2 * Copyright (c) 1988 University of Utah.
3 * Copyright (c) 1982, 1986, 1990 The Regents of the University of California.
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.
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
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.
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
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 $
44 #include <sys/param.h>
45 #include <sys/systm.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>
56 #include <sys/random.h>
57 #include <sys/signalvar.h>
59 #include <sys/vnode.h>
60 #include <sys/sysctl.h>
62 #include <sys/signal2.h>
66 #include <vm/vm_extern.h>
69 static d_open_t mmopen
;
70 static d_close_t mmclose
;
71 static d_read_t mmread
;
72 static d_write_t mmwrite
;
73 static d_ioctl_t mmioctl
;
75 static d_mmap_t memmmap
;
77 static d_kqfilter_t mmkqfilter
;
78 static int memuksmap(cdev_t dev
, vm_page_t fake
);
81 static struct dev_ops mem_ops
= {
82 { "mem", 0, D_MPSAFE
| D_QUICK
},
88 .d_kqfilter
= mmkqfilter
,
95 static struct dev_ops mem_ops_mem
= {
96 { "mem", 0, D_MEM
| D_MPSAFE
| D_QUICK
},
102 .d_kqfilter
= mmkqfilter
,
106 .d_uksmap
= memuksmap
109 static struct dev_ops mem_ops_noq
= {
110 { "mem", 0, D_MPSAFE
},
116 .d_kqfilter
= mmkqfilter
,
120 .d_uksmap
= memuksmap
123 static int rand_bolt
;
125 static cdev_t zerodev
= NULL
;
126 static struct lock mem_lock
= LOCK_INITIALIZER("memlk", 0, 0);
128 MALLOC_DEFINE(M_MEMDESC
, "memdesc", "memory range descriptors");
129 static int mem_ioctl (cdev_t
, u_long
, caddr_t
, int, struct ucred
*);
130 static int random_ioctl (cdev_t
, u_long
, caddr_t
, int, struct ucred
*);
132 struct mem_range_softc mem_range_softc
;
134 static int seedenable
;
135 SYSCTL_INT(_kern
, OID_AUTO
, seedenable
, CTLFLAG_RW
, &seedenable
, 0, "");
138 mmopen(struct dev_open_args
*ap
)
140 cdev_t dev
= ap
->a_head
.a_dev
;
143 switch (minor(dev
)) {
147 * /dev/mem and /dev/kmem
149 if (ap
->a_oflags
& FWRITE
) {
150 if (securelevel
> 0 || kernel_mem_readonly
)
157 * /dev/kpmap can only be opened for reading.
159 if (ap
->a_oflags
& FWRITE
)
164 error
= priv_check_cred(ap
->a_cred
, PRIV_ROOT
, 0);
167 if (securelevel
> 0 || kernel_mem_readonly
) {
171 error
= cpu_set_iopl();
181 mmclose(struct dev_close_args
*ap
)
183 cdev_t dev
= ap
->a_head
.a_dev
;
186 switch (minor(dev
)) {
188 error
= cpu_clr_iopl();
199 mmrw(cdev_t dev
, struct uio
*uio
, int flags
)
209 while (uio
->uio_resid
> 0 && error
== 0) {
211 if (iov
->iov_len
== 0) {
214 if (uio
->uio_iovcnt
< 0)
218 switch (minor(dev
)) {
221 * minor device 0 is physical memory, /dev/mem
224 v
&= ~(long)PAGE_MASK
;
225 pmap_kenter((vm_offset_t
)ptvmmap
, v
);
226 o
= (int)uio
->uio_offset
& PAGE_MASK
;
227 c
= (u_int
)(PAGE_SIZE
- ((uintptr_t)iov
->iov_base
& PAGE_MASK
));
228 c
= min(c
, (u_int
)(PAGE_SIZE
- o
));
229 c
= min(c
, (u_int
)iov
->iov_len
);
230 error
= uiomove((caddr_t
)&ptvmmap
[o
], (int)c
, uio
);
231 pmap_kremove((vm_offset_t
)ptvmmap
);
236 * minor device 1 is kernel memory, /dev/kmem
238 vm_offset_t saddr
, eaddr
;
244 * Make sure that all of the pages are currently
245 * resident so that we don't create any zero-fill
248 saddr
= trunc_page(uio
->uio_offset
);
249 eaddr
= round_page(uio
->uio_offset
+ c
);
254 * Make sure the kernel addresses are mapped.
255 * platform_direct_mapped() can be used to bypass
256 * default mapping via the page table (virtual kernels
257 * contain a lot of out-of-band data).
260 if (uio
->uio_rw
!= UIO_READ
)
261 prot
|= VM_PROT_WRITE
;
262 error
= kvm_access_check(saddr
, eaddr
, prot
);
265 error
= uiomove((caddr_t
)(vm_offset_t
)uio
->uio_offset
,
271 * minor device 2 (/dev/null) is EOF/RATHOLE
273 if (uio
->uio_rw
== UIO_READ
)
279 * minor device 3 (/dev/random) is source of filth
280 * on read, seeder on write
283 buf
= kmalloc(PAGE_SIZE
, M_TEMP
, M_WAITOK
);
284 c
= min(iov
->iov_len
, PAGE_SIZE
);
285 if (uio
->uio_rw
== UIO_WRITE
) {
286 error
= uiomove(buf
, (int)c
, uio
);
290 error
= add_buffer_randomness_src(buf
, c
, RAND_SRC_SEEDING
);
291 } else if (error
== 0) {
295 poolsize
= read_random(buf
, c
);
299 if ((flags
& IO_NDELAY
) != 0)
300 return (EWOULDBLOCK
);
303 c
= min(c
, poolsize
);
304 error
= uiomove(buf
, (int)c
, uio
);
309 * minor device 4 (/dev/urandom) is source of muck
310 * on read, writes are disallowed.
312 c
= min(iov
->iov_len
, PAGE_SIZE
);
313 if (uio
->uio_rw
== UIO_WRITE
) {
317 if (CURSIG(curthread
->td_lwp
) != 0) {
319 * Use tsleep() to get the error code right.
320 * It should return immediately.
322 error
= tsleep(&rand_bolt
, PCATCH
, "urand", 1);
323 if (error
!= 0 && error
!= EWOULDBLOCK
)
327 buf
= kmalloc(PAGE_SIZE
, M_TEMP
, M_WAITOK
);
328 poolsize
= read_random_unlimited(buf
, c
);
329 c
= min(c
, poolsize
);
330 error
= uiomove(buf
, (int)c
, uio
);
332 /* case 5: read/write not supported, mmap only */
333 /* case 6: read/write not supported, mmap only */
336 * minor device 12 (/dev/zero) is source of nulls
337 * on read, write are disallowed.
339 if (uio
->uio_rw
== UIO_WRITE
) {
344 zbuf
= (caddr_t
)kmalloc(PAGE_SIZE
, M_TEMP
,
347 c
= min(iov
->iov_len
, PAGE_SIZE
);
348 error
= uiomove(zbuf
, (int)c
, uio
);
355 iov
->iov_base
= (char *)iov
->iov_base
+ c
;
357 uio
->uio_offset
+= c
;
366 mmread(struct dev_read_args
*ap
)
368 return(mmrw(ap
->a_head
.a_dev
, ap
->a_uio
, ap
->a_ioflag
));
372 mmwrite(struct dev_write_args
*ap
)
374 return(mmrw(ap
->a_head
.a_dev
, ap
->a_uio
, ap
->a_ioflag
));
377 /*******************************************************\
378 * allow user processes to MMAP some memory sections *
379 * instead of going through read/write *
380 \*******************************************************/
382 static int user_kernel_mapping(int num
, vm_ooffset_t offset
,
383 vm_ooffset_t
*resultp
);
388 memmmap(struct dev_mmap_args
*ap
)
390 cdev_t dev
= ap
->a_head
.a_dev
;
394 switch (minor(dev
)) {
397 * minor device 0 is physical memory
399 ap
->a_result
= atop(ap
->a_offset
);
404 * minor device 1 is kernel memory
406 ap
->a_result
= atop(vtophys(ap
->a_offset
));
412 * minor device 5 is /dev/upmap (see sys/upmap.h)
413 * minor device 6 is /dev/kpmap (see sys/upmap.h)
416 error
= user_kernel_mapping(minor(dev
), ap
->a_offset
, &result
);
417 ap
->a_result
= atop(result
);
429 memuksmap(cdev_t dev
, vm_page_t fake
)
434 switch (minor(dev
)) {
437 * minor device 0 is physical memory
439 fake
->phys_addr
= ptoa(fake
->pindex
);
444 * minor device 1 is kernel memory
446 fake
->phys_addr
= vtophys(ptoa(fake
->pindex
));
452 * minor device 5 is /dev/upmap (see sys/upmap.h)
453 * minor device 6 is /dev/kpmap (see sys/upmap.h)
456 error
= user_kernel_mapping(minor(dev
),
457 ptoa(fake
->pindex
), &result
);
458 fake
->phys_addr
= result
;
468 mmioctl(struct dev_ioctl_args
*ap
)
470 cdev_t dev
= ap
->a_head
.a_dev
;
473 lockmgr(&mem_lock
, LK_EXCLUSIVE
);
475 switch (minor(dev
)) {
477 error
= mem_ioctl(dev
, ap
->a_cmd
, ap
->a_data
,
478 ap
->a_fflag
, ap
->a_cred
);
482 error
= random_ioctl(dev
, ap
->a_cmd
, ap
->a_data
,
483 ap
->a_fflag
, ap
->a_cred
);
490 lockmgr(&mem_lock
, LK_RELEASE
);
496 * Operations for changing memory attributes.
498 * This is basically just an ioctl shim for mem_range_attr_get
499 * and mem_range_attr_set.
502 mem_ioctl(cdev_t dev
, u_long cmd
, caddr_t data
, int flags
, struct ucred
*cred
)
505 struct mem_range_op
*mo
= (struct mem_range_op
*)data
;
506 struct mem_range_desc
*md
;
508 /* is this for us? */
509 if ((cmd
!= MEMRANGE_GET
) &&
510 (cmd
!= MEMRANGE_SET
))
513 /* any chance we can handle this? */
514 if (mem_range_softc
.mr_op
== NULL
)
517 /* do we have any descriptors? */
518 if (mem_range_softc
.mr_ndesc
== 0)
523 nd
= imin(mo
->mo_arg
[0], mem_range_softc
.mr_ndesc
);
525 md
= (struct mem_range_desc
*)
526 kmalloc(nd
* sizeof(struct mem_range_desc
),
527 M_MEMDESC
, M_WAITOK
);
528 error
= mem_range_attr_get(md
, &nd
);
530 error
= copyout(md
, mo
->mo_desc
,
531 nd
* sizeof(struct mem_range_desc
));
532 kfree(md
, M_MEMDESC
);
534 nd
= mem_range_softc
.mr_ndesc
;
540 md
= (struct mem_range_desc
*)kmalloc(sizeof(struct mem_range_desc
),
541 M_MEMDESC
, M_WAITOK
);
542 error
= copyin(mo
->mo_desc
, md
, sizeof(struct mem_range_desc
));
543 /* clamp description string */
544 md
->mr_owner
[sizeof(md
->mr_owner
) - 1] = 0;
546 error
= mem_range_attr_set(md
, &mo
->mo_arg
[0]);
547 kfree(md
, M_MEMDESC
);
554 * Implementation-neutral, kernel-callable functions for manipulating
555 * memory range attributes.
558 mem_range_attr_get(struct mem_range_desc
*mrd
, int *arg
)
560 /* can we handle this? */
561 if (mem_range_softc
.mr_op
== NULL
)
565 *arg
= mem_range_softc
.mr_ndesc
;
567 bcopy(mem_range_softc
.mr_desc
, mrd
, (*arg
) * sizeof(struct mem_range_desc
));
573 mem_range_attr_set(struct mem_range_desc
*mrd
, int *arg
)
575 /* can we handle this? */
576 if (mem_range_softc
.mr_op
== NULL
)
579 return (mem_range_softc
.mr_op
->set(&mem_range_softc
, mrd
, arg
));
583 mem_range_AP_init(void)
585 if (mem_range_softc
.mr_op
&& mem_range_softc
.mr_op
->initAP
)
586 mem_range_softc
.mr_op
->initAP(&mem_range_softc
);
590 random_ioctl(cdev_t dev
, u_long cmd
, caddr_t data
, int flags
, struct ucred
*cred
)
596 * Even inspecting the state is privileged, since it gives a hint
597 * about how easily the randomness might be guessed.
602 /* Really handled in upper layer */
606 intr
= *(int16_t *)data
;
607 if ((error
= priv_check_cred(cred
, PRIV_ROOT
, 0)) != 0)
609 if (intr
< 0 || intr
>= MAX_INTS
)
611 register_randintr(intr
);
614 intr
= *(int16_t *)data
;
615 if ((error
= priv_check_cred(cred
, PRIV_ROOT
, 0)) != 0)
617 if (intr
< 0 || intr
>= MAX_INTS
)
619 unregister_randintr(intr
);
625 intr
= *(int16_t *)data
;
626 if ((error
= priv_check_cred(cred
, PRIV_ROOT
, 0)) != 0)
628 if (intr
< 0 || intr
>= MAX_INTS
)
630 intr
= next_registered_randintr(intr
);
631 if (intr
== MAX_INTS
)
633 *(u_int16_t
*)data
= intr
;
643 mm_filter_read(struct knote
*kn
, long hint
)
649 mm_filter_write(struct knote
*kn
, long hint
)
655 dummy_filter_detach(struct knote
*kn
) {}
657 /* Implemented in kern_nrandom.c */
658 static struct filterops random_read_filtops
=
659 { FILTEROP_ISFD
|FILTEROP_MPSAFE
, NULL
, dummy_filter_detach
, random_filter_read
};
661 static struct filterops mm_read_filtops
=
662 { FILTEROP_ISFD
|FILTEROP_MPSAFE
, NULL
, dummy_filter_detach
, mm_filter_read
};
664 static struct filterops mm_write_filtops
=
665 { FILTEROP_ISFD
|FILTEROP_MPSAFE
, NULL
, dummy_filter_detach
, mm_filter_write
};
668 mmkqfilter(struct dev_kqfilter_args
*ap
)
670 struct knote
*kn
= ap
->a_kn
;
671 cdev_t dev
= ap
->a_head
.a_dev
;
674 switch (kn
->kn_filter
) {
676 switch (minor(dev
)) {
678 kn
->kn_fop
= &random_read_filtops
;
681 kn
->kn_fop
= &mm_read_filtops
;
686 kn
->kn_fop
= &mm_write_filtops
;
689 ap
->a_result
= EOPNOTSUPP
;
697 iszerodev(cdev_t dev
)
699 return (zerodev
== dev
);
703 * /dev/upmap and /dev/kpmap.
706 user_kernel_mapping(int num
, vm_ooffset_t offset
, vm_ooffset_t
*resultp
)
712 if ((p
= curproc
) == NULL
)
716 * If this is a child currently in vfork the pmap is shared with
717 * the parent! We need to actually set-up the parent's p_upmap,
718 * not the child's, and we need to set the invfork flag. Userland
719 * will probably adjust its static state so it must be consistent
720 * with the parent or userland will be really badly confused.
722 * (this situation can happen when user code in vfork() calls
723 * libc's getpid() or some other function which then decides
724 * it wants the upmap).
726 if (p
->p_flags
& P_PPWAIT
) {
740 * /dev/upmap - maps RW per-process shared user-kernel area.
742 if (p
->p_upmap
== NULL
)
743 proc_usermap(p
, invfork
);
745 p
->p_upmap
->invfork
= invfork
;
748 offset
< roundup2(sizeof(*p
->p_upmap
), PAGE_SIZE
)) {
749 /* only good for current process */
750 *resultp
= pmap_kextract((vm_offset_t
)p
->p_upmap
+
757 * /dev/kpmap - maps RO shared kernel global page
760 offset
< roundup2(sizeof(*kpmap
), PAGE_SIZE
)) {
761 *resultp
= pmap_kextract((vm_offset_t
)kpmap
+
773 mem_drvinit(void *unused
)
776 /* Initialise memory range handling */
777 if (mem_range_softc
.mr_op
!= NULL
)
778 mem_range_softc
.mr_op
->init(&mem_range_softc
);
780 make_dev(&mem_ops_mem
, 0, UID_ROOT
, GID_KMEM
, 0640, "mem");
781 make_dev(&mem_ops_mem
, 1, UID_ROOT
, GID_KMEM
, 0640, "kmem");
782 make_dev(&mem_ops
, 2, UID_ROOT
, GID_WHEEL
, 0666, "null");
783 make_dev(&mem_ops
, 3, UID_ROOT
, GID_WHEEL
, 0644, "random");
784 make_dev(&mem_ops
, 4, UID_ROOT
, GID_WHEEL
, 0644, "urandom");
785 make_dev(&mem_ops
, 5, UID_ROOT
, GID_WHEEL
, 0666, "upmap");
786 make_dev(&mem_ops
, 6, UID_ROOT
, GID_WHEEL
, 0444, "kpmap");
787 zerodev
= make_dev(&mem_ops
, 12, UID_ROOT
, GID_WHEEL
, 0666, "zero");
788 make_dev(&mem_ops_noq
, 14, UID_ROOT
, GID_WHEEL
, 0600, "io");
791 SYSINIT(memdev
, SI_SUB_DRIVERS
, SI_ORDER_MIDDLE
+ CDEV_MAJOR
, mem_drvinit
,