xref: /freebsd/sys/amd64/amd64/mem.c (revision 4cf49a43559ed9fdad601bdcccd2c55963008675)
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  * $FreeBSD$
42  */
43 
44 /*
45  * Memory special file
46  */
47 
48 #include <sys/param.h>
49 #include <sys/systm.h>
50 #include <sys/conf.h>
51 #include <sys/buf.h>
52 #include <sys/kernel.h>
53 #include <sys/uio.h>
54 #include <sys/ioccom.h>
55 #include <sys/malloc.h>
56 #include <sys/memrange.h>
57 #include <sys/proc.h>
58 #include <sys/signalvar.h>
59 
60 #include <machine/frame.h>
61 #include <machine/random.h>
62 #include <machine/psl.h>
63 #include <machine/specialreg.h>
64 #include <i386/isa/intr_machdep.h>
65 
66 #include <vm/vm.h>
67 #include <vm/vm_prot.h>
68 #include <vm/pmap.h>
69 #include <vm/vm_extern.h>
70 
71 
72 static	d_open_t	mmopen;
73 static	d_close_t	mmclose;
74 static	d_read_t	mmrw;
75 static	d_ioctl_t	mmioctl;
76 static	d_mmap_t	memmmap;
77 static	d_poll_t	mmpoll;
78 
79 #define CDEV_MAJOR 2
80 static struct cdevsw mem_cdevsw = {
81 	/* open */	mmopen,
82 	/* close */	mmclose,
83 	/* read */	mmrw,
84 	/* write */	mmrw,
85 	/* ioctl */	mmioctl,
86 	/* poll */	mmpoll,
87 	/* mmap */	memmmap,
88 	/* strategy */	nostrategy,
89 	/* name */	"mem",
90 	/* maj */	CDEV_MAJOR,
91 	/* dump */	nodump,
92 	/* psize */	nopsize,
93 	/* flags */	D_MEM,
94 	/* bmaj */	-1
95 };
96 
97 static struct random_softc random_softc[16];
98 static caddr_t	zbuf;
99 
100 MALLOC_DEFINE(M_MEMDESC, "memdesc", "memory range descriptors");
101 static int mem_ioctl __P((dev_t, u_long, caddr_t, int, struct proc *));
102 static int random_ioctl __P((dev_t, u_long, caddr_t, int, struct proc *));
103 
104 struct mem_range_softc mem_range_softc;
105 
106 
107 static int
108 mmclose(dev, flags, fmt, p)
109 	dev_t dev;
110 	int flags;
111 	int fmt;
112 	struct proc *p;
113 {
114 	switch (minor(dev)) {
115 	case 14:
116 		curproc->p_md.md_regs->tf_eflags &= ~PSL_IOPL;
117 		break;
118 	default:
119 		break;
120 	}
121 	return (0);
122 }
123 
124 static int
125 mmopen(dev, flags, fmt, p)
126 	dev_t dev;
127 	int flags;
128 	int fmt;
129 	struct proc *p;
130 {
131 	int error;
132 
133 	switch (minor(dev)) {
134 	case 14:
135 		error = suser(p);
136 		if (error != 0)
137 			return (error);
138 		if (securelevel > 0)
139 			return (EPERM);
140 		curproc->p_md.md_regs->tf_eflags |= PSL_IOPL;
141 		break;
142 	default:
143 		break;
144 	}
145 	return (0);
146 }
147 
148 static int
149 mmrw(dev, uio, flags)
150 	dev_t dev;
151 	struct uio *uio;
152 	int flags;
153 {
154 	register int o;
155 	register u_int c, v;
156 	u_int poolsize;
157 	register struct iovec *iov;
158 	int error = 0;
159 	caddr_t buf = NULL;
160 
161 	while (uio->uio_resid > 0 && error == 0) {
162 		iov = uio->uio_iov;
163 		if (iov->iov_len == 0) {
164 			uio->uio_iov++;
165 			uio->uio_iovcnt--;
166 			if (uio->uio_iovcnt < 0)
167 				panic("mmrw");
168 			continue;
169 		}
170 		switch (minor(dev)) {
171 
172 /* minor device 0 is physical memory */
173 		case 0:
174 			v = uio->uio_offset;
175 			pmap_enter(kernel_pmap, (vm_offset_t)ptvmmap, v,
176 				uio->uio_rw == UIO_READ ? VM_PROT_READ : VM_PROT_WRITE,
177 				TRUE);
178 			o = (int)uio->uio_offset & PAGE_MASK;
179 			c = (u_int)(PAGE_SIZE - ((int)iov->iov_base & PAGE_MASK));
180 			c = min(c, (u_int)(PAGE_SIZE - o));
181 			c = min(c, (u_int)iov->iov_len);
182 			error = uiomove((caddr_t)&ptvmmap[o], (int)c, uio);
183 			pmap_remove(kernel_pmap, (vm_offset_t)ptvmmap,
184 				    (vm_offset_t)&ptvmmap[PAGE_SIZE]);
185 			continue;
186 
187 /* minor device 1 is kernel memory */
188 		case 1: {
189 			vm_offset_t addr, eaddr;
190 			c = iov->iov_len;
191 
192 			/*
193 			 * Make sure that all of the pages are currently resident so
194 			 * that we don't create any zero-fill pages.
195 			 */
196 			addr = trunc_page(uio->uio_offset);
197 			eaddr = round_page(uio->uio_offset + c);
198 
199 			if (addr < (vm_offset_t)VADDR(PTDPTDI, 0))
200 				return EFAULT;
201 			if (eaddr >= (vm_offset_t)VADDR(APTDPTDI, 0))
202 				return EFAULT;
203 			for (; addr < eaddr; addr += PAGE_SIZE)
204 				if (pmap_extract(kernel_pmap, addr) == 0)
205 					return EFAULT;
206 
207 			if (!kernacc((caddr_t)(int)uio->uio_offset, c,
208 			    uio->uio_rw == UIO_READ ? B_READ : B_WRITE))
209 				return (EFAULT);
210 			error = uiomove((caddr_t)(int)uio->uio_offset, (int)c, uio);
211 			continue;
212 		}
213 
214 /* minor device 2 is EOF/RATHOLE */
215 		case 2:
216 			if (uio->uio_rw == UIO_READ)
217 				return (0);
218 			c = iov->iov_len;
219 			break;
220 
221 /* minor device 3 (/dev/random) is source of filth on read, rathole on write */
222 		case 3:
223 			if (uio->uio_rw == UIO_WRITE) {
224 				c = iov->iov_len;
225 				break;
226 			}
227 			if (buf == NULL)
228 				buf = (caddr_t)
229 				    malloc(PAGE_SIZE, M_TEMP, M_WAITOK);
230 			c = min(iov->iov_len, PAGE_SIZE);
231 			poolsize = read_random(buf, c);
232 			if (poolsize == 0) {
233 				if (buf)
234 					free(buf, M_TEMP);
235 				return (0);
236 			}
237 			c = min(c, poolsize);
238 			error = uiomove(buf, (int)c, uio);
239 			continue;
240 
241 /* minor device 4 (/dev/urandom) is source of muck on read, rathole on write */
242 		case 4:
243 			if (uio->uio_rw == UIO_WRITE) {
244 				c = iov->iov_len;
245 				break;
246 			}
247 			if (CURSIG(curproc) != 0) {
248 				/*
249 				 * Use tsleep() to get the error code right.
250 				 * It should return immediately.
251 				 */
252 				error = tsleep(&random_softc[0],
253 				    PZERO | PCATCH, "urand", 1);
254 				if (error != 0 && error != EWOULDBLOCK)
255 					continue;
256 			}
257 			if (buf == NULL)
258 				buf = (caddr_t)
259 				    malloc(PAGE_SIZE, M_TEMP, M_WAITOK);
260 			c = min(iov->iov_len, PAGE_SIZE);
261 			poolsize = read_random_unlimited(buf, c);
262 			c = min(c, poolsize);
263 			error = uiomove(buf, (int)c, uio);
264 			continue;
265 
266 /* minor device 12 (/dev/zero) is source of nulls on read, rathole on write */
267 		case 12:
268 			if (uio->uio_rw == UIO_WRITE) {
269 				c = iov->iov_len;
270 				break;
271 			}
272 			if (zbuf == NULL) {
273 				zbuf = (caddr_t)
274 				    malloc(PAGE_SIZE, M_TEMP, M_WAITOK);
275 				bzero(zbuf, PAGE_SIZE);
276 			}
277 			c = min(iov->iov_len, PAGE_SIZE);
278 			error = uiomove(zbuf, (int)c, uio);
279 			continue;
280 
281 #ifdef notyet
282 /* 386 I/O address space (/dev/ioport[bwl]) is a read/write access to seperate
283    i/o device address bus, different than memory bus. Semantics here are
284    very different than ordinary read/write, as if iov_len is a multiple
285    an implied string move from a single port will be done. Note that lseek
286    must be used to set the port number reliably. */
287 		case 14:
288 			if (iov->iov_len == 1) {
289 				u_char tmp;
290 				tmp = inb(uio->uio_offset);
291 				error = uiomove (&tmp, iov->iov_len, uio);
292 			} else {
293 				if (!useracc((caddr_t)iov->iov_base,
294 					iov->iov_len, uio->uio_rw))
295 					return (EFAULT);
296 				insb(uio->uio_offset, iov->iov_base,
297 					iov->iov_len);
298 			}
299 			break;
300 		case 15:
301 			if (iov->iov_len == sizeof (short)) {
302 				u_short tmp;
303 				tmp = inw(uio->uio_offset);
304 				error = uiomove (&tmp, iov->iov_len, uio);
305 			} else {
306 				if (!useracc((caddr_t)iov->iov_base,
307 					iov->iov_len, uio->uio_rw))
308 					return (EFAULT);
309 				insw(uio->uio_offset, iov->iov_base,
310 					iov->iov_len/ sizeof (short));
311 			}
312 			break;
313 		case 16:
314 			if (iov->iov_len == sizeof (long)) {
315 				u_long tmp;
316 				tmp = inl(uio->uio_offset);
317 				error = uiomove (&tmp, iov->iov_len, uio);
318 			} else {
319 				if (!useracc((caddr_t)iov->iov_base,
320 					iov->iov_len, uio->uio_rw))
321 					return (EFAULT);
322 				insl(uio->uio_offset, iov->iov_base,
323 					iov->iov_len/ sizeof (long));
324 			}
325 			break;
326 #endif
327 
328 		default:
329 			return (ENXIO);
330 		}
331 		if (error)
332 			break;
333 		iov->iov_base += c;
334 		iov->iov_len -= c;
335 		uio->uio_offset += c;
336 		uio->uio_resid -= c;
337 	}
338 	if (buf)
339 		free(buf, M_TEMP);
340 	return (error);
341 }
342 
343 
344 
345 
346 /*******************************************************\
347 * allow user processes to MMAP some memory sections	*
348 * instead of going through read/write			*
349 \*******************************************************/
350 static int
351 memmmap(dev_t dev, vm_offset_t offset, int nprot)
352 {
353 	switch (minor(dev))
354 	{
355 
356 /* minor device 0 is physical memory */
357 	case 0:
358         	return i386_btop(offset);
359 
360 /* minor device 1 is kernel memory */
361 	case 1:
362         	return i386_btop(vtophys(offset));
363 
364 	default:
365 		return -1;
366 	}
367 }
368 
369 static int
370 mmioctl(dev, cmd, data, flags, p)
371 	dev_t dev;
372 	u_long cmd;
373 	caddr_t data;
374 	int flags;
375 	struct proc *p;
376 {
377 
378 	switch (minor(dev)) {
379 	case 0:
380 		return mem_ioctl(dev, cmd, data, flags, p);
381 	case 3:
382 	case 4:
383 		return random_ioctl(dev, cmd, data, flags, p);
384 	}
385 	return (ENODEV);
386 }
387 
388 /*
389  * Operations for changing memory attributes.
390  *
391  * This is basically just an ioctl shim for mem_range_attr_get
392  * and mem_range_attr_set.
393  */
394 static int
395 mem_ioctl(dev, cmd, data, flags, p)
396 	dev_t dev;
397 	u_long cmd;
398 	caddr_t data;
399 	int flags;
400 	struct proc *p;
401 {
402 	int nd, error = 0;
403 	struct mem_range_op *mo = (struct mem_range_op *)data;
404 	struct mem_range_desc *md;
405 
406 	/* is this for us? */
407 	if ((cmd != MEMRANGE_GET) &&
408 	    (cmd != MEMRANGE_SET))
409 		return (ENOTTY);
410 
411 	/* any chance we can handle this? */
412 	if (mem_range_softc.mr_op == NULL)
413 		return (EOPNOTSUPP);
414 
415 	/* do we have any descriptors? */
416 	if (mem_range_softc.mr_ndesc == 0)
417 		return (ENXIO);
418 
419 	switch (cmd) {
420 	case MEMRANGE_GET:
421 		nd = imin(mo->mo_arg[0], mem_range_softc.mr_ndesc);
422 		if (nd > 0) {
423 			md = (struct mem_range_desc *)
424 				malloc(nd * sizeof(struct mem_range_desc),
425 				       M_MEMDESC, M_WAITOK);
426 			error = mem_range_attr_get(md, &nd);
427 			if (!error)
428 				error = copyout(md, mo->mo_desc,
429 					nd * sizeof(struct mem_range_desc));
430 			free(md, M_MEMDESC);
431 		} else {
432 			nd = mem_range_softc.mr_ndesc;
433 		}
434 		mo->mo_arg[0] = nd;
435 		break;
436 
437 	case MEMRANGE_SET:
438 		md = (struct mem_range_desc *)malloc(sizeof(struct mem_range_desc),
439 						    M_MEMDESC, M_WAITOK);
440 		error = copyin(mo->mo_desc, md, sizeof(struct mem_range_desc));
441 		/* clamp description string */
442 		md->mr_owner[sizeof(md->mr_owner) - 1] = 0;
443 		if (error == 0)
444 			error = mem_range_attr_set(md, &mo->mo_arg[0]);
445 		free(md, M_MEMDESC);
446 		break;
447 	}
448 	return (error);
449 }
450 
451 /*
452  * Implementation-neutral, kernel-callable functions for manipulating
453  * memory range attributes.
454  */
455 int
456 mem_range_attr_get(mrd, arg)
457 	struct mem_range_desc *mrd;
458 	int *arg;
459 {
460 	/* can we handle this? */
461 	if (mem_range_softc.mr_op == NULL)
462 		return (EOPNOTSUPP);
463 
464 	if (*arg == 0) {
465 		*arg = mem_range_softc.mr_ndesc;
466 	} else {
467 		bcopy(mem_range_softc.mr_desc, mrd, (*arg) * sizeof(struct mem_range_desc));
468 	}
469 	return (0);
470 }
471 
472 int
473 mem_range_attr_set(mrd, arg)
474 	struct mem_range_desc *mrd;
475 	int *arg;
476 {
477 	/* can we handle this? */
478 	if (mem_range_softc.mr_op == NULL)
479 		return (EOPNOTSUPP);
480 
481 	return (mem_range_softc.mr_op->set(&mem_range_softc, mrd, arg));
482 }
483 
484 #ifdef SMP
485 void
486 mem_range_AP_init(void)
487 {
488 	if (mem_range_softc.mr_op && mem_range_softc.mr_op->initAP)
489 		return (mem_range_softc.mr_op->initAP(&mem_range_softc));
490 }
491 #endif
492 
493 static int
494 random_ioctl(dev, cmd, data, flags, p)
495 	dev_t dev;
496 	u_long cmd;
497 	caddr_t data;
498 	int flags;
499 	struct proc *p;
500 {
501 	static intrmask_t interrupt_allowed;
502 	intrmask_t interrupt_mask;
503 	int error, intr;
504 	struct random_softc *sc;
505 
506 	/*
507 	 * We're the random or urandom device.  The only ioctls are for
508 	 * selecting and inspecting which interrupts are used in the muck
509 	 * gathering business.
510 	 */
511 	if (cmd != MEM_SETIRQ && cmd != MEM_CLEARIRQ && cmd != MEM_RETURNIRQ)
512 		return (ENOTTY);
513 
514 	/*
515 	 * Even inspecting the state is privileged, since it gives a hint
516 	 * about how easily the randomness might be guessed.
517 	 */
518 	error = suser(p);
519 	if (error != 0)
520 		return (error);
521 
522 	/*
523 	 * XXX the data is 16-bit due to a historical botch, so we use
524 	 * magic 16's instead of ICU_LEN and can't support 24 interrupts
525 	 * under SMP.
526 	 */
527 	intr = *(int16_t *)data;
528 	if (cmd != MEM_RETURNIRQ && (intr < 0 || intr >= 16))
529 		return (EINVAL);
530 
531 	interrupt_mask = 1 << intr;
532 	sc = &random_softc[intr];
533 	switch (cmd) {
534 	case MEM_SETIRQ:
535 		if (interrupt_allowed & interrupt_mask)
536 			break;
537 		interrupt_allowed |= interrupt_mask;
538 		sc->sc_intr = intr;
539 		disable_intr();
540 		sc->sc_handler = intr_handler[intr];
541 		intr_handler[intr] = add_interrupt_randomness;
542 		sc->sc_arg = intr_unit[intr];
543 		intr_unit[intr] = sc;
544 		enable_intr();
545 		break;
546 	case MEM_CLEARIRQ:
547 		if (!(interrupt_allowed & interrupt_mask))
548 			break;
549 		interrupt_allowed &= ~interrupt_mask;
550 		disable_intr();
551 		intr_handler[intr] = sc->sc_handler;
552 		intr_unit[intr] = sc->sc_arg;
553 		enable_intr();
554 		break;
555 	case MEM_RETURNIRQ:
556 		*(u_int16_t *)data = interrupt_allowed;
557 		break;
558 	}
559 	return (0);
560 }
561 
562 int
563 mmpoll(dev, events, p)
564 	dev_t dev;
565 	int events;
566 	struct proc *p;
567 {
568 	switch (minor(dev)) {
569 	case 3:		/* /dev/random */
570 		return random_poll(dev, events, p);
571 	case 4:		/* /dev/urandom */
572 	default:
573 		return seltrue(dev, events, p);
574 	}
575 }
576 
577 /*
578  * Routine that identifies /dev/mem and /dev/kmem.
579  *
580  * A minimal stub routine can always return 0.
581  */
582 int
583 iskmemdev(dev)
584 	dev_t dev;
585 {
586 
587 	return ((major(dev) == mem_cdevsw.d_maj)
588 	      && (minor(dev) == 0 || minor(dev) == 1));
589 }
590 
591 int
592 iszerodev(dev)
593 	dev_t dev;
594 {
595 	return ((major(dev) == mem_cdevsw.d_maj)
596 	  && minor(dev) == 12);
597 }
598 
599 static void
600 mem_drvinit(void *unused)
601 {
602 
603 	/* Initialise memory range handling */
604 	if (mem_range_softc.mr_op != NULL)
605 		mem_range_softc.mr_op->init(&mem_range_softc);
606 
607 	make_dev(&mem_cdevsw, 0, UID_ROOT, GID_KMEM, 0640, "mem");
608 	make_dev(&mem_cdevsw, 1, UID_ROOT, GID_KMEM, 0640, "kmem");
609 	make_dev(&mem_cdevsw, 2, UID_ROOT, GID_WHEEL, 0666, "null");
610 	make_dev(&mem_cdevsw, 3, UID_ROOT, GID_WHEEL, 0644, "random");
611 	make_dev(&mem_cdevsw, 4, UID_ROOT, GID_WHEEL, 0644, "urandom");
612 	make_dev(&mem_cdevsw, 12, UID_ROOT, GID_WHEEL, 0666, "zero");
613 	make_dev(&mem_cdevsw, 14, UID_ROOT, GID_WHEEL, 0600, "io");
614 }
615 
616 SYSINIT(memdev,SI_SUB_DRIVERS,SI_ORDER_MIDDLE+CDEV_MAJOR,mem_drvinit,NULL)
617 
618