xref: /freebsd/sys/amd64/amd64/mem.c (revision a14a0223ae1b172e96dd2a1d849e22026a98b692)
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/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	mmrw;
74 static	d_ioctl_t	mmioctl;
75 static	d_mmap_t	memmmap;
76 static	d_poll_t	mmpoll;
77 
78 #define CDEV_MAJOR 2
79 static struct cdevsw mem_cdevsw = {
80 	/* open */	mmopen,
81 	/* close */	mmclose,
82 	/* read */	mmrw,
83 	/* write */	mmrw,
84 	/* ioctl */	mmioctl,
85 	/* poll */	mmpoll,
86 	/* mmap */	memmmap,
87 	/* strategy */	nostrategy,
88 	/* name */	"mem",
89 	/* maj */	CDEV_MAJOR,
90 	/* dump */	nodump,
91 	/* psize */	nopsize,
92 	/* flags */	D_MEM,
93 	/* bmaj */	-1
94 };
95 
96 static struct random_softc random_softc[16];
97 static caddr_t	zbuf;
98 
99 MALLOC_DEFINE(M_MEMDESC, "memdesc", "memory range descriptors");
100 static int mem_ioctl __P((dev_t, u_long, caddr_t, int, struct proc *));
101 static int random_ioctl __P((dev_t, u_long, caddr_t, int, struct proc *));
102 
103 struct mem_range_softc mem_range_softc;
104 
105 
106 static int
107 mmclose(dev, flags, fmt, p)
108 	dev_t dev;
109 	int flags;
110 	int fmt;
111 	struct proc *p;
112 {
113 	switch (minor(dev)) {
114 	case 14:
115 		curproc->p_md.md_regs->tf_eflags &= ~PSL_IOPL;
116 		break;
117 	default:
118 		break;
119 	}
120 	return (0);
121 }
122 
123 static int
124 mmopen(dev, flags, fmt, p)
125 	dev_t dev;
126 	int flags;
127 	int fmt;
128 	struct proc *p;
129 {
130 	int error;
131 
132 	switch (minor(dev)) {
133 	case 14:
134 		error = suser(p);
135 		if (error != 0)
136 			return (error);
137 		if (securelevel > 0)
138 			return (EPERM);
139 		curproc->p_md.md_regs->tf_eflags |= PSL_IOPL;
140 		break;
141 	default:
142 		break;
143 	}
144 	return (0);
145 }
146 
147 static int
148 mmrw(dev, uio, flags)
149 	dev_t dev;
150 	struct uio *uio;
151 	int flags;
152 {
153 	register int o;
154 	register u_int c, v;
155 	u_int poolsize;
156 	register struct iovec *iov;
157 	int error = 0;
158 	caddr_t buf = NULL;
159 
160 	while (uio->uio_resid > 0 && error == 0) {
161 		iov = uio->uio_iov;
162 		if (iov->iov_len == 0) {
163 			uio->uio_iov++;
164 			uio->uio_iovcnt--;
165 			if (uio->uio_iovcnt < 0)
166 				panic("mmrw");
167 			continue;
168 		}
169 		switch (minor(dev)) {
170 
171 /* minor device 0 is physical memory */
172 		case 0:
173 			v = uio->uio_offset;
174 			pmap_enter(kernel_pmap, (vm_offset_t)ptvmmap, v,
175 				uio->uio_rw == UIO_READ ? VM_PROT_READ : VM_PROT_WRITE,
176 				TRUE);
177 			o = (int)uio->uio_offset & PAGE_MASK;
178 			c = (u_int)(PAGE_SIZE - ((int)iov->iov_base & PAGE_MASK));
179 			c = min(c, (u_int)(PAGE_SIZE - o));
180 			c = min(c, (u_int)iov->iov_len);
181 			error = uiomove((caddr_t)&ptvmmap[o], (int)c, uio);
182 			pmap_remove(kernel_pmap, (vm_offset_t)ptvmmap,
183 				    (vm_offset_t)&ptvmmap[PAGE_SIZE]);
184 			continue;
185 
186 /* minor device 1 is kernel memory */
187 		case 1: {
188 			vm_offset_t addr, eaddr;
189 			c = iov->iov_len;
190 
191 			/*
192 			 * Make sure that all of the pages are currently resident so
193 			 * that we don't create any zero-fill pages.
194 			 */
195 			addr = trunc_page(uio->uio_offset);
196 			eaddr = round_page(uio->uio_offset + c);
197 
198 			if (addr < (vm_offset_t)VADDR(PTDPTDI, 0))
199 				return EFAULT;
200 			if (eaddr >= (vm_offset_t)VADDR(APTDPTDI, 0))
201 				return EFAULT;
202 			for (; addr < eaddr; addr += PAGE_SIZE)
203 				if (pmap_extract(kernel_pmap, addr) == 0)
204 					return EFAULT;
205 
206 			if (!kernacc((caddr_t)(int)uio->uio_offset, c,
207 			    uio->uio_rw == UIO_READ ?
208 			    VM_PROT_READ : VM_PROT_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 		default:
282 			return (ENXIO);
283 		}
284 		if (error)
285 			break;
286 		iov->iov_base += c;
287 		iov->iov_len -= c;
288 		uio->uio_offset += c;
289 		uio->uio_resid -= c;
290 	}
291 	if (buf)
292 		free(buf, M_TEMP);
293 	return (error);
294 }
295 
296 
297 
298 
299 /*******************************************************\
300 * allow user processes to MMAP some memory sections	*
301 * instead of going through read/write			*
302 \*******************************************************/
303 static int
304 memmmap(dev_t dev, vm_offset_t offset, int nprot)
305 {
306 	switch (minor(dev))
307 	{
308 
309 /* minor device 0 is physical memory */
310 	case 0:
311         	return i386_btop(offset);
312 
313 /* minor device 1 is kernel memory */
314 	case 1:
315         	return i386_btop(vtophys(offset));
316 
317 	default:
318 		return -1;
319 	}
320 }
321 
322 static int
323 mmioctl(dev, cmd, data, flags, p)
324 	dev_t dev;
325 	u_long cmd;
326 	caddr_t data;
327 	int flags;
328 	struct proc *p;
329 {
330 
331 	switch (minor(dev)) {
332 	case 0:
333 		return mem_ioctl(dev, cmd, data, flags, p);
334 	case 3:
335 	case 4:
336 		return random_ioctl(dev, cmd, data, flags, p);
337 	}
338 	return (ENODEV);
339 }
340 
341 /*
342  * Operations for changing memory attributes.
343  *
344  * This is basically just an ioctl shim for mem_range_attr_get
345  * and mem_range_attr_set.
346  */
347 static int
348 mem_ioctl(dev, cmd, data, flags, p)
349 	dev_t dev;
350 	u_long cmd;
351 	caddr_t data;
352 	int flags;
353 	struct proc *p;
354 {
355 	int nd, error = 0;
356 	struct mem_range_op *mo = (struct mem_range_op *)data;
357 	struct mem_range_desc *md;
358 
359 	/* is this for us? */
360 	if ((cmd != MEMRANGE_GET) &&
361 	    (cmd != MEMRANGE_SET))
362 		return (ENOTTY);
363 
364 	/* any chance we can handle this? */
365 	if (mem_range_softc.mr_op == NULL)
366 		return (EOPNOTSUPP);
367 
368 	/* do we have any descriptors? */
369 	if (mem_range_softc.mr_ndesc == 0)
370 		return (ENXIO);
371 
372 	switch (cmd) {
373 	case MEMRANGE_GET:
374 		nd = imin(mo->mo_arg[0], mem_range_softc.mr_ndesc);
375 		if (nd > 0) {
376 			md = (struct mem_range_desc *)
377 				malloc(nd * sizeof(struct mem_range_desc),
378 				       M_MEMDESC, M_WAITOK);
379 			error = mem_range_attr_get(md, &nd);
380 			if (!error)
381 				error = copyout(md, mo->mo_desc,
382 					nd * sizeof(struct mem_range_desc));
383 			free(md, M_MEMDESC);
384 		} else {
385 			nd = mem_range_softc.mr_ndesc;
386 		}
387 		mo->mo_arg[0] = nd;
388 		break;
389 
390 	case MEMRANGE_SET:
391 		md = (struct mem_range_desc *)malloc(sizeof(struct mem_range_desc),
392 						    M_MEMDESC, M_WAITOK);
393 		error = copyin(mo->mo_desc, md, sizeof(struct mem_range_desc));
394 		/* clamp description string */
395 		md->mr_owner[sizeof(md->mr_owner) - 1] = 0;
396 		if (error == 0)
397 			error = mem_range_attr_set(md, &mo->mo_arg[0]);
398 		free(md, M_MEMDESC);
399 		break;
400 	}
401 	return (error);
402 }
403 
404 /*
405  * Implementation-neutral, kernel-callable functions for manipulating
406  * memory range attributes.
407  */
408 int
409 mem_range_attr_get(mrd, arg)
410 	struct mem_range_desc *mrd;
411 	int *arg;
412 {
413 	/* can we handle this? */
414 	if (mem_range_softc.mr_op == NULL)
415 		return (EOPNOTSUPP);
416 
417 	if (*arg == 0) {
418 		*arg = mem_range_softc.mr_ndesc;
419 	} else {
420 		bcopy(mem_range_softc.mr_desc, mrd, (*arg) * sizeof(struct mem_range_desc));
421 	}
422 	return (0);
423 }
424 
425 int
426 mem_range_attr_set(mrd, arg)
427 	struct mem_range_desc *mrd;
428 	int *arg;
429 {
430 	/* can we handle this? */
431 	if (mem_range_softc.mr_op == NULL)
432 		return (EOPNOTSUPP);
433 
434 	return (mem_range_softc.mr_op->set(&mem_range_softc, mrd, arg));
435 }
436 
437 #ifdef SMP
438 void
439 mem_range_AP_init(void)
440 {
441 	if (mem_range_softc.mr_op && mem_range_softc.mr_op->initAP)
442 		return (mem_range_softc.mr_op->initAP(&mem_range_softc));
443 }
444 #endif
445 
446 static int
447 random_ioctl(dev, cmd, data, flags, p)
448 	dev_t dev;
449 	u_long cmd;
450 	caddr_t data;
451 	int flags;
452 	struct proc *p;
453 {
454 	static intrmask_t interrupt_allowed;
455 	intrmask_t interrupt_mask;
456 	int error, intr;
457 	struct random_softc *sc;
458 
459 	/*
460 	 * We're the random or urandom device.  The only ioctls are for
461 	 * selecting and inspecting which interrupts are used in the muck
462 	 * gathering business.
463 	 */
464 	if (cmd != MEM_SETIRQ && cmd != MEM_CLEARIRQ && cmd != MEM_RETURNIRQ)
465 		return (ENOTTY);
466 
467 	/*
468 	 * Even inspecting the state is privileged, since it gives a hint
469 	 * about how easily the randomness might be guessed.
470 	 */
471 	error = suser(p);
472 	if (error != 0)
473 		return (error);
474 
475 	/*
476 	 * XXX the data is 16-bit due to a historical botch, so we use
477 	 * magic 16's instead of ICU_LEN and can't support 24 interrupts
478 	 * under SMP.
479 	 */
480 	intr = *(int16_t *)data;
481 	if (cmd != MEM_RETURNIRQ && (intr < 0 || intr >= 16))
482 		return (EINVAL);
483 
484 	interrupt_mask = 1 << intr;
485 	sc = &random_softc[intr];
486 	switch (cmd) {
487 	case MEM_SETIRQ:
488 		if (interrupt_allowed & interrupt_mask)
489 			break;
490 		interrupt_allowed |= interrupt_mask;
491 		sc->sc_intr = intr;
492 		disable_intr();
493 		sc->sc_handler = intr_handler[intr];
494 		intr_handler[intr] = add_interrupt_randomness;
495 		sc->sc_arg = intr_unit[intr];
496 		intr_unit[intr] = sc;
497 		enable_intr();
498 		break;
499 	case MEM_CLEARIRQ:
500 		if (!(interrupt_allowed & interrupt_mask))
501 			break;
502 		interrupt_allowed &= ~interrupt_mask;
503 		disable_intr();
504 		intr_handler[intr] = sc->sc_handler;
505 		intr_unit[intr] = sc->sc_arg;
506 		enable_intr();
507 		break;
508 	case MEM_RETURNIRQ:
509 		*(u_int16_t *)data = interrupt_allowed;
510 		break;
511 	}
512 	return (0);
513 }
514 
515 int
516 mmpoll(dev, events, p)
517 	dev_t dev;
518 	int events;
519 	struct proc *p;
520 {
521 	switch (minor(dev)) {
522 	case 3:		/* /dev/random */
523 		return random_poll(dev, events, p);
524 	case 4:		/* /dev/urandom */
525 	default:
526 		return seltrue(dev, events, p);
527 	}
528 }
529 
530 /*
531  * Routine that identifies /dev/mem and /dev/kmem.
532  *
533  * A minimal stub routine can always return 0.
534  */
535 int
536 iskmemdev(dev)
537 	dev_t dev;
538 {
539 
540 	return ((major(dev) == mem_cdevsw.d_maj)
541 	      && (minor(dev) == 0 || minor(dev) == 1));
542 }
543 
544 int
545 iszerodev(dev)
546 	dev_t dev;
547 {
548 	return ((major(dev) == mem_cdevsw.d_maj)
549 	  && minor(dev) == 12);
550 }
551 
552 static void
553 mem_drvinit(void *unused)
554 {
555 
556 	/* Initialise memory range handling */
557 	if (mem_range_softc.mr_op != NULL)
558 		mem_range_softc.mr_op->init(&mem_range_softc);
559 
560 	make_dev(&mem_cdevsw, 0, UID_ROOT, GID_KMEM, 0640, "mem");
561 	make_dev(&mem_cdevsw, 1, UID_ROOT, GID_KMEM, 0640, "kmem");
562 	make_dev(&mem_cdevsw, 2, UID_ROOT, GID_WHEEL, 0666, "null");
563 	make_dev(&mem_cdevsw, 3, UID_ROOT, GID_WHEEL, 0644, "random");
564 	make_dev(&mem_cdevsw, 4, UID_ROOT, GID_WHEEL, 0644, "urandom");
565 	make_dev(&mem_cdevsw, 12, UID_ROOT, GID_WHEEL, 0666, "zero");
566 	make_dev(&mem_cdevsw, 14, UID_ROOT, GID_WHEEL, 0600, "io");
567 }
568 
569 SYSINIT(memdev,SI_SUB_DRIVERS,SI_ORDER_MIDDLE+CDEV_MAJOR,mem_drvinit,NULL)
570 
571