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. Neither the name of the University nor the names of its contributors
20 * may be used to endorse or promote products derived from this software
21 * without specific prior written permission.
22 *
23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * SUCH DAMAGE.
34 *
35 * from: Utah $Hdr: mem.c 1.13 89/10/08$
36 * from: @(#)mem.c 7.2 (Berkeley) 5/9/91
37 * $FreeBSD: src/sys/i386/i386/mem.c,v 1.79.2.9 2003/01/04 22:58:01 njl Exp $
38 */
39
40 /*
41 * Memory special file
42 */
43
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/buf.h>
47 #include <sys/conf.h>
48 #include <sys/fcntl.h>
49 #include <sys/filio.h>
50 #include <sys/interrupt.h>
51 #include <sys/kernel.h>
52 #include <sys/malloc.h>
53 #include <sys/memrange.h>
54 #include <sys/proc.h>
55 #include <sys/caps.h>
56 #include <sys/queue.h>
57 #include <sys/random.h>
58 #include <sys/signalvar.h>
59 #include <sys/uio.h>
60 #include <sys/vnode.h>
61 #include <sys/sysctl.h>
62
63 #include <sys/signal2.h>
64 #include <sys/spinlock2.h>
65
66 #include <vm/vm.h>
67 #include <vm/pmap.h>
68 #include <vm/vm_map.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 mmread;
75 static d_write_t mmwrite;
76 static d_ioctl_t mmioctl;
77 #if 0
78 static d_mmap_t memmmap;
79 #endif
80 static d_kqfilter_t mmkqfilter;
81 static int memuksmap(vm_map_backing_t ba, int op, cdev_t dev, vm_page_t fake);
82
83 #define CDEV_MAJOR 2
84 static struct dev_ops mem_ops = {
85 { "mem", 0, D_MPSAFE | D_QUICK },
86 .d_open = mmopen,
87 .d_close = mmclose,
88 .d_read = mmread,
89 .d_write = mmwrite,
90 .d_ioctl = mmioctl,
91 .d_kqfilter = mmkqfilter,
92 #if 0
93 .d_mmap = memmmap,
94 #endif
95 .d_uksmap = memuksmap
96 };
97
98 static struct dev_ops mem_ops_mem = {
99 { "mem", 0, D_MEM | D_MPSAFE | D_QUICK },
100 .d_open = mmopen,
101 .d_close = mmclose,
102 .d_read = mmread,
103 .d_write = mmwrite,
104 .d_ioctl = mmioctl,
105 .d_kqfilter = mmkqfilter,
106 #if 0
107 .d_mmap = memmmap,
108 #endif
109 .d_uksmap = memuksmap
110 };
111
112 static struct dev_ops mem_ops_noq = {
113 { "mem", 0, D_MPSAFE },
114 .d_open = mmopen,
115 .d_close = mmclose,
116 .d_read = mmread,
117 .d_write = mmwrite,
118 .d_ioctl = mmioctl,
119 .d_kqfilter = mmkqfilter,
120 #if 0
121 .d_mmap = memmmap,
122 #endif
123 .d_uksmap = memuksmap
124 };
125
126 static int rand_bolt;
127 static caddr_t zbuf;
128 static cdev_t zerodev = NULL;
129 static struct lock mem_lock = LOCK_INITIALIZER("memlk", 0, 0);
130
131 MALLOC_DEFINE(M_MEMDESC, "memdesc", "memory range descriptors");
132 static int mem_ioctl (cdev_t, u_long, caddr_t, int, struct ucred *);
133 static int random_ioctl (cdev_t, u_long, caddr_t, int, struct ucred *);
134
135 struct mem_range_softc mem_range_softc;
136
137 static int seedenable;
138 SYSCTL_INT(_kern, OID_AUTO, seedenable, CTLFLAG_RW, &seedenable, 0, "");
139
140 static int
mmopen(struct dev_open_args * ap)141 mmopen(struct dev_open_args *ap)
142 {
143 cdev_t dev = ap->a_head.a_dev;
144 int error;
145
146 switch (minor(dev)) {
147 case 0:
148 case 1:
149 /*
150 * /dev/mem and /dev/kmem
151 */
152 error = caps_priv_check(ap->a_cred, SYSCAP_RESTRICTEDROOT |
153 __SYSCAP_NOROOTTEST);
154 if (error == 0) {
155 if (ap->a_oflags & FWRITE) {
156 if (securelevel > 0 || kernel_mem_readonly)
157 error = EPERM;
158 }
159 }
160 break;
161 case 3:
162 case 4:
163 /*
164 * /dev/random
165 * /dev/urandom
166 *
167 * Cannot be written to from RESTRICTEDROOT environments.
168 */
169 error = 0;
170 if (ap->a_oflags & FWRITE) {
171 error = caps_priv_check(ap->a_cred,
172 SYSCAP_RESTRICTEDROOT |
173 __SYSCAP_NOROOTTEST);
174 }
175 break;
176 case 6:
177 /*
178 * /dev/kpmap can only be opened for reading.
179 */
180 error = 0;
181 if (ap->a_oflags & FWRITE)
182 error = EPERM;
183 break;
184 case 14:
185 /*
186 * /dev/io
187 */
188 error = caps_priv_check(ap->a_cred, SYSCAP_RESTRICTEDROOT);
189 if (error == 0) {
190 if (securelevel > 0 || kernel_mem_readonly)
191 error = EPERM;
192 else
193 error = cpu_set_iopl();
194 }
195 break;
196 default:
197 error = 0;
198 break;
199 }
200 return (error);
201 }
202
203 static int
mmclose(struct dev_close_args * ap)204 mmclose(struct dev_close_args *ap)
205 {
206 cdev_t dev = ap->a_head.a_dev;
207 int error;
208
209 switch (minor(dev)) {
210 case 14:
211 error = cpu_clr_iopl();
212 break;
213 default:
214 error = 0;
215 break;
216 }
217 return (error);
218 }
219
220
221 static int
mmrw(cdev_t dev,struct uio * uio,int flags)222 mmrw(cdev_t dev, struct uio *uio, int flags)
223 {
224 int o;
225 u_int c;
226 u_int poolsize;
227 u_long v;
228 struct iovec *iov;
229 int error = 0;
230 caddr_t buf = NULL;
231
232 while (uio->uio_resid > 0 && error == 0) {
233 iov = uio->uio_iov;
234 if (iov->iov_len == 0) {
235 uio->uio_iov++;
236 uio->uio_iovcnt--;
237 if (uio->uio_iovcnt < 0)
238 panic("mmrw");
239 continue;
240 }
241 switch (minor(dev)) {
242 case 0:
243 /*
244 * minor device 0 is physical memory, /dev/mem
245 */
246 v = uio->uio_offset;
247 v &= ~(long)PAGE_MASK;
248 pmap_kenter((vm_offset_t)ptvmmap, v);
249 o = (int)uio->uio_offset & PAGE_MASK;
250 c = (u_int)(PAGE_SIZE - ((uintptr_t)iov->iov_base & PAGE_MASK));
251 c = min(c, (u_int)(PAGE_SIZE - o));
252 c = min(c, (u_int)iov->iov_len);
253 error = uiomove((caddr_t)&ptvmmap[o], (int)c, uio);
254 pmap_kremove((vm_offset_t)ptvmmap);
255 continue;
256
257 case 1: {
258 /*
259 * minor device 1 is kernel memory, /dev/kmem
260 */
261 vm_offset_t saddr, eaddr;
262 int prot;
263
264 c = iov->iov_len;
265
266 /*
267 * Make sure that all of the pages are currently
268 * resident so that we don't create any zero-fill
269 * pages.
270 */
271 saddr = trunc_page(uio->uio_offset);
272 eaddr = round_page(uio->uio_offset + c);
273 if (saddr > eaddr)
274 return EFAULT;
275
276 /*
277 * Make sure the kernel addresses are mapped.
278 * platform_direct_mapped() can be used to bypass
279 * default mapping via the page table (virtual kernels
280 * contain a lot of out-of-band data).
281 */
282 prot = VM_PROT_READ;
283 if (uio->uio_rw != UIO_READ)
284 prot |= VM_PROT_WRITE;
285 error = kvm_access_check(saddr, eaddr, prot);
286 if (error)
287 return (error);
288 error = uiomove((caddr_t)(vm_offset_t)uio->uio_offset,
289 (int)c, uio);
290 continue;
291 }
292 case 2:
293 /*
294 * minor device 2 (/dev/null) is EOF/RATHOLE
295 */
296 if (uio->uio_rw == UIO_READ)
297 return (0);
298 c = iov->iov_len;
299 break;
300 case 3:
301 /*
302 * minor device 3 (/dev/random) is source of filth
303 * on read, seeder on write
304 */
305 if (buf == NULL)
306 buf = kmalloc(PAGE_SIZE, M_TEMP, M_WAITOK);
307 c = min(iov->iov_len, PAGE_SIZE);
308 if (uio->uio_rw == UIO_WRITE) {
309 error = uiomove(buf, (int)c, uio);
310 if (error == 0 &&
311 seedenable &&
312 securelevel <= 0)
313 {
314 error = add_buffer_randomness_src(buf, c, RAND_SRC_SEEDING);
315 } else if (error == 0) {
316 error = EPERM;
317 }
318 } else {
319 poolsize = read_random(buf, c, 0);
320 if (poolsize == 0) {
321 if (buf)
322 kfree(buf, M_TEMP);
323 if ((flags & IO_NDELAY) != 0)
324 return (EWOULDBLOCK);
325 return (0);
326 }
327 c = min(c, poolsize);
328 error = uiomove(buf, (int)c, uio);
329 }
330 continue;
331 case 4:
332 /*
333 * minor device 4 (/dev/urandom) is source of muck
334 * on read, writes are disallowed.
335 */
336 c = min(iov->iov_len, PAGE_SIZE);
337 if (uio->uio_rw == UIO_WRITE) {
338 error = EPERM;
339 break;
340 }
341 if (CURSIG(curthread->td_lwp) != 0) {
342 /*
343 * Use tsleep() to get the error code right.
344 * It should return immediately.
345 */
346 error = tsleep(&rand_bolt, PCATCH, "urand", 1);
347 if (error != 0 && error != EWOULDBLOCK)
348 continue;
349 }
350 if (buf == NULL)
351 buf = kmalloc(PAGE_SIZE, M_TEMP, M_WAITOK);
352 poolsize = read_random(buf, c, 1);
353 c = min(c, poolsize);
354 error = uiomove(buf, (int)c, uio);
355 continue;
356 /* case 5: read/write not supported, mmap only */
357 /* case 6: read/write not supported, mmap only */
358 case 12:
359 /*
360 * minor device 12 (/dev/zero) is source of nulls
361 * on read, write are disallowed.
362 */
363 if (uio->uio_rw == UIO_WRITE) {
364 c = iov->iov_len;
365 break;
366 }
367 if (zbuf == NULL) {
368 zbuf = (caddr_t)kmalloc(PAGE_SIZE, M_TEMP,
369 M_WAITOK | M_ZERO);
370 }
371 c = min(iov->iov_len, PAGE_SIZE);
372 error = uiomove(zbuf, (int)c, uio);
373 continue;
374 default:
375 return (ENODEV);
376 }
377 if (error)
378 break;
379 iov->iov_base = (char *)iov->iov_base + c;
380 iov->iov_len -= c;
381 uio->uio_offset += c;
382 uio->uio_resid -= c;
383 }
384 if (buf)
385 kfree(buf, M_TEMP);
386 return (error);
387 }
388
389 static int
mmread(struct dev_read_args * ap)390 mmread(struct dev_read_args *ap)
391 {
392 return(mmrw(ap->a_head.a_dev, ap->a_uio, ap->a_ioflag));
393 }
394
395 static int
mmwrite(struct dev_write_args * ap)396 mmwrite(struct dev_write_args *ap)
397 {
398 return(mmrw(ap->a_head.a_dev, ap->a_uio, ap->a_ioflag));
399 }
400
401 /*******************************************************\
402 * allow user processes to MMAP some memory sections *
403 * instead of going through read/write *
404 \*******************************************************/
405
406 static int user_kernel_mapping(vm_map_backing_t ba, int num,
407 vm_ooffset_t offset, vm_ooffset_t *resultp);
408
409 static int
memuksmap(vm_map_backing_t ba,int op,cdev_t dev,vm_page_t fake)410 memuksmap(vm_map_backing_t ba, int op, cdev_t dev, vm_page_t fake)
411 {
412 vm_ooffset_t result;
413 int error;
414 struct lwp *lp;
415
416 error = 0;
417
418 switch(op) {
419 case UKSMAPOP_ADD:
420 /*
421 * We only need to track mappings for /dev/lpmap, all process
422 * mappings will be deleted when the process exits and we
423 * do not need to track kernel mappings.
424 */
425 if (minor(dev) == 7) {
426 lp = ba->aux_info;
427 spin_lock(&lp->lwp_spin);
428 TAILQ_INSERT_TAIL(&lp->lwp_lpmap_backing_list,
429 ba, entry);
430 spin_unlock(&lp->lwp_spin);
431 }
432 break;
433 case UKSMAPOP_REM:
434 /*
435 * We only need to track mappings for /dev/lpmap, all process
436 * mappings will be deleted when the process exits and we
437 * do not need to track kernel mappings.
438 */
439 if (minor(dev) == 7) {
440 lp = ba->aux_info;
441 spin_lock(&lp->lwp_spin);
442 TAILQ_REMOVE(&lp->lwp_lpmap_backing_list, ba, entry);
443 spin_unlock(&lp->lwp_spin);
444 }
445 break;
446 case UKSMAPOP_FAULT:
447 switch (minor(dev)) {
448 case 0:
449 /*
450 * minor device 0 is physical memory
451 */
452 fake->phys_addr = ptoa(fake->pindex);
453 break;
454 case 1:
455 /*
456 * minor device 1 is kernel memory
457 */
458 fake->phys_addr = vtophys(ptoa(fake->pindex));
459 break;
460 case 5:
461 case 6:
462 case 7:
463 /*
464 * minor device 5 is /dev/upmap (see sys/upmap.h)
465 * minor device 6 is /dev/kpmap (see sys/upmap.h)
466 * minor device 7 is /dev/lpmap (see sys/upmap.h)
467 */
468 result = 0;
469 error = user_kernel_mapping(ba,
470 minor(dev),
471 ptoa(fake->pindex),
472 &result);
473 fake->phys_addr = result;
474 break;
475 default:
476 error = EINVAL;
477 break;
478 }
479 break;
480 default:
481 error = EINVAL;
482 break;
483 }
484 return error;
485 }
486
487 static int
mmioctl(struct dev_ioctl_args * ap)488 mmioctl(struct dev_ioctl_args *ap)
489 {
490 cdev_t dev = ap->a_head.a_dev;
491 int error;
492
493 lockmgr(&mem_lock, LK_EXCLUSIVE);
494
495 switch (minor(dev)) {
496 case 0:
497 error = mem_ioctl(dev, ap->a_cmd, ap->a_data,
498 ap->a_fflag, ap->a_cred);
499 break;
500 case 3:
501 case 4:
502 error = random_ioctl(dev, ap->a_cmd, ap->a_data,
503 ap->a_fflag, ap->a_cred);
504 break;
505 default:
506 error = ENODEV;
507 break;
508 }
509
510 lockmgr(&mem_lock, LK_RELEASE);
511
512 return (error);
513 }
514
515 /*
516 * Operations for changing memory attributes.
517 *
518 * This is basically just an ioctl shim for mem_range_attr_get
519 * and mem_range_attr_set.
520 */
521 static int
mem_ioctl(cdev_t dev,u_long cmd,caddr_t data,int flags,struct ucred * cred)522 mem_ioctl(cdev_t dev, u_long cmd, caddr_t data, int flags, struct ucred *cred)
523 {
524 int nd, error = 0;
525 struct mem_range_op *mo = (struct mem_range_op *)data;
526 struct mem_range_desc *md;
527
528 /* is this for us? */
529 if ((cmd != MEMRANGE_GET) &&
530 (cmd != MEMRANGE_SET))
531 return (ENOTTY);
532
533 /* any chance we can handle this? */
534 if (mem_range_softc.mr_op == NULL)
535 return (EOPNOTSUPP);
536
537 /* do we have any descriptors? */
538 if (mem_range_softc.mr_ndesc == 0)
539 return (ENXIO);
540
541 switch (cmd) {
542 case MEMRANGE_GET:
543 nd = imin(mo->mo_arg[0], mem_range_softc.mr_ndesc);
544 if (nd > 0) {
545 md = (struct mem_range_desc *)
546 kmalloc(nd * sizeof(struct mem_range_desc),
547 M_MEMDESC, M_WAITOK);
548 error = mem_range_attr_get(md, &nd);
549 if (!error)
550 error = copyout(md, mo->mo_desc,
551 nd * sizeof(struct mem_range_desc));
552 kfree(md, M_MEMDESC);
553 } else {
554 nd = mem_range_softc.mr_ndesc;
555 }
556 mo->mo_arg[0] = nd;
557 break;
558
559 case MEMRANGE_SET:
560 md = (struct mem_range_desc *)kmalloc(sizeof(struct mem_range_desc),
561 M_MEMDESC, M_WAITOK);
562 error = copyin(mo->mo_desc, md, sizeof(struct mem_range_desc));
563 /* clamp description string */
564 md->mr_owner[sizeof(md->mr_owner) - 1] = 0;
565 if (error == 0)
566 error = mem_range_attr_set(md, &mo->mo_arg[0]);
567 kfree(md, M_MEMDESC);
568 break;
569 }
570 return (error);
571 }
572
573 /*
574 * Implementation-neutral, kernel-callable functions for manipulating
575 * memory range attributes.
576 */
577 int
mem_range_attr_get(struct mem_range_desc * mrd,int * arg)578 mem_range_attr_get(struct mem_range_desc *mrd, int *arg)
579 {
580 /* can we handle this? */
581 if (mem_range_softc.mr_op == NULL)
582 return (EOPNOTSUPP);
583
584 if (*arg == 0) {
585 *arg = mem_range_softc.mr_ndesc;
586 } else {
587 bcopy(mem_range_softc.mr_desc, mrd, (*arg) * sizeof(struct mem_range_desc));
588 }
589 return (0);
590 }
591
592 int
mem_range_attr_set(struct mem_range_desc * mrd,int * arg)593 mem_range_attr_set(struct mem_range_desc *mrd, int *arg)
594 {
595 /* can we handle this? */
596 if (mem_range_softc.mr_op == NULL)
597 return (EOPNOTSUPP);
598
599 return (mem_range_softc.mr_op->set(&mem_range_softc, mrd, arg));
600 }
601
602 void
mem_range_AP_init(void)603 mem_range_AP_init(void)
604 {
605 if (mem_range_softc.mr_op && mem_range_softc.mr_op->initAP)
606 mem_range_softc.mr_op->initAP(&mem_range_softc);
607 }
608
609 static int
random_ioctl(cdev_t dev,u_long cmd,caddr_t data,int flags,struct ucred * cred)610 random_ioctl(cdev_t dev, u_long cmd, caddr_t data, int flags, struct ucred *cred)
611 {
612 int error;
613 int intr;
614
615 /*
616 * Even inspecting the state is privileged, since it gives a hint
617 * about how easily the randomness might be guessed.
618 */
619 error = 0;
620
621 switch (cmd) {
622 /* Really handled in upper layer */
623 case FIOASYNC:
624 break;
625 case MEM_SETIRQ:
626 intr = *(int16_t *)data;
627 if ((error = caps_priv_check(cred, SYSCAP_RESTRICTEDROOT)) != 0)
628 break;
629 if (intr < 0 || intr >= MAX_INTS)
630 return (EINVAL);
631 register_randintr(intr);
632 break;
633 case MEM_CLEARIRQ:
634 intr = *(int16_t *)data;
635 if ((error = caps_priv_check(cred, SYSCAP_RESTRICTEDROOT)) != 0)
636 break;
637 if (intr < 0 || intr >= MAX_INTS)
638 return (EINVAL);
639 unregister_randintr(intr);
640 break;
641 case MEM_RETURNIRQ:
642 error = ENOTSUP;
643 break;
644 case MEM_FINDIRQ:
645 intr = *(int16_t *)data;
646 if ((error = caps_priv_check(cred, SYSCAP_RESTRICTEDROOT)) != 0)
647 break;
648 if (intr < 0 || intr >= MAX_INTS)
649 return (EINVAL);
650 intr = next_registered_randintr(intr);
651 if (intr == MAX_INTS)
652 return (ENOENT);
653 *(u_int16_t *)data = intr;
654 break;
655 default:
656 error = ENOTSUP;
657 break;
658 }
659 return (error);
660 }
661
662 static int
mm_filter_read(struct knote * kn,long hint)663 mm_filter_read(struct knote *kn, long hint)
664 {
665 return (1);
666 }
667
668 static int
mm_filter_write(struct knote * kn,long hint)669 mm_filter_write(struct knote *kn, long hint)
670 {
671 return (1);
672 }
673
674 static void
dummy_filter_detach(struct knote * kn)675 dummy_filter_detach(struct knote *kn) {}
676
677 /* Implemented in kern_nrandom.c */
678 static struct filterops random_read_filtops =
679 { FILTEROP_ISFD|FILTEROP_MPSAFE, NULL, dummy_filter_detach, random_filter_read };
680
681 static struct filterops mm_read_filtops =
682 { FILTEROP_ISFD|FILTEROP_MPSAFE, NULL, dummy_filter_detach, mm_filter_read };
683
684 static struct filterops mm_write_filtops =
685 { FILTEROP_ISFD|FILTEROP_MPSAFE, NULL, dummy_filter_detach, mm_filter_write };
686
687 static int
mmkqfilter(struct dev_kqfilter_args * ap)688 mmkqfilter(struct dev_kqfilter_args *ap)
689 {
690 struct knote *kn = ap->a_kn;
691 cdev_t dev = ap->a_head.a_dev;
692
693 ap->a_result = 0;
694 switch (kn->kn_filter) {
695 case EVFILT_READ:
696 switch (minor(dev)) {
697 case 3:
698 kn->kn_fop = &random_read_filtops;
699 break;
700 default:
701 kn->kn_fop = &mm_read_filtops;
702 break;
703 }
704 break;
705 case EVFILT_WRITE:
706 kn->kn_fop = &mm_write_filtops;
707 break;
708 default:
709 ap->a_result = EOPNOTSUPP;
710 return (0);
711 }
712
713 return (0);
714 }
715
716 int
iszerodev(cdev_t dev)717 iszerodev(cdev_t dev)
718 {
719 return (zerodev == dev);
720 }
721
722 /*
723 * /dev/lpmap, /dev/upmap, /dev/kpmap.
724 */
725 static int
user_kernel_mapping(vm_map_backing_t ba,int num,vm_ooffset_t offset,vm_ooffset_t * resultp)726 user_kernel_mapping(vm_map_backing_t ba, int num, vm_ooffset_t offset,
727 vm_ooffset_t *resultp)
728 {
729 struct proc *p;
730 struct lwp *lp;
731 int error;
732 int invfork;
733
734 if (offset < 0)
735 return (EINVAL);
736
737 error = EINVAL;
738
739 switch(num) {
740 case 5:
741 /*
742 * /dev/upmap - maps RW per-process shared user-kernel area.
743 */
744
745 /*
746 * If this is a child currently in vfork the pmap is shared
747 * with the parent! We need to actually set-up the parent's
748 * p_upmap, not the child's, and we need to set the invfork
749 * flag. Userland will probably adjust its static state so
750 * it must be consistent with the parent or userland will be
751 * really badly confused.
752 *
753 * (this situation can happen when user code in vfork() calls
754 * libc's getpid() or some other function which then decides
755 * it wants the upmap).
756 */
757 p = ba->aux_info;
758 if (p == NULL)
759 break;
760 if (p->p_flags & P_PPWAIT) {
761 p = p->p_pptr;
762 if (p == NULL)
763 return (EINVAL);
764 invfork = 1;
765 } else {
766 invfork = 0;
767 }
768
769 /*
770 * Create the kernel structure as required, set the invfork
771 * flag if we are faulting in on a vfork().
772 */
773 if (p->p_upmap == NULL)
774 proc_usermap(p, invfork);
775 if (p->p_upmap && invfork)
776 p->p_upmap->invfork = invfork;
777
778 /*
779 * Extract address for pmap
780 */
781 if (p->p_upmap &&
782 offset < roundup2(sizeof(*p->p_upmap), PAGE_SIZE)) {
783 /* only good for current process */
784 *resultp = pmap_kextract((vm_offset_t)p->p_upmap +
785 offset);
786 error = 0;
787 }
788 break;
789 case 6:
790 /*
791 * /dev/kpmap - maps RO shared kernel global page
792 *
793 * Extract address for pmap
794 */
795 if (kpmap &&
796 offset < roundup2(sizeof(*kpmap), PAGE_SIZE)) {
797 *resultp = pmap_kextract((vm_offset_t)kpmap + offset);
798 error = 0;
799 }
800 break;
801 case 7:
802 /*
803 * /dev/lpmap - maps RW per-thread shared user-kernel area.
804 */
805 lp = ba->aux_info;
806 if (lp == NULL)
807 break;
808
809 /*
810 * Create the kernel structure as required
811 */
812 if (lp->lwp_lpmap == NULL)
813 lwp_usermap(lp, -1); /* second arg not yet XXX */
814
815 /*
816 * Extract address for pmap
817 */
818 if (lp->lwp_lpmap &&
819 offset < roundup2(sizeof(*lp->lwp_lpmap), PAGE_SIZE)) {
820 /* only good for current process */
821 *resultp = pmap_kextract((vm_offset_t)lp->lwp_lpmap +
822 offset);
823 error = 0;
824 }
825 break;
826 default:
827 break;
828 }
829 return error;
830 }
831
832 static void
mem_drvinit(void * unused)833 mem_drvinit(void *unused)
834 {
835
836 /* Initialise memory range handling */
837 if (mem_range_softc.mr_op != NULL)
838 mem_range_softc.mr_op->init(&mem_range_softc);
839
840 make_dev(&mem_ops_mem, 0, UID_ROOT, GID_KMEM, 0640, "mem");
841 make_dev(&mem_ops_mem, 1, UID_ROOT, GID_KMEM, 0640, "kmem");
842 make_dev(&mem_ops, 2, UID_ROOT, GID_WHEEL, 0666, "null");
843 make_dev(&mem_ops, 3, UID_ROOT, GID_WHEEL, 0644, "random");
844 make_dev(&mem_ops, 4, UID_ROOT, GID_WHEEL, 0644, "urandom");
845 make_dev(&mem_ops, 5, UID_ROOT, GID_WHEEL, 0666, "upmap");
846 make_dev(&mem_ops, 6, UID_ROOT, GID_WHEEL, 0444, "kpmap");
847 make_dev(&mem_ops, 7, UID_ROOT, GID_WHEEL, 0666, "lpmap");
848 zerodev = make_dev(&mem_ops, 12, UID_ROOT, GID_WHEEL, 0666, "zero");
849 make_dev(&mem_ops_noq, 14, UID_ROOT, GID_WHEEL, 0600, "io");
850 }
851
852 SYSINIT(memdev, SI_SUB_DRIVERS, SI_ORDER_MIDDLE + CDEV_MAJOR, mem_drvinit,
853 NULL);
854
855