}
529 	return (error);
530 }
531 
532 int
533 mmpoll(struct dev_poll_args *ap)
534 {
535 	cdev_t dev = ap->a_head.a_dev;
536 	int revents;
537 
538 	switch (minor(dev)) {
539 	case 3:		/* /dev/random */
540 		revents = random_poll(dev, ap->a_events);
541 		break;
542 	case 4:		/* /dev/urandom */
543 	default:
544 		revents = seltrue(dev, ap->a_events);
545 		break;
546 	}
547 	ap->a_events = revents;
548 	return (0);
549 }
550 
551 static int
552 mm_filter_read(struct knote *kn, long hint)
553 {
554 	return (1);
555 }
556 
557 static void
558 dummy_filter_detach(struct knote *kn) {}
559 
560 static struct filterops random_read_filtops =
561         { 1, NULL, dummy_filter_detach, random_filter_read };
562 
563 static struct filterops mm_read_filtops =
564         { 1, NULL, dummy_filter_detach, mm_filter_read };
565 
566 int
567 mmkqfilter(struct dev_kqfilter_args *ap)
568 {
569 	struct knote *kn = ap->a_kn;
570 	cdev_t dev = ap->a_head.a_dev;
571 
572 	ap->a_result = 0;
573 	switch (kn->kn_filter) {
574 	case EVFILT_READ:
575 		switch (minor(dev)) {
576 		case 3:
577 			kn->kn_fop = &random_read_filtops;
578 			break;
579 		default:
580 			kn->kn_fop = &mm_read_filtops;
581 			break;
582 		}
583 		break;
584 	default:
585 		ap->a_result = 1;
586 		return (0);
587 	}
588 
589 	return (0);
590 }
591 
592 int
593 iszerodev(cdev_t dev)
594 {
595 	return (zerodev == dev);
596 }
597 
598 static void
599 mem_drvinit(void *unused)
600 {
601 
602 	/* Initialise memory range handling */
603 	if (mem_range_softc.mr_op != NULL)
604 		mem_range_softc.mr_op->init(&mem_range_softc);
605 
606 	make_dev(&mem_ops, 0, UID_ROOT, GID_KMEM, 0640, "mem");
607 	make_dev(&mem_ops, 1, UID_ROOT, GID_KMEM, 0640, "kmem");
608 	make_dev(&mem_ops, 2, UID_ROOT, GID_WHEEL, 0666, "null");
609 	make_dev(&mem_ops, 3, UID_ROOT, GID_WHEEL, 0644, "random");
610 	make_dev(&mem_ops, 4, UID_ROOT, GID_WHEEL, 0644, "urandom");
611 	zerodev = make_dev(&mem_ops, 12, UID_ROOT, GID_WHEEL, 0666, "zero");
612 	make_dev(&mem_ops, 14, UID_ROOT, GID_WHEEL, 0600, "io");
613 }
614 
615 SYSINIT(memdev,SI_SUB_DRIVERS,SI_ORDER_MIDDLE+CDEV_MAJOR,mem_drvinit,NULL)
616 
617