xref: /freebsd/sys/amd64/amd64/mem.c (revision a79b71281cd63ad7a6cc43a6d5673a2510b51630)
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/conf.h>
50 #include <sys/fcntl.h>
51 #include <sys/ioccom.h>
52 #include <sys/kernel.h>
53 #include <sys/malloc.h>
54 #include <sys/memrange.h>
55 #include <sys/proc.h>
56 #include <sys/random.h>
57 #include <sys/signalvar.h>
58 #include <sys/systm.h>
59 #include <sys/uio.h>
60 
61 #include <machine/frame.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 		p->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 0:
134 	case 1:
135 		if ((flags & FWRITE) && securelevel > 0)
136 			return (EPERM);
137 		break;
138 	case 14:
139 		error = suser(p);
140 		if (error != 0)
141 			return (error);
142 		if (securelevel > 0)
143 			return (EPERM);
144 		p->p_md.md_regs->tf_eflags |= PSL_IOPL;
145 		break;
146 	default:
147 		break;
148 	}
149 	return (0);
150 }
151 
152 static int
153 mmrw(dev, uio, flags)
154 	dev_t dev;
155 	struct uio *uio;
156 	int flags;
157 {
158 	register int o;
159 	register u_int c, v;
160 	u_int poolsize;
161 	register struct iovec *iov;
162 	int error = 0;
163 	caddr_t buf = NULL;
164 
165 	while (uio->uio_resid > 0 && error == 0) {
166 		iov = uio->uio_iov;
167 		if (iov->iov_len == 0) {
168 			uio->uio_iov++;
169 			uio->uio_iovcnt--;
170 			if (uio->uio_iovcnt < 0)
171 				panic("mmrw");
172 			continue;
173 		}
174 		switch (minor(dev)) {
175 
176 /* minor device 0 is physical memory */
177 		case 0:
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 - ((int)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 /* minor device 1 is kernel memory */
190 		case 1: {
191 			vm_offset_t addr, eaddr;
192 			c = iov->iov_len;
193 
194 			/*
195 			 * Make sure that all of the pages are currently resident so
196 			 * that we don't create any zero-fill pages.
197 			 */
198 			addr = trunc_page(uio->uio_offset);
199 			eaddr = round_page(uio->uio_offset + c);
200 
201 			if (addr < (vm_offset_t)VADDR(PTDPTDI, 0))
202 				return EFAULT;
203 			if (eaddr >= (vm_offset_t)VADDR(APTDPTDI, 0))
204 				return EFAULT;
205 			for (; addr < eaddr; addr += PAGE_SIZE)
206 				if (pmap_extract(kernel_pmap, addr) == 0)
207 					return EFAULT;
208 
209 			if (!kernacc((caddr_t)(int)uio->uio_offset, c,
210 			    uio->uio_rw == UIO_READ ?
211 			    VM_PROT_READ : VM_PROT_WRITE))
212 				return (EFAULT);
213 			error = uiomove((caddr_t)(int)uio->uio_offset, (int)c, uio);
214 			continue;
215 		}
216 
217 /* minor device 2 is EOF/RATHOLE */
218 		case 2:
219 			if (uio->uio_rw == UIO_READ)
220 				return (0);
221 			c = iov->iov_len;
222 			break;
223 
224 /* minor device 3 (/dev/random) is source of filth on read, rathole on write */
225 		case 3:
226 			if (uio->uio_rw == UIO_WRITE) {
227 				c = iov->iov_len;
228 				break;
229 			}
230 			if (buf == NULL)
231 				buf = (caddr_t)
232 				    malloc(PAGE_SIZE, M_TEMP, M_WAITOK);
233 			c = min(iov->iov_len, PAGE_SIZE);
234 			poolsize = read_random(buf, c);
235 			if (poolsize == 0) {
236 				if (buf)
237 					free(buf, M_TEMP);
238 				return (0);
239 			}
240 			c = min(c, poolsize);
241 			error = uiomove(buf, (int)c, uio);
242 			continue;
243 
244 /* minor device 4 (/dev/urandom) is source of muck on read, rathole on write */
245 		case 4:
246 			if (uio->uio_rw == UIO_WRITE) {
247 				c = iov->iov_len;
248 				break;
249 			}
250 			if (CURSIG(curproc) != 0) {
251 				/*
252 				 * Use tsleep() to get the error code right.
253 				 * It should return immediately.
254 				 */
255 				error = tsleep(&random_softc[0],
256 				    PZERO | PCATCH, "urand", 1);
257 				if (error != 0 && error != EWOULDBLOCK)
258 					continue;
259 			}
260 			if (buf == NULL)
261 				buf = (caddr_t)
262 				    malloc(PAGE_SIZE, M_TEMP, M_WAITOK);
263 			c = min(iov->iov_len, PAGE_SIZE);
264 			poolsize = read_random_unlimited(buf, c);
265 			c = min(c, poolsize);
266 			error = uiomove(buf, (int)c, uio);
267 			continue;
268 
269 /* minor device 12 (/dev/zero) is source of nulls on read, rathole on write */
270 		case 12:
271 			if (uio->uio_rw == UIO_WRITE) {
272 				c = iov->iov_len;
273 				break;
274 			}
275 			if (zbuf == NULL) {
276 				zbuf = (caddr_t)
277 				    malloc(PAGE_SIZE, M_TEMP, M_WAITOK);
278 				bzero(zbuf, PAGE_SIZE);
279 			}
280 			c = min(iov->iov_len, PAGE_SIZE);
281 			error = uiomove(zbuf, (int)c, uio);
282 			continue;
283 
284 		default:
285 			return (ENXIO);
286 		}
287 		if (error)
288 			break;
289 		iov->iov_base += c;
290 		iov->iov_len -= c;
291 		uio->uio_offset += c;
292 		uio->uio_resid -= c;
293 	}
294 	if (buf)
295 		free(buf, M_TEMP);
296 	return (error);
297 }
298 
299 
300 
301 
302 /*******************************************************\
303 * allow user processes to MMAP some memory sections	*
304 * instead of going through read/write			*
305 \*******************************************************/
306 static int
307 memmmap(dev_t dev, vm_offset_t offset, int nprot)
308 {
309 	switch (minor(dev))
310 	{
311 
312 /* minor device 0 is physical memory */
313 	case 0:
314         	return i386_btop(offset);
315 
316 /* minor device 1 is kernel memory */
317 	case 1:
318         	return i386_btop(vtophys(offset));
319 
320 	default:
321 		return -1;
322 	}
323 }
324 
325 static int
326 mmioctl(dev, cmd, data, flags, p)
327 	dev_t dev;
328 	u_long cmd;
329 	caddr_t data;
330 	int flags;
331 	struct proc *p;
332 {
333 
334 	switch (minor(dev)) {
335 	case 0:
336 		return mem_ioctl(dev, cmd, data, flags, p);
337 	case 3:
338 	case 4:
339 		return random_ioctl(dev, cmd, data, flags, p);
340 	}
341 	return (ENODEV);
342 }
343 
344 /*
345  * Operations for changing memory attributes.
346  *
347  * This is basically just an ioctl shim for mem_range_attr_get
348  * and mem_range_attr_set.
349  */
350 static int
351 mem_ioctl(dev, cmd, data, flags, p)
352 	dev_t dev;
353 	u_long cmd;
354 	caddr_t data;
355 	int flags;
356 	struct proc *p;
357 {
358 	int nd, error = 0;
359 	struct mem_range_op *mo = (struct mem_range_op *)data;
360 	struct mem_range_desc *md;
361 
362 	/* is this for us? */
363 	if ((cmd != MEMRANGE_GET) &&
364 	    (cmd != MEMRANGE_SET))
365 		return (ENOTTY);
366 
367 	/* any chance we can handle this? */
368 	if (mem_range_softc.mr_op == NULL)
369 		return (EOPNOTSUPP);
370 
371 	/* do we have any descriptors? */
372 	if (mem_range_softc.mr_ndesc == 0)
373 		return (ENXIO);
374 
375 	switch (cmd) {
376 	case MEMRANGE_GET:
377 		nd = imin(mo->mo_arg[0], mem_range_softc.mr_ndesc);
378 		if (nd > 0) {
379 			md = (struct mem_range_desc *)
380 				malloc(nd * sizeof(struct mem_range_desc),
381 				       M_MEMDESC, M_WAITOK);
382 			error = mem_range_attr_get(md, &nd);
383 			if (!error)
384 				error = copyout(md, mo->mo_desc,
385 					nd * sizeof(struct mem_range_desc));
386 			free(md, M_MEMDESC);
387 		} else {
388 			nd = mem_range_softc.mr_ndesc;
389 		}
390 		mo->mo_arg[0] = nd;
391 		break;
392 
393 	case MEMRANGE_SET:
394 		md = (struct mem_range_desc *)malloc(sizeof(struct mem_range_desc),
395 						    M_MEMDESC, M_WAITOK);
396 		error = copyin(mo->mo_desc, md, sizeof(struct mem_range_desc));
397 		/* clamp description string */
398 		md->mr_owner[sizeof(md->mr_owner) - 1] = 0;
399 		if (error == 0)
400 			error = mem_range_attr_set(md, &mo->mo_arg[0]);
401 		free(md, M_MEMDESC);
402 		break;
403 	}
404 	return (error);
405 }
406 
407 /*
408  * Implementation-neutral, kernel-callable functions for manipulating
409  * memory range attributes.
410  */
411 int
412 mem_range_attr_get(mrd, arg)
413 	struct mem_range_desc *mrd;
414 	int *arg;
415 {
416 	/* can we handle this? */
417 	if (mem_range_softc.mr_op == NULL)
418 		return (EOPNOTSUPP);
419 
420 	if (*arg == 0) {
421 		*arg = mem_range_softc.mr_ndesc;
422 	} else {
423 		bcopy(mem_range_softc.mr_desc, mrd, (*arg) * sizeof(struct mem_range_desc));
424 	}
425 	return (0);
426 }
427 
428 int
429 mem_range_attr_set(mrd, arg)
430 	struct mem_range_desc *mrd;
431 	int *arg;
432 {
433 	/* can we handle this? */
434 	if (mem_range_softc.mr_op == NULL)
435 		return (EOPNOTSUPP);
436 
437 	return (mem_range_softc.mr_op->set(&mem_range_softc, mrd, arg));
438 }
439 
440 #ifdef SMP
441 void
442 mem_range_AP_init(void)
443 {
444 	if (mem_range_softc.mr_op && mem_range_softc.mr_op->initAP)
445 		return (mem_range_softc.mr_op->initAP(&mem_range_softc));
446 }
447 #endif
448 
449 static int
450 random_ioctl(dev, cmd, data, flags, p)
451 	dev_t dev;
452 	u_long cmd;
453 	caddr_t data;
454 	int flags;
455 	struct proc *p;
456 {
457 	static intrmask_t interrupt_allowed;
458 	intrmask_t interrupt_mask;
459 	int error, intr;
460 	struct random_softc *sc;
461 
462 	/*
463 	 * We're the random or urandom device.  The only ioctls are for
464 	 * selecting and inspecting which interrupts are used in the muck
465 	 * gathering business.
466 	 */
467 	if (cmd != MEM_SETIRQ && cmd != MEM_CLEARIRQ && cmd != MEM_RETURNIRQ)
468 		return (ENOTTY);
469 
470 	/*
471 	 * Even inspecting the state is privileged, since it gives a hint
472 	 * about how easily the randomness might be guessed.
473 	 */
474 	error = suser(p);
475 	if (error != 0)
476 		return (error);
477 
478 	/*
479 	 * XXX the data is 16-bit due to a historical botch, so we use
480 	 * magic 16's instead of ICU_LEN and can't support 24 interrupts
481 	 * under SMP.
482 	 */
483 	intr = *(int16_t *)data;
484 	if (cmd != MEM_RETURNIRQ && (intr < 0 || intr >= 16))
485 		return (EINVAL);
486 
487 	interrupt_mask = 1 << intr;
488 	sc = &random_softc[intr];
489 	switch (cmd) {
490 	case MEM_SETIRQ:
491 		if (interrupt_allowed & interrupt_mask)
492 			break;
493 		interrupt_allowed |= interrupt_mask;
494 		sc->sc_intr = intr;
495 		disable_intr();
496 		sc->sc_handler = intr_handler[intr];
497 		intr_handler[intr] = add_interrupt_randomness;
498 		sc->sc_arg = intr_unit[intr];
499 		intr_unit[intr] = sc;
500 		enable_intr();
501 		break;
502 	case MEM_CLEARIRQ:
503 		if (!(interrupt_allowed & interrupt_mask))
504 			break;
505 		interrupt_allowed &= ~interrupt_mask;
506 		disable_intr();
507 		intr_handler[intr] = sc->sc_handler;
508 		intr_unit[intr] = sc->sc_arg;
509 		enable_intr();
510 		break;
511 	case MEM_RETURNIRQ:
512 		*(u_int16_t *)data = interrupt_allowed;
513 		break;
514 	}
515 	return (0);
516 }
517 
518 int
519 mmpoll(dev, events, p)
520 	dev_t dev;
521 	int events;
522 	struct proc *p;
523 {
524 	switch (minor(dev)) {
525 	case 3:		/* /dev/random */
526 		return random_poll(dev, events, p);
527 	case 4:		/* /dev/urandom */
528 	default:
529 		return seltrue(dev, events, p);
530 	}
531 }
532 
533 int
534 iszerodev(dev)
535 	dev_t dev;
536 {
537 	return ((major(dev) == mem_cdevsw.d_maj)
538 	  && minor(dev) == 12);
539 }
540 
541 static void
542 mem_drvinit(void *unused)
543 {
544 
545 	/* Initialise memory range handling */
546 	if (mem_range_softc.mr_op != NULL)
547 		mem_range_softc.mr_op->init(&mem_range_softc);
548 
549 	make_dev(&mem_cdevsw, 0, UID_ROOT, GID_KMEM, 0640, "mem");
550 	make_dev(&mem_cdevsw, 1, UID_ROOT, GID_KMEM, 0640, "kmem");
551 	make_dev(&mem_cdevsw, 2, UID_ROOT, GID_WHEEL, 0666, "null");
552 	make_dev(&mem_cdevsw, 3, UID_ROOT, GID_WHEEL, 0644, "random");
553 	make_dev(&mem_cdevsw, 4, UID_ROOT, GID_WHEEL, 0644, "urandom");
554 	make_dev(&mem_cdevsw, 12, UID_ROOT, GID_WHEEL, 0666, "zero");
555 	make_dev(&mem_cdevsw, 14, UID_ROOT, GID_WHEEL, 0600, "io");
556 }
557 
558 SYSINIT(memdev,SI_SUB_DRIVERS,SI_ORDER_MIDDLE+CDEV_MAJOR,mem_drvinit,NULL)
559 
560